<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Robert D. Brown III on Medium]]></title>
        <description><![CDATA[Stories by Robert D. Brown III on Medium]]></description>
        <link>https://medium.com/@robertdbrowniii?source=rss-f345147f5249------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*qSOzxavmHzo9QKNlK_MhHQ.jpeg</url>
            <title>Stories by Robert D. Brown III on Medium</title>
            <link>https://medium.com/@robertdbrowniii?source=rss-f345147f5249------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 06:01:36 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@robertdbrowniii/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Big Data Analytics: Avoiding the Hype]]></title>
            <link>https://medium.com/@robertdbrowniii/big-data-analytics-avoiding-the-hype-83da40a87087?source=rss-f345147f5249------2</link>
            <guid isPermaLink="false">https://medium.com/p/83da40a87087</guid>
            <dc:creator><![CDATA[Robert D. Brown III]]></dc:creator>
            <pubDate>Sat, 10 Aug 2024 15:44:15 GMT</pubDate>
            <atom:updated>2024-08-10T15:44:15.729Z</atom:updated>
            <content:encoded><![CDATA[<h3>Decision quality is the pathway to avoid the hype of big #data #analytics.</h3><p>(I originally published this post on <a href="https://www.linkedin.com/pulse/big-data-analytics-avoiding-hype-robert-d-brown-iii?utm_source=share&amp;utm_medium=member_ios&amp;utm_campaign=share_via">LinkedIn</a> October 29, 2017. I have made slight editing changes from the original here.)</p><p>One evening this past week I received an email from a friend in the leadership of a marketing department in a large corporation. She made a simple request and asked a simple question: “Rob, talk to me about big data analytics. What do I need to know?” What followed was a kind of Jerry Maguire screed fueled by an espresso bender pulled too late in the evening (as I sometimes do, but I swear I’m going to stop any day now). I’m sharing a slightly edited version of my response here.</p><p>Obviously, “big data analytics” is a big field, so I’m not quite sure what you might be wanting me to talk about, but I can tell you what my concerns are that I think any company that is investing in big data analytics initiatives needs to consider. This has actually become a mission for me over the past four years or more (ed. note: now more than ten years).</p><p>To start, let me be very clear that I think what is going on in this field is great. I really do. It’s fascinating to me as it is to many others, and rightly so. There are a lot of exciting developments emerging in terms of computing resources to the actual math behind the analytics pieces of AI and machine learning to the compelling graphical display of quantitative information to the real benefits that are being obtained. A lot of the math has been around for a long time already, but especially in areas of AI, some of that old math is maturing rapidly. But some big data problems are showing up big.</p><h3>There’s Gold in Them Thar Hills — Sell Shovels, Too</h3><p>First, there is a huge gold rush to find value with big data analytics. While it’s true that a lot within the field is maturing, there’s still a lot of hype about the value and what we can obtain with it. There seems to be a tacit belief that data will solve all our problems, and you will see this belief being shared with catchphrases like “we should be a data-driven company,” or “we believe in data-driven decision making.” The idea being that decision making by humans is inherently biased, and only the cold hard facts of data should be trusted to support unbiased decision making. I get that. I don’t like bias, either, and I think we should have less of it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/585/1*t6thkNMGZGTNo-bpjE34Tw@2x.jpeg" /></figure><p>(<em>http://upload.wikimedia.org/wikipedia/en/0/0a/Georgia_Gold_Rush_Prospectors.jpg</em>)</p><p>Unfortunately, people seem to be forgetting that data is not free of impressions from people’s values and biases. All data that we have access to is collected and distilled according to human decision making, which, as we’ve all conceded, is colored by bias. These biases affect all aspects of the acquisition of data: what, how, and when it’s taken; and which data is discarded due to ideas about what constitutes non-informative data (i.e., outliers). Data is not free of bias. It is not values-free. We need to do a better job of mining data in as unbiased a manner as possible, using tools and processes that identify, account for, and make transparent the sources of bias that infiltrate data. We need better shovels.</p><h3>Be Values-Driven, Data-Supported</h3><p>Second, mankind does not live by data alone, meaning that good decisions (the meaningful thing that mature humans do) are not made on the basis of information or data only. In fact, technically, a good decision rests on three legs: values and preferences, decision alternatives, and information. If any of those components are missing from the framing of a problem to be addressed by decision making, as all problems are, one cannot say that they have made a good decision. People seem to believe that data will tell them what they ought to do, but it can’t because even though data is not free of values, it does not possess the volition driven by values/preferences. While data is tainted by values, it does not possess a values frame. Only humans (or sentient agents, let’s say) have volition and a values frame. We only make decisions because we want to achieve some goal that we value, and those goals will most likely be different for each person.</p><p>For example, my wife and I often argue about the temperature in the house. We both have access to the same data (i.e., the temperature indicated on the thermostat), but I like it warmer, and she likes it cooler. Decision making is driven by values and preferences (what we want to achieve and how much), but the alternatives we exercise are conditionally related to the information we obtain about how each alternative affects our ability to achieve our values at preferred levels. So, what I’m getting to is that “big data analytics” must be pursued with this three-legged stool in mind, that the ultimate reason any of us want analytics of any kind is to achieve that which we value, and what we value varies differentially across people and is inherently subjective.</p><p>https://youtu.be/NOalmQUnjoo</p><p><em>Temperature is a uniquely personal preference. Thermostats tell us data about the temperature, but they do not tell us what it ought to be.</em></p><h3>Decision Makers, Heal Thyself</h3><p>Third, the success rate of big data analytics initiatives is depressingly and ironically low. Why should that be? After all, data should solve all our problems, right? Big data analytics ought to tell us how to succeed in big data analytics initiatives. But it isn’t doing so, and we should be asking why. Quite a bit of <a href="https://medium.com/@roilavan/the-dirty-little-secret-of-big-data-ai-analytics-platforms-they-cannot-solve-all-of-your-81b0f2c3a4b4">reporting</a> about this reveals that the reasons for initiative failures are due to multiple causes, the first order level being that the data being used is just low quality. One of the seven or so quality metrics for data to be used in data analytics initiatives is that it must be relevant to the problem at hand. Data relevancy is the most important quality, but confirming the relevance of data is routinely violated. (Someday I’ll have to tell you the story about a client’s marketing group that was using a regression model based on data that was, at least, remotely relevant to the problem they were applying their model. I guess it felt nice to be “data-driven.” ) Relevance is the key word here because it also elevates to the second-order level of why data analytics initiatives fail, and that is the initiatives themselves are often not really relevant to the concerns of the enterprise. Or to put it even more bluntly, the data analytics initiatives that fail were failures from their start.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/585/1*l8fiQQsDehl8UllU2TiVVw@2x.jpeg" /></figure><p>(<em>http://ufacts.blogspot.com/2015/05/first-surgeon-ever-in-history-to.html</em>)</p><p>I’ve become convinced that ultimately the reason many data analytics initiatives fail is directly related to the same reasons all other types of projects fail, and that is they are not…</p><ul><li>selected and executed on the basis of relevance to the enterprise (i.e., the values and preferences);</li><li>prioritized on the basis of a thorough consideration of opportunity costs for alternate uses of the resources used to execute (i.e., a clear identification of alternatives);</li><li>supported by relevant and efficiently acquired data (i.e., information).</li></ul><p>When people don’t recognize that the three-legged stool must be in place and held together by decision quality governance to make good decisions about data analytics initiatives, they wind up selecting these initiatives because they think the solutions will provide an answer to interesting local technical problems, or because it’s the “great idea” of a sufficiently politically powerful person who drives the initiative forward, etc., etc., etc.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/585/1*POIGW8H8AvDMRUCRivd2cA@2x.jpeg" /></figure><p><a href="http://gph.is/1QfaJiv">http://gph.is/1QfaJiv</a></p><p>Most of the problems associated with data analytics initiatives can be solved by placing a decision quality wrapper around them. Until people realize that, though, big data analytics will, if the statistics hold up, waste valuable resources and time, erode competitive advantage, and defer relevant stakeholders from achieving what they value most.</p><p>My mission is to change that.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/536/1*zWms5ajlg90HNqM7hfjhPg@2x.jpeg" /></figure><blockquote>You had me at Decision Quality.</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=83da40a87087" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What Do You Expect?]]></title>
            <link>https://medium.com/@robertdbrowniii/what-do-you-expect-8e2a885509b6?source=rss-f345147f5249------2</link>
            <guid isPermaLink="false">https://medium.com/p/8e2a885509b6</guid>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[cyber-risk-quantification]]></category>
            <dc:creator><![CDATA[Robert D. Brown III]]></dc:creator>
            <pubDate>Fri, 03 May 2024 23:03:32 GMT</pubDate>
            <atom:updated>2024-05-04T13:36:03.943Z</atom:updated>
            <content:encoded><![CDATA[<h4>Why expected values provide a valuable way to support cybersecurity investment decisions and distinctions</h4><h4>Introduction</h4><p>We often hear that calculating expected values is not informative for cyber security investment decisions because they are difficult to understand and don’t necessarily represent values that can occur. As a result, decision-makers resort to less informative but easier to use heuristics to guide them in making decisions. In the following sections, I hope to clarify the value of using expected values, and then point to more informative, productive ways to manage risk with cybersecurity investments.</p><h4>Flipping Coins and Craps for Dollars</h4><p>Imagine a situation where I offer you the opportunity to win a prize by playing one of two games. In one game, I toss a fair coin. If the coin shows heads up, you win $10. If the coin shows tails up, you pay me $5. In the other game, I roll a fairly weighted six-sided die. If a six shows up, you win $18; otherwise, you pay me $4. Which game do you prefer to play?</p><p>Because there are several trade offs to consider, the choice that most likely works to your favor isn’t immediately obvious. The second game offers an 80% larger benefit than the first, but the probability of the first (50%) favors your winning something compared to that of the second game (⅙ = 16.7%). However, the second game presents a larger probability of losing (1 — ⅙ = 83.3%) compared to the second (50%), but the loss of the second is smaller compared to the first. How do you choose? This is where expected value enters to help us make a decision.</p><p>Expected value is just a compact way of saying “the total probability-weighted value of potential possibilities.” In our example above, each game presents only two possibilities: a win or a loss. Each possibility carries a probability that it can occur based on the structure of the game and reasonable assumptions about the fairness, or weighting, of that structure. You can reasonably think of the probabilities in this case as informed degrees of belief of the possible outcomes. It often helps to visualize the game possibilities and their probabilities for these two game options as a tree diagram:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/827/0*rUfQl2Jbavx-LdsZ" /></figure><p>To calculate the expected value (EV) of the two choices you now face, you multiply backwards through the branches then sum the products at the nodes. We call this process “rolling back the tree.” The formula and table that follows below illustrate this process.</p><pre>EV = (probability of benefit) x (value of benefit) + (probability of loss) x (value of loss)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BcPmjeJOlCqPEn332WHLMw.png" /></figure><h4>Do NOT Expect the Expected Value!</h4><p>Let’s unpack what these results mean because the outcomes in this case (I’ll explain the exception to this later) imply something different than what anyone today would typically infer from the term “expected.” The calculated outcomes are <strong>not a prediction of what will occur</strong> in the choice of either game. Under no circumstances when you play either game will you actually gain $2.50 (game 1) or pay out $0.33 (game 2). <strong>Nor do expected values tell us anticipated values as a reasonable approximation of what to plan on happening.</strong> “Expected” does not mean “anticipated.” As stated earlier, <em>expected value</em> is a statistical term that is the total probability-weighted value of potential possibilities. Those two numbers represent all of the risk information (i.e., probability and impact) about either game in a single number that allows us to make a rational preference distinction between the two–as long as we’re risk neutral. While they do not represent real outcomes, they do represent informative direction. Now, we need to unpack what we mean by “rational” and “risk neutral.”</p><h4>On Being Rational and Risk Neutral</h4><p>I’m going to state several assumptions about you, and if they hold, for the purposes of this discussion I will consider you to be rational and risk neutral. The first assumption is that you prefer more money to less, or more generally, that you prefer more of what you value than less. Second, you understand the probabilities and possibilities involved in the game, and your potential loss from playing either game would easily come from your wallet or a quick Venmo transfer. In other words, you comprehend the situation and accept the results unflinchingly. Finally, if I were to present to you the opportunity to take a certain $2.50 from me or play the first game, you would be indifferent to either because you understand that they are essentially probabilistic proxies for each other. We could say they have the same utility to you; thus, receiving $2.50 now versus an equal probability to win $10 or lose $5 is approximately the same to you. The combined effect of these assumptions is that you can consider expected values to be a proxy for actual values, and that you prefer games of chance (or decisions under uncertainty) that yield higher expected values to lower ones[1].</p><p>The informative direction of our expected value calculation above is that you should take game one over game two, not because game one will put $2.50 in your pocket versus game two extracting $0.33 from it (remember, it won’t do either), but because by the definition of rationality above, game one presents a better choice considering not only the game structure but your own value preferences and risk tolerance.</p><p>The above description only applies to situations in which outcomes are discrete with a gap between possible event values. If we were presented with another situation in which the possibilities of an event and their probabilities both vary across a continuous range, the expected value would fall within the event’s range. In cases like that, the expected value can be taken as a reasonable approximation of what you can anticipate. Regardless, it is still true that the expected value of this continuous range is the probability-weighted value of potential possibilities. As we will see in the example that follows, understanding the expected value depends entirely on the context of the situation of how it is calculated and communicated.</p><p>All of this is great, considering toy games like the one above, but how do we use this information in real world situations? In the next section, we’ll elevate our consideration to something more consistent with security investment decisions, but the same observations will still apply:</p><ol><li>Expected values do not (necessarily) tell us what to anticipate.</li><li>Expected values allow us to make rational distinctions (as long as we’re risk neutral) of preferences between mutually exclusive games of chance or decisions under uncertainty.</li></ol><h4>The Security Game of Chance Is Like Flipping Coins</h4><p>We can think about cybersecurity similarly to the game examples discussed above in the following way. If we think about the upcoming year, there is a probability that your organization will face some financial losses due to cyber threats materializing or not. It’s as if the cybersecurity gods roll a many-sided die with one side that says “Losses” (with a probability of 1/N, where N = number of dice sides), and the die’s other sides say “No Losses” (with a probability of 1–1/N). The game looks like this tree.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/452/0*U6WKKWhthAwXv0B_" /></figure><p>Unlike the game example above, the Financial Losses will not be predetermined, discrete values. Instead, they will be uncertain, but you can think of that uncertainty as a distribution of losses until the actual losses are tallied up.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/548/0*1oX12XQvnIxiFK8U" /></figure><p>Over the last three hundred years, the mathematics of uncertainty have provided us with several ways to collapse that distribution into a single expected value, either through calculus, Monte Carlo simulation, or rules about how to convert a continuous distribution into discrete values with associated probabilities[2]. To keep things simple, we’ll think about the latter approach, breaking up the distribution into adjoining sections.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/603/0*tZKOcVM0nYMh0Arz" /></figure><p>Given that we now have three distinct possibilities and their probabilities, we can calculate the expected value EV of the future potential losses given that future losses occur by doing the following roll-back:</p><pre>EV Financial Losses (given Losses occur) = (Prob Low x Low Value) + (Prob Interior x Interior Value) + (Prob High x High Value)</pre><p>Can you visualize how the tree might look?</p><p>In this case, although we don’t know the exact value of the losses, we assume that the losses actually occur; therefore, <strong>the expected value is a real number that can be used as a forecast.</strong> But you must remember that this does not take into account the $0 cases in which no losses occur. We have to think about <strong>the complete set of possibilities</strong>, including the no-loss $0 cases, to get the total expected value of cyber threats. Again, an event tree will help us visualize how the possibilities play out.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1003/0*GEyBETrRgxXu3vBK" /></figure><p>The final expected value calculation for cyber threat looks like this:</p><pre>EV Cyber Threat = (Prob No Losses x $0) + (Prob Loss Occurs x EV Financial Losses)<br><br># The first term after the equals sign is just $0, so the formula reduces to…<br><br>EV Cyber Threat = Prob Loss Occurs x EV Financial Losses</pre><p>Now we’re back to the same kind of expected value that we originally considered in which a gap exists between the values that can occur (assuming the smallest value in the distribution is still significantly higher than the No Losses case). The expected value calculated here does not tell you what you should anticipate paying in the upcoming year for financial losses. Rather, it tells you the probability-weighted value of all potential<em> possibilities, losses&gt;&gt;$0, and no losses = $0.</em></p><p>Similar to the game examples above, even though we now have a calculation for a value that won’t ever exist, that doesn’t mean that it isn’t a useful number. We can use the number to make rational preference distinctions among investment alternatives for security controls. Let’s consider how to do that.</p><p>Imagine now that you face a base probability <em>Pb</em> of <em>EV of Financial Losses</em> (i.e., the expected value of the distribution of losses given that losses occur) over the next three years. You are considering at least two investment strategies for security controls. The first strategy, with a cost <em>C1</em> results in a probability <em>P1</em> of losses over the same time frame; the second, <em>C2</em> and <em>P2</em>. Neither of these strategies reduces the <em>EV of Financial Losses</em> because once a material breach occurs the organization still faces exposure to the same kinds and levels of losses. The primary advantage presented by the strategies is represented by a tradeoff between the probability of loss and the cost to implement and maintain the controls.</p><p>We already know how to calculate the current base expected value:</p><pre>EVb = Pb x EV of Financial Losses</pre><p>The new expected value delivered by strategy one is</p><pre>EV1 = P1 x EV of Financial Losses</pre><p>The expected value of the <strong>avoided loss</strong> then is</p><pre>EV Avoided Loss1 = EVb — EV1</pre><p>If we repeat the previous calculations for strategy two, we get</p><pre>EV Avoided Loss2 = EVb — (P2 x EV of Financial Losses)<br><br>EV Avoided Loss2 = EVb — EV2</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/593/0*PbB950DQnYbUuq85" /></figure><p>Since we want to achieve the expected avoided losses with the highest “bang for our buck,” we can calculate the expected ROI for each strategy as</p><pre>ROI1 = (EV Avoided Loss1 — C1) / C1</pre><p>and</p><pre>ROI2 = (EV Avoided Loss2 — C2) / C2</pre><p>In other words, the ROI is the ratio of the net avoided loss to the cost to achieve the avoided loss. It’s a measure of the cost efficiency of our efforts to reduce the probability of loss, and hence the efficiency of our efforts to increase expected avoided loss. Rationally, we want more ROI rather than less. Therefore, we simply compare <em>ROI1</em> to <em>ROI2</em>, and select the strategy that gives the highest expected ROI.</p><h4>Conclusion</h4><p>We should think of expected values as a good first pass approximation that helps us prioritize security investment strategies when we need to get started focusing on what to do first or next. However, we should never constrain our attention to risk only on the expected values. We need to keep a very healthy eye on the things that happen in the tails of loss distributions. I like to say, “it’s not the first moment (the expected value) that only matters. It’s the second moment (the variance) that takes your breath away.” (That’s a cheap statistics joke!) Which is to say, there is much more to cyber risk management than merely understanding expected values. The surprises that catch our breath occur in the tails, which means that any fully mature approach to risk management must include preemptive risk mitigation plans, risk transfer, and contingency plans for when they occur. But a proper understanding and application of expected loss values will help you get started securing your organization in a rational and cost-efficient manner.</p><h4>Footnotes</h4><ol><li>The other two risk attitudes that you could exhibit are risk averse and risk seeking. In either of these cases, the probability weighting on the gain versus loss for you to be indifferent to receiving a certain equivalent is not equal to the structure of the game-weighted probabilities on gain versus loss. The concepts behind how to find the preferred probability given your risk attitude is beyond the intended scope of this document. Without further justification here, suffice it to say that if the loss you face in a gamble is within 15% of your total assets, you can reasonably behave as risk neutral.</li><li>One common method for discretizing distributions into ranges and probabilities is the Extended Swanson-Megill. It assumes that the Low Value, Interior Value, and High Value sections represent 10th, 50th (or median), and 90th percentile values, respectively. The Prob Low, Prob Interior, and Prob High use 30%, 40%, 30% weightings on the sections. Without going into the derivation for these weights, simply understand that they are derived in such a way that the mean and variance of the continuous distribution are preserved when those weights are used as simulation sampling weights. The expected value formula becomes…</li></ol><pre>EV Financial Losses (given Losses occur) = (0.3 x p10) + (0.4 x p50) + (0.3 x p90)</pre><h4>Acknowledgment</h4><p>Many thanks to <a href="https://medium.com/u/d7edd6a4f65d">Peyman Kor</a> for his invaluable feedback on this document. Please, visit <a href="https://medium.com/@peymankor">his Medium account</a> for more insights into applications of decision and risk analysis and decision intelligence.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8e2a885509b6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Business Case Analysis with R]]></title>
            <link>https://medium.com/@robertdbrowniii/business-case-analysis-with-r-9d9ba3d00c15?source=rss-f345147f5249------2</link>
            <guid isPermaLink="false">https://medium.com/p/9d9ba3d00c15</guid>
            <dc:creator><![CDATA[Robert D. Brown III]]></dc:creator>
            <pubDate>Sun, 31 Dec 2023 19:12:58 GMT</pubDate>
            <atom:updated>2023-12-31T19:12:58.930Z</atom:updated>
            <content:encoded><![CDATA[<p>(The following is reposted with permission from the original at <a href="https://blog.ephorie.de/"><strong>Learning Machines</strong></a><strong> </strong>on<strong> </strong><a href="https://blog.ephorie.de/business-case-analysis-with-r">January 14, 2020</a><strong>. </strong>Please be sure to check out Holger K. von Jouanne-Diedrich’s fantastic blog and <a href="https://www.youtube.com/user/vonjd">YouTube</a> channel about data, science, and learning machines.)</p><h3>Introduction</h3><p>While Excel remains the tool of choice among analysts for business case analysis, presumably because of its low barrier to entry and ease of use, it nonetheless continues to be the source of many vexing and disastrous <a href="http://panko.shidler.hawaii.edu/My%20Publications/Whatknow.htm">errors</a>. As the research from Panko and others have revealed, the source of these errors arise in part for the following reasons:</p><ul><li>Analysts continue in a persistent lack of QA standards and practices to ensure that spreadsheets are as error free as possible.</li><li>Cell referencing and replicating formulas across relevant dimensions propagate errors at exponential rates.</li><li>Effective auditing gets obscured because formulas’ conceptual meanings are difficult to interpret from formulas constructed from grid coordinates.</li></ul><p>To suggest a means to remedy that situation, in the spring of 2018, I published <a href="https://www.apress.com/us/book/9781484234945">Business Case Analysis with R: Simulation Tutorials to Support Complex Business Decisions</a> (<em>BCAwR</em>). I also wanted to showcase how the popular programming language <strong>R</strong> could be employed to evaluate business case opportunities that are fraught with uncertainty nor supported by an abundance of well structured, pedigreed, and relevant data, a situation that plagues just about all business decisions faced in the real world. <em>BCAwR </em>takes the reader through the process of analyzing a rather complex commercial investment problem of bringing a chemical processing plant on line. For my post here, I will present a simplified model to give a preview of the contents of the book. I will also provide a little surprise at the end that I hope will contribute to bridging the efforts of the decision and data science communities.</p><p>But before we delve into any coding or quantitative activity related to analysis of any business case, we ought to thoroughly frame our problem by narrowing the scope of our inquiry to the appropriate level for the context of the problem at hand, identifying the goals and objectives we want to achieve when we eventually commit to action, partitioning the set of actionable decision we can make to achieve our goals, and creating an inventory of the influences and relationships that connect decisions and uncertainties to the measures of success. Although we will look at a much more simplified business problem here, a little framing will still serve us well.</p><p>Imagine that we are considering investing in a project that solves a particular market problem with the ultimate desire to generate a financial return to our organization. We identify three possible strategic pathways to compare for the investment, each with a different level of capital commitment required to address certain nuances of the market problem. Each (1) investment, in turn, will generate (2) an initial post investment net cash flow, (3) a peak cash flow, and varying durations of (4) time to reach the peak from the initial cash flow.</p><p>To complicate the matter, for each strategy, we do not know the precise quantitative value of the four parameters that characterize the potential cash flow. For this reason, when we evaluate the three strategies, we will employ a Monte Carlo simulation approach that samples a large number of potential realizations for each parameter from probability distributions. With these ideas in mind, we can now set up the coding to handle the simulation problem.</p><pre>library(leonRdo) # https://www.incitedecisiontech.com/packages/packages.html<br>library(tidyverse) # (on CRAN)<br># Set the simulation seed.<br>set.seed(42)<br> <br># The number of simulation runs per uncertain variable.<br>trials &lt;- 1000<br> <br># The time frame across which our cash flow simulation will run.<br># The initial investment will occur in year 0.<br>year &lt;- 0:15<br> <br># The weighted average cost of capital without any additional risk premium.<br>discount.rate &lt;- 0.1<br> <br># The distinctive capital allocation strategic decisions.<br>strategy &lt;- c(&quot;Strategy 1&quot;, &quot;Strategy 2&quot;, &quot;Strategy 3&quot;)</pre><p><strong>R</strong> employs simple Monte Carlo techniques, which doesn’t typically yield stable means for result distributions until many thousands of trials are used. However, When developing a model like this, I suggest setting the trials to 50–100 to make sure that the logic is running correctly in short responsive time frames. Then, when you’re ready to produce real results, change the values to greater than 10,000. Given that Monte Carlo in <strong>R</strong> is noisy even for reasonably large samples (~1000) and for different seed settings, I recently developed the leonRdo package (<a href="https://www.incitedecisiontech.com/packages/packages.html">R Packages: leonRdo &amp; inteRest</a>) to provide a median Latin hypercube approach to sampling that produces much more stable means with approximately 1/10th the number of trials regardless of the seed value used.</p><p>In this particular showcase discussion, I have included the tidyversepackage (on CRAN) to help with data manipulation and graphing. In the original <em>BCAwR</em>, I based all the code on base <strong>R</strong> so that readers new to <strong>R</strong> and simulation could focus more on learning the key principles over the idiosyncratic syntax of secondary packages.</p><p>The set.seed() function ensures that we will produce the same sample trial set on each run of our model. This will help us debug problems as we develop code. Eventually as others might interact with our code, they will be able to observe the same set of results we work with. Later, we can reset the seed to other values to make sure that our inferences on a given set of trials remain consistent on a different trial set.</p><h3>Model Functions</h3><p>Next, we need to declare some important functions that we will call in our model. The first is a function that provides an analytic solution to the differential equation that relates the proportional absorption of an entity into a fixed environment. This ramp up function takes as parameters the amount of time Tp required to go from, say, Y = 1% absorption to Y = 99% across Time. We will use this function to model the growth of our cash flow over the year index.</p><pre># Ramp up function.<br>calcRampUp &lt;- function(Y0, Time, Tp) {<br>  # Y0 = initial absorption<br>  # Time = time index<br>  # tp = time to Y = 1 - Y0, the time to peak.<br>  Y &lt;- 1 / (1 + (Y0 / (1 - Y0)) ^ (2 * (Time - 1) / (Tp - 1) - 1))<br>  return(Y)<br>}</pre><p>The second function we need is the actual business model we want to represent. The business model logic will remain fixed across strategies, and the results will vary only on the basis of the strategy conditional parameters we supply. By “functionalizing” our model, we can iterate the entire model over variable ranges to understand the sensitivity of the output function to those variables, as we shall later see. Please note, this is a toy model we are using for illustration purposes only. <em>BCAwR</em> demonstrates a much more complex consideration for the implementation of a chemical processing plant.</p><pre># Business model cash flow.<br>calcCashFlow &lt;- function(I, SCF, PCF, YP, Y) {<br>  # I = initial investment<br>  # SCF = starting cash flow<br>  # PCF = peak cash flow<br>  # YP = year of peak cash flow<br>  # Y = year index<br>  cf &lt;- (Y == 0) * (-I) +<br>    (Y &gt; 0) * (SCF + (PCF - SCF) * calcRampUp(0.01, Y, YP))<br>  return(cf)<br>}</pre><p>The last function we will employ is the net present value (NPV) function that we will apply to the individual trial results of the cash flow simulation across strategies.</p><pre># The net present value of the cash flow function.<br>calcNPV &lt;- function(CF, Y, DR) {<br>  # CF = cash flow vector<br>  # Y = year index<br>  # DR = discount rate<br>  npv &lt;- sum(CF / (1 + DR) ^ Y)<br>  return(npv)<br>}</pre><h3>Initialize Simulation Trial Samples</h3><p>Our next task is to create trial samples for the business model cash flow function parameters for the three project strategy investments. For this particular example, I have chosen to use canonical distributions as if their parameters were based on empirical or historical data and to utilize a simple method for generating trials. However, typically I would use a combination of both empirical data and subject matter expert guidance reflected as cumulative probabilities across the range of the assessed variables’ potential values. I explain how to take this latter approach in greater detail in <strong>Section III</strong> of <em>BCAwR</em>.</p><pre>investment1 &lt;-<br>  rlnorm_mlhs(n = trials,<br>         meanlog = log(800),<br>         sdlog = log(1 + 200 / 800))<br>investment2 &lt;-<br>  rlnorm_mlhs(n = trials,<br>         meanlog = log(700),<br>         sdlog = log(1 + 50 / 700))<br>investment3 &lt;-<br>  rlnorm_mlhs(n = trials,<br>         meanlog = log(1000),<br>         sdlog = log(1 + 150 / 1000))<br> <br>start.cash.flow1 &lt;- rnorm_mlhs(n = trials, mean = -100, sd = 20)<br>start.cash.flow2 &lt;- rnorm_mlhs(n = trials, mean = -90, sd = 5)<br>start.cash.flow3 &lt;- rnorm_mlhs(n = trials, mean = -120, sd = 15)<br> <br>peak.cash.flow1 &lt;- rnorm_mlhs(n = trials, mean = 300, sd = 20)<br>peak.cash.flow2 &lt;- rnorm_mlhs(n = trials, mean = 280, sd = 5)<br>peak.cash.flow3 &lt;- rnorm_mlhs(n = trials, mean = 375, sd = 15)<br> <br>yr.peak1 &lt;- rnorm_mlhs(n = trials, mean = 8.5, sd = 0.75)<br>yr.peak2 &lt;- rnorm_mlhs(n = trials, mean = 10, sd = 0.85)<br>yr.peak3 &lt;- rnorm_mlhs(n = trials, mean = 11, sd = 1)<br> <br># Store the business model parameter samples in a list of data frames.<br>proj.data &lt;-<br>  list(<br>    investment = data.frame(investment1, investment2, investment3),<br>    start.cash.flow = data.frame(start.cash.flow1, start.cash.flow2, start.cash.flow3),<br>    peak.cash.flow = data.frame(peak.cash.flow1, peak.cash.flow2, peak.cash.flow3),<br>    yr.peak = data.frame(yr.peak1, yr.peak2, yr.peak3)<br>  )</pre><h3>Model Results</h3><p>For each strategy, we apply the samples from each parameter to the business model cash flow function. This will result in a list of cash flows for the three project strategies. For each strategy, there will be as many cash flows as defined by trials, and each cash flow trial will be as long as the year index. We can use the lapply() and sapply() functions to avoid using for loops [see also <a href="https://blog.ephorie.de/learning-r-a-gentle-introduction-to-higher-order-functions">Learning R: A gentle introduction to higher-order functions</a>].</p><pre>proj.cf &lt;- lapply(1:length(strategy), function(s) {<br>  sapply(1:trials, function(t) {<br>    calcCashFlow(<br>      I = proj.data$investment[[s]][t],<br>      SCF = proj.data$start.cash.flow[[s]][t],<br>      PCF = proj.data$peak.cash.flow[[s]][t],<br>      YP = proj.data$yr.peak[[s]][t],<br>      Y = year<br>    )<br>  })<br>})<br>names(proj.cf) &lt;- strategy</pre><p>By running the first five trials of Strategy 1, we can see what the cash flows look like.</p><pre>head(round(proj.cf[[1]][, 1:5], digits = 1), n=length(year))<br>##         [,1]   [,2]   [,3]   [,4]   [,5]<br>##  [1,] -852.7 -852.2 -766.7 -974.3 -912.6<br>##  [2,]  -85.8 -103.1 -117.2 -116.1  -88.1<br>##  [3,]  -77.8  -93.6 -107.3 -106.9  -79.9<br>##  [4,]  -55.3  -64.3  -76.7  -79.9  -57.1<br>##  [5,]    0.0    9.4    0.2  -13.1   -1.6<br>##  [6,]   98.0  128.7  123.7  102.2   97.9<br>##  [7,]  201.5  230.2  227.2  215.6  206.8<br>##  [8,]  265.3  279.2  276.2  279.8  276.9<br>##  [9,]  292.5  296.5  293.3  305.3  308.0<br>## [10,]  302.3  301.8  298.5  314.1  319.5<br>## [11,]  305.6  303.4  300.1  316.9  323.5<br>## [12,]  306.7  303.9  300.5  317.8  324.8<br>## [13,]  307.1  304.1  300.7  318.1  325.2<br>## [14,]  307.2  304.1  300.7  318.2  325.4<br>## [15,]  307.2  304.1  300.7  318.2  325.4<br>## [16,]  307.2  304.1  300.7  318.2  325.5</pre><p>We can calculate some summary results for the cash flows for each strategy and plot them. First, we might like to plot the mean cash flow over time.</p><pre>proj.cf.mean &lt;- as.data.frame(lapply(strategy, function(s) {<br>  rowMeans(proj.cf[[s]])<br>}))<br>names(proj.cf.mean) &lt;- strategy<br>proj.cf.mean &lt;- cbind(year, proj.cf.mean)<br> <br># Plot the mean strategy cash flows.<br>gg.mean.cf &lt;-<br>  ggplot(gather(proj.cf.mean, &quot;strategy&quot;, &quot;mean&quot;, -year)) +<br>  geom_line(aes(x = year, y = mean, color = strategy)) +<br>  geom_point(aes(x = year, y = mean, color = strategy)) +<br>  labs(title = &quot;Mean Strategy Cash Flow&quot;,<br>       y = &quot;[$M]&quot;)<br>plot(gg.mean.cf)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/840/0*WUwI_21n3U0o92Gi.png" /></figure><p><em>Figure 1: The mean cash flows over time show tradeoffs in initial outlays, times to peak, level of peak, and the timing of breakeven.</em></p><p>Then we can calculate the cumulative mean cash flow and plot that for each strategy, too.</p><pre>proj.ccf.mean &lt;- proj.cf.mean %&gt;%<br>  mutate_at(vars(strategy), list(~ cumsum(.)))<br> <br># Plot the cumulative mean strategy cash flows.<br>gg.mean.ccf &lt;-<br>  ggplot(gather(proj.ccf.mean, &quot;strategy&quot;, &quot;mean&quot;, -year)) +<br>  geom_line(aes(x = year, y = mean, color = strategy)) +<br>  geom_point(aes(x = year, y = mean, color = strategy)) +<br>  labs(title = &quot;Cumulative Mean Strategy Cash Flow&quot;,<br>       y = &quot;[$M]&quot;)<br>plot(gg.mean.ccf)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/840/0*OIyWFOnO8kJMrdwH.png" /></figure><p><em>Figure 2: The cumulative mean cash flows over time show tradeoffs in drawdown, nominal payback timing, and terminal cash recovery level.</em></p><p>Now we can observe the risk profile of the cash flows by calculating the trial NPVs of the cash flows.</p><pre>proj.npv &lt;- as.data.frame(lapply(1:length(strategy), function(s) {<br>  sapply(1:trials, function(t) {<br>    calcNPV(CF = proj.cf[[s]][, t],<br>            Y = year,<br>            DR = discount.rate)<br>  })<br>}))<br>names(proj.npv) &lt;- strategy<br> <br># Plot the CDF of the strategies&#39; sample NPVs.<br>gg.ecdf &lt;-<br>  ggplot(gather(proj.npv, &quot;strategy&quot;, &quot;NPV&quot;), aes(x = NPV, color = strategy)) +<br>  stat_ecdf(geom = &quot;point&quot;) +<br>  labs(title = &quot;Strategy NPV Risk Profile&quot;,<br>       x = &quot;Strategy NPV [$M]&quot;,<br>       y = &quot;Cumulative Probability&quot;)<br>plot(gg.ecdf)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/840/0*sgToa5u5uQKGSvRU.png" /></figure><p><em>Figure 3: The cumulative probability chart shows tradeoffs in stategy dominance, range of value, and relative value volatility.</em></p><p>And we can calculate the mean NPV of the cash flows.</p><pre>proj.npv.mean &lt;- round(colMeans(proj.npv), digits = 1)<br>print(proj.npv.mean)<br>## Strategy 1 Strategy 2 Strategy 3 <br>##      157.4       59.3     -123.6</pre><p>The first observation we make about the risk profiles of the strategies is that we can dismiss Strategy 3 immediately because it presents negative mean NPV, the probability that it produces a negative economic value is ~75% (i.e., the probability(NPV&lt;=0) = 0.75), and practically none of its trials present any opportunity for dominance over any other strategy.</p><p>When we observe the relative volatility and dominance of the remaining strategies, we realize that we face a bit of ambiguity about how to choose the best pathway forward. Strategy 1 exhibits the best overall mean NPV, but it does so with the greatest relative volatility. While Strategy 2 exhibits approximately the same risk of failure (~25%) as Strategy 1 (~27%), it also exhibits the least maximum exposure and relative volatility. To reduce the ambiguity of choosing, we might like to know which uncertainty, due to the overall quality of the information we possess about it, contributes most to switching dominance from Strategy 1 over Strategy 2. Knowing which uncertainty our strategy values are most sensitive to gives us the ability to stop worrying about the other uncertainties for the purpose of choosing clearly and focus only on improving our understanding of those that matter most.</p><p>To accomplish that latter feat, we run our functionalized model to test the sensitivity of the strategy means to the 80th percentile range in each of the variables. We start by initializing the lists to hold the sensitivity responses.</p><pre>data.temp &lt;- proj.data<br>proj.npv.sens &lt;- list(p10 = proj.data,<br>                      p90 = proj.data)<br>d &lt;- data.frame(0, 0, 0)<br>dd &lt;-<br>  list(<br>    investment = d,<br>    start.cash.flow = d,<br>    peak.cash.flow = d,<br>    yr.peak = d<br>  )<br>proj.npv.sens.mean &lt;- list(p10 = dd,<br>                           p90 = dd)</pre><p>We calculate the sensitivity of the strategies’ mean NPVs by fixing each variable to its p10 and p90 quantile values sequentially while all the other variables run according to their defined variation. The result is a chart that shows how sensitive we might be to changing our decision to pursue the best strategy over the next best strategy on the outcome of a given uncertainty.</p><pre>p1090 &lt;- c(0.1, 0.9)<br>for (q in 1:length(p1090)) {<br>  for (v in 1:length(proj.data)) {<br>    for (s in 1:length(strategy)) {<br>      data.temp[[v]][s] &lt;-<br>        rep(quantile(unlist(proj.data[[v]][s]), probs = p1090[q]), trials)<br>      for (t in 1:trials) {<br>        proj.npv.sens[[q]][[v]][t, s] &lt;- calcNPV(<br>          CF = calcCashFlow(<br>            I = data.temp$investment[[s]][t],<br>            SCF = data.temp$start.cash.flow[[s]][t],<br>            PCF = data.temp$peak.cash.flow[[s]][t],<br>            YP = data.temp$yr.peak[[s]][t],<br>            Y = year<br>          ),<br>          Y = year,<br>          DR = discount.rate<br>        )<br>      }<br>      proj.npv.sens.mean[[q]][[v]][s] &lt;-<br>        mean(proj.npv.sens[[q]][[v]][, s])<br>      data.temp &lt;- proj.data<br>    }<br>  }<br>}<br> <br># Recast the sensitivity values to a form that can be plotted.<br>variable.names &lt;-<br>  c(&quot;investment&quot;, &quot;start.cash.flow&quot;, &quot;peak.cash.flow&quot;, &quot;yr.peak&quot;)<br>proj.npv.sens.mean2 &lt;- as.data.frame(sapply(1:length(p1090), function(q) {<br>  sapply(1:length(proj.data), function(v) {<br>    sapply(1:length(strategy), function(s) {<br>      proj.npv.sens.mean[[q]][[v]][[s]]<br>    })<br>  })<br>})) %&gt;%<br>  rename(p10 = V1,<br>         p90 = V2) %&gt;%<br>  mutate(<br>    variable = rep(variable.names, each = length(strategy)),<br>    strategy = rep(strategy, times = length(proj.data)),<br>    strategy.mean = rep(t(proj.npv.mean), times = length(proj.data))<br>  ) %&gt;%<br>  select(variable, strategy, strategy.mean, p10, p90)<br> <br>gg.sens &lt;-<br>  ggplot(proj.npv.sens.mean2,<br>         aes(<br>           x = variable,<br>           y = p10,<br>           yend = p90,<br>           color = strategy<br>         )) +<br>  geom_segment(aes(<br>    x = variable,<br>    xend = variable,<br>    y = p10,<br>    yend = p90<br>  ),<br>  color = &quot;gray50&quot;,<br>  size = 0.75) +<br>  geom_point(<br>    aes(x = variable, y = strategy.mean, color = strategy),<br>    size = 1,<br>    color = &quot;black&quot;,<br>    fill = &quot;black&quot;,<br>    shape = 23<br>  ) +<br>  geom_point(<br>    aes(x = variable, y = p90),<br>    size = 3) +<br>  geom_point(<br>    aes(x = variable, y = p10),<br>    size = 3) +<br>  labs(title = &quot;Sensitivity of Strategy Mean NPV to Uncertain Variable Ranges&quot;,<br>       y = &quot;Strategy NPV [$M]&quot;) +<br>  coord_flip() +<br>  theme(legend.position = &quot;none&quot;) +<br>  facet_wrap(~ strategy)<br>plot(gg.sens)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/840/0*pmwxS_hR4YZu7D6o.png" /></figure><p><em>Figure 4: The sensitivity chart shows the range of variation in a strategy mean due to an 80th percentile variation in a given uncertainty. The mean value of each strategy is marked by the black diamond within the range of each variable. According to our results here, we should focus additional research attention on the investment range first and then the year to peak.</em></p><p>This sensitivity chart is valuable not only to the decision analyst, but to the data analyst as well for this one simple reason: now that we know which uncertainties potentially exposes us to the greatest risk or reward, we can calculate the value of information on those uncertainties. This value of information represents the rational maximum budget we should allocate to acquire better data and insight on those uncertainties. While a typical business case analysis of sufficient complexity may contain 10–100 uncertainties, we should focus our research budget only on those that are critical to making a decision. This simple “stopping function” helps us ensure that our data science resources are allocated properly and most efficiently. <strong>Section IV</strong> of <em>BCAwR</em> provides a means to calculate the value of information on continuous variables like the ones we have utilized in our example here.</p><h3>Conclusion</h3><p>Business decision analysts can definitely use <strong>R</strong> effectively to conduct many of the business case analyses in a more transparent and less error prone environment than their current tools typically allow, especially through the use of clearly named variables that array abstract and the use of functionalized structural models. However, the real power that R provides is the means to simulate information when high quality, large scale data may be scarce, infeasible, or impossible to obtain. Then, by using such concepts as value of information, data scientists can be called upon to refine the quality of information on those critical uncertainties that may frustrate success when commitment to action turns into real decisions. I hope the simple example presented here inspires you to learn more about it.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9d9ba3d00c15" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Cost of Personal Records Exfiltration: Regression Derivation]]></title>
            <link>https://medium.com/@robertdbrowniii/the-cost-of-personal-records-exfiltration-regression-derivation-c3fc1482d7f0?source=rss-f345147f5249------2</link>
            <guid isPermaLink="false">https://medium.com/p/c3fc1482d7f0</guid>
            <dc:creator><![CDATA[Robert D. Brown III]]></dc:creator>
            <pubDate>Wed, 14 Dec 2022 01:32:37 GMT</pubDate>
            <atom:updated>2023-11-13T02:13:51.100Z</atom:updated>
            <content:encoded><![CDATA[<p>(continuing from <a href="https://www.linkedin.com/pulse/can-you-afford-ignore-financial-liability-sensitive-brown-iii/">Can You Afford to Ignore the Financial Liability of Sensitive Records</a>?)</p><p>For those with a deeper interest in cyber risk quantification methods, we provide the derivation to our regression model below. Warning! Really geeky stuff is about to happen.</p><p>The digitized data set contained 181 pairs of values in log10 (records lost) — log10 (unit cost) format. We converted these to <strong>natural log format</strong>. After importing this data into RStudio as a dataframe (<em>Verizon_2015_DBIR_df</em>), we ran a linear regression using</p><pre>unit_cost_record &lt;-<br>  lm(y ~ x, Verizon_2015_DBIR_df)</pre><p>and obtained the following result:</p><pre>Coefficients:<br>(Intercept)            x  <br>     8.18         -0.577</pre><p>This gives the formula for predicting the average log unit cost.</p><pre>log_pred_cost &lt;-<br>  Verizon_2015_DBIR_df$x * unit_cost_record$coefficients[2] + <br>  unit_cost_record$coefficients[1]</pre><p>or</p><pre>log_pred_cost &lt;-<br>  -0.577 * R + 8.18</pre><p>where <strong><em>R</em></strong> is the natural log of the number of lost or exposed records.</p><p>Using this formula to plot the predicted values against the records data yields the following chart (Figure 4).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IrLW-Wh2SKjWLFPHdb7z3g.png" /><figcaption><strong>Figure 4: </strong>Scatter plot of the reported data of natural log unit costs to natural log of lost data (red dots) compared to the predicted average natural log unit costs for the same lost data values (green dots).</figcaption></figure><p>Remember that the data we are using has been displayed and processed in a log-log form. We need to make a final cost prediction in real dollar units. But converting these data to real units as an exponent of <em>e</em> (the natural base number 2.7182…) will distort the symmetry we observe in the dispersion of the data in the current log-log form. As a result, the predicted average real value cannot be found by simply raising <em>e</em> to the predicted log value. Doing that would return only the median value. To get the average value, we must adjust that operation by some factor that accounts for the error distribution about the green line. As math would have it, our lives will be much easier if this distribution is similar in shape to the theoretical Normal distribution. Therefore, to calculate the average in real dollars from our regression, we will first need to confirm that the error distribution conforms to Normal, and then we must find what the standard deviation of the error is. We calculate the error as</p><p><em>error = predicted log unit cost — reported log unit cost</em></p><p>for all the points in the data. A quick visual inspection of the error distribution (Figure 5) and its Normal Q-Q plot (Figure 6) reveals that treating the error as Normally distributed appears reasonable. Figure 5 presents a familiar bell-shape that approximates the Normal distribution, and the Normal Q-Q plot points fall very closely to the line of best fit for a perfect Normal distribution.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LohP_yZkbh4bhADT7EMX3g.png" /><figcaption><strong>Figure 5: </strong>The distribution of the log error from the actual reported unit costs to the predicted unit costs at each level of lost records.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UOYRU4vaOsTcp6N_QfSMPg.png" /><figcaption><strong>Figure 6: </strong>The quantile-quantile scatter plot of the error distribution compared to the theoretical normal distribution with the same implied standard deviation of the error distribution. If the error distribution was truly Normal, its quantile values would fall on the blue line that maps the quantile values of the theoretical Normal distribution.</figcaption></figure><p>The mean and standard deviation of the error distribution are <em>mean = 0</em> and <em>sd = 1.587</em>. We can use this information to simulate the reported data distribution using an updated prediction cost formula</p><pre>simulated_log_pred_cost &lt;-<br>  -0.577 * R + 8.18 + rnorm(n=181, mean=0, sd=1.587)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*71K5_KChMqf-FALSGqLL3g.png" /><figcaption><strong>Figure 7: </strong>Figure 4 overlaid with simulated scatter values implied by the distribution of the error between the original data and the predicted values.</figcaption></figure><p>Figure 7 shows that our simulation formula coheres rather nicely with the scatter of the reported data. However, using the full standard deviation of 1.587 ultimately produces results in real dollars that are probably much too wide. To adjust for this, we reduced the standard deviation by 4%. We now formalize our formula into an average <strong>real dollar format</strong>:</p><pre>avg_pred_cost &lt;-<br>  exp(-0.577 * log(R) + 8.18 + (0.96 * sd)^2 /2),</pre><p>where <em>sd = 1.587</em>. Notice that we have replaced the error generating part (i.e., <em>rnorm(n=181, mean=0, sd=1.587)</em>) with half of the adjusted squared standard deviation (or, half the variance). Simplifying gives</p><pre>avg_pred_cost &lt;-<br>  11390 * R^(-0.577).</pre><p>The version that retains the simulation capacity is</p><pre>simulated_pred_cost &lt;-<br>  3568 * R^(-0.577) * exp(N(0, 0.96 * sd))</pre><p>where <em>N(0, 0.96 * sd)</em> is the Normal distribution generator with <em>mean = 0</em> and <em>standard deviation = 0.96 * sd</em>.</p><p>(Back to <a href="https://www.linkedin.com/pulse/can-you-afford-ignore-financial-liability-sensitive-brown-iii/">Can You Afford to Ignore the Financial Liability of Sensitive Records?</a>)</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c3fc1482d7f0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Writing Stuff Down]]></title>
            <link>https://medium.com/@robertdbrowniii/writing-stuff-down-6135dc953310?source=rss-f345147f5249------2</link>
            <guid isPermaLink="false">https://medium.com/p/6135dc953310</guid>
            <category><![CDATA[decision-making]]></category>
            <dc:creator><![CDATA[Robert D. Brown III]]></dc:creator>
            <pubDate>Wed, 23 May 2018 00:00:00 GMT</pubDate>
            <atom:updated>2018-05-27T17:35:24.284Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*wl0r3VpR89Ts7wzS." /><figcaption>Photo by Glenn Carstens-Peters on Unsplash</figcaption></figure><p><a href="https://www.chicagobooth.edu/faculty/directory/t/richard-h-thaler">Richard Thaler</a>, recent Nobel recipient for his contribution to <a href="https://en.wikipedia.org/wiki/Behavioral_economics">behavioral economics</a>, offers some valuable and powerful advice <a href="https://www.mckinsey.com/business-functions/strategy-and-corporate-finance/our-insights/debiasing-the-corporation-an-interview-with-nobel-laureate-richard-thaler">here</a>:</p><blockquote><em>Write stuff down.</em></blockquote><p>Good stuff from a Nobel laureate, right?</p><p>But why should you write stuff down?</p><blockquote><em>So, any company that can learn to distinguish between bad decisions and bad outcomes has a leg up.</em></blockquote><p>Writing stuff down provides an effective way to record and memorialize your thinking so that you can review how your thinking efforts led you to the decisions you made and the outcomes you experienced. It allows you to record critical distinctions and appropriate discriminations, and when the time is right, to keep yourself humble so that you remain open to learning. It helps you avoid cheating yourself of a better life via the cognitive bias of post hoc rationalization, as good as that might feel in the moment when you seek justification or an ego boost.</p><p>For organizations, one of the key contributors to realizing Thaler’s advice is to develop and use a decision ledger to document and track the collaborative thinking that goes into making high value, complex business decisions from problem/opportunity recognition through framing, quantitative evaluation, agreement, and finally detailed planning. The purpose of the decision ledger (when used correctly and not just as a check list) is to ensure that your team avoids making bad decisions by failing to address the crucial information required to make high quality decisions, quality that when repeatedly observed leads to an increase in the likelihood of experiencing good outcomes. From the perspective of satisfying the fiduciary responsibilities of directors, your organization will be able to provide the <strong>essential warranties</strong> that its directors have observed their duty of care. But more importantly, they will have actually done so.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LSDBSSf9BM3B26sm7fuC6g.png" /></figure><p>Furthermore, a decision ledger helps to eliminate the effects of adversarial and advocacy driven decision making by documenting early in a decision making process what the criteria for selecting a pathway forward will be and how the implemented decisions (as policies or program/project/portfolio initiatives) will be judged as successful and to what degree. When all the final evidence and reasoning is presented and considered, the decision team will usually possess all the right information to know what to do and why (and how to resolve situations that remain ambiguous), eliminating the need for final voting that is often swayed by political influence…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*QjKyvPgPkYxrinbRawhi3Q.jpeg" /><figcaption>Photo by <a href="https://unsplash.com/photos/IBWJsMObnnU?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Arnaud Jaegers</a> on <a href="https://unsplash.com/search/photos/voting?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></figcaption></figure><p>…or driven by overriding agendas.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*GGX9L0tXAcQc5p6IpARwkA.jpeg" /><figcaption>Photo by <a href="https://unsplash.com/photos/BbOXC95sxlE?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Sagar Dani</a> on <a href="https://unsplash.com/search/photos/puppets?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></figcaption></figure><p>Want to learn more? <a href="https://calendly.com/rdbrown">Schedule a free call</a> with me to discuss what your organization needs to do to distinguish between bad decisions and bad outcomes, and how establishing a decision ledger for your organization to write stuff down can help you “get a leg up.”</p><p><em>Originally published at </em><a href="https://www.linkedin.com/pulse/writing-stuff-down-robert-d-brown-iii/"><em>https://www.linkedin.com</em></a><em> on May 23, 2018.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6135dc953310" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Which Part of Your Data Analytics Dollar is Wasted?]]></title>
            <link>https://medium.com/@robertdbrowniii/which-part-of-your-data-analytics-dollar-is-wasted-2bc96cbf96fd?source=rss-f345147f5249------2</link>
            <guid isPermaLink="false">https://medium.com/p/2bc96cbf96fd</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[portfolio]]></category>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[big-data]]></category>
            <category><![CDATA[decision-making]]></category>
            <dc:creator><![CDATA[Robert D. Brown III]]></dc:creator>
            <pubDate>Sun, 15 Apr 2018 19:17:40 GMT</pubDate>
            <atom:updated>2018-04-15T19:17:40.012Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Improve your chances of achieving successful data analytics initiatives</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uBZPF6_qXKDs699af27tBw.jpeg" /><figcaption>Photo by Sharon McCutcheon on Unsplash</figcaption></figure><blockquote><strong>“Half the money I spend on advertising is wasted; the trouble is, I don’t know which half.”</strong></blockquote><p>You’ve probably heard that saying by department-store magnate John Wanamaker (1838–1922). Given that data analytics initiatives are facing approximately a <a href="https://www.gartner.com/newsroom/id/3130017">60% failure rate</a>, those in corporate leadership who are allocating resources to mining BigData, becoming data-driven organizations (I still roll my eyes at that bumper sticker phrase), or investing in the new, new, new economy of “performing on analytics” are effectively asking the same question, whether they know it or not. For every $5 spent on new data analytics initiatives, $3 will be wasted. And if you’re the spender, you won’t know which $3 until it’s too late. Didn’t you think things were supposed to be more spectacular than this?</p><blockquote><strong>Where’s the kaboom? There was supposed to be an Earth-shattering kaboom!</strong></blockquote><p>Why is this situation the case? Why doesn’t the data analytics revolution — whose exponents seem to suggest that data analytics will relentlessly solve all our problems — solve its own problem of value delivery? Several culprits likely contribute to the current reality, for example…</p><ul><li>The talent pool is inexperienced, undersupplied to the call of the market, and overpriced. A lot of people are getting certified as data scientists through a range of bootcamps of varying quality, but they still lack the critical business acumen to design solutions for effective usability by decision makers.</li><li>The actual data quality used in data analytics continues to defy standards of high quality, including having access to data that is relevant to the problem at hand.</li><li>The solution approaches are over-engineered to the problem at hand. Recurrent and convolutional neural networks are cool, but not every problem needs to be addressed with one.</li><li>Project mismanagement: schedule and budget overruns, and deliverables failing to meet specifications.</li></ul><p>I suspect that more mundane explanations exist. But I do not believe that the reasons for failure are that data analytics cannot deliver value. We’ll get to that latter point in a moment. The problem as I see it is that a data analytics initiative is just like any other initiative or project, and all other kinds of projects are known to fail at similar rates (see this summary of the <a href="https://www.pmiwdc.org/sites/default/files/presentations/201703/PMIW_LocalCommunity_Tysons_presentation_2017-02.pdf">Chaos Report</a>) and for repeatedly known categories of reasons. According to the research by <a href="http://aax-us-east.amazon-adsystem.com/x/c/Qjo495EvTkQ0BIjJYAs0jcwAAAFilAB35wEAAAFKAbtzBJA/http://www.amazon.com/gp/product/1576751503/ref=as_at/?imprToken=BefXYBPN886iR0FwKe7MDQ&amp;slotNum=39&amp;ie=UTF8&amp;camp=1789&amp;creative=9325&amp;creativeASIN=1576751503&amp;linkCode=w61&amp;tag=thapre-20&amp;linkId=MCTFATILRZ6AKGGY">Paul C. Nutt</a>, most projects fail before execution begins in the problem solving and planning phases of management. Among the worst things that decision-makers can do to ensure that they waste good money is to solve irrelevant problems for the wrong reasons (A close second is to throw good money after bad in an attempt to recover sunk costs once execution begins).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iu5zEj69tOZZgt_0c25zhA.png" /><figcaption><em>Fig. 1 — Projects of all kinds fail or become challenged at an alarmingly high rate. Data analytics initiatives are just another kind of project. (Thanks to </em><a href="https://www.linkedin.com/in/etorkia/"><em>@Eric Torkia</em></a><em> with </em><a href="https://www.crystalballservices.com/"><em>Technology Partnerz</em></a><em> for compiling the information in the above chart.)</em></figcaption></figure><p>Guess what? It gets worse.</p><blockquote>A year ago, Gartner estimated that 60% of big data projects fail. As bad as that sounds, the reality is actually worse. <a href="https://twitter.com/nheudecker/status/928720268662530048">According to Gartner analyst Nick Heudecker</a>‏ this week, Gartner was ‘too conservative’ with its 60% estimate. The real failure rate? ‘[C]loser to 85 percent.’ In other words, abandon hope all ye who enter here, especially because ‘[T]he problem isn’t technology,’ <a href="https://www.techrepublic.com/article/85-of-big-data-projects-fail-but-your-developers-can-help-yours-succeed/">Heudecker said</a>. ‘It’s you.’</blockquote><blockquote><a href="https://siliconangle.com/blog/2016/10/04/gartner-warns-big-datas-bubble-may-burst-as-enterprises-plot-investments-elsewhere/">Gartner says</a> that one of the main reasons big data projects have failed to come together is a lack of proper management, combined with lots of trial and error from companies that just throw systems together willy-nilly.</blockquote><p>So, really, for every $5 spent on new data analytics initiatives, $4.25 will be wasted.</p><p>The problem isn’t data science, per se. It’s the way we choose which initiatives to do and why, followed by how. And so, just like for any other project, the biggest reason that data analytics initiatives fail is that many problems applied to data analytics are irrelevant to the business. Instead of framing a data analytics project in terms of business case value, very often these projects are just interesting science fair projects. We’re all for fairness and science, but nothing represents a bigger waste of time and money than irrelevant projects.</p><p>So, how can you know for sure before you spend any money on any data analytics initiative whether you will waste more than half your money or not? You can’t. But, you can get better at making better bets. Consider the following example.</p><p>A few years ago, a business unit leader within a development group requested some help justifying an increase of the group’s budget to accommodate a raft of really interesting new ideas. The suggested portfolio contained 122 new projects. But my Spidey sense began to tingle. In my nearly 20 years of experience doing decision and risk analysis, I have rarely seen an initial portfolio of initiatives stand completely justified. For every incremental dollar expended to acquire value, a diminishing return effect occurs that eventually makes a few projects economically unjustifiable. And so it was with this case.</p><p>To address the issue, we first framed every initiative in terms of the objectives of the company each should achieve and identified at least a notional way to know that each was supporting the achievement of those objectives. We then devised a way to measure the value of each initiative using “small data” (most of which was derived by the help of subject matter expert guidance and assessment) and probabilistic reasoning using simulation, a kind of virtual analysis of variance. And what did we find? Well, approximately 50% of the projects generated 98% of the portfolio’s value, and that 50% suggested a required budget of only 15% more than the previous year’s budget.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4J8MIqw-C1WD8aNpqrm2VA.png" /><figcaption><em>Fig. 2 — The cumulative incremental cost of each project ordered by mean return ($Value/$Investment) yields a declining marginal return. Not all projects are feasible, and they should be pruned from consideration when their marginal contribution of mean $Value ≤ 0.</em></figcaption></figure><p>We pruned off the 50% that remained beyond the “rational” budget (i.e., original budget plus the incremental 15%), many of which were actually just science fair projects. I recall that one engineer explained to me that if he solved the heat transfer problem he was so concerned about, he would gain some fame in the mechanical engineering world — he’d get the blue ribbon. Unfortunately, that project didn’t appear to get out of the red. Fortunately, we caught it before it became a part of the development plan. <strong>We saw which dollars most likely would have been wasted before they even got a chance to be wasted.</strong></p><p>For the initiatives that consumed the largest amount of budget, we used <a href="http://www.incitedecisiontech.com/intro2.shtml#valueofinfo">Value of Information and Value of Control</a> to find ways to tweak their value even more, adding more slope to the marginal contribution of value of each initiative.</p><p>For the next four years thereafter, we calibrated the portfolio budget using the same approach, and each year the project portfolio marginal contribution curve increased its slope and its total contribution with fewer projects and with about the same budget. The group simply got better at identifying potentially valuable projects before we evaluated them economically, most likely because they developed a deeper understanding of the features of initiatives that supported organizational objectives and why.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Q6KrRLmfuwboqPnLkvmZMQ.png" /><figcaption><em>Fig. 3 — Over time, with each planning period, you will learn how to recognize better and better things to do with your budget using a portfolio analysis approach to guide the selection and prioritization of data analytics initiatives.</em></figcaption></figure><p>The analytics maturity evolution model tells us that the most sophisticated organizations achieve upright status when they embrace <a href="https://www.gartner.com/it-glossary/prescriptive-analytics/">prescriptive analytics</a>; i.e., those analytics that tell you what you ought to do beyond the basics of merely describing what the state of nature is (descriptive), why the state of nature presents the way it does (diagnostic), and what the future state of nature will be (predictive). However, that organizational evolutionary process takes too long and it leaves behind a bloody trail of extinct dollars and projects. What we need is a kind of black <a href="https://en.wikipedia.org/wiki/Monolith_(Space_Odyssey)">monolith</a> that tells us what kinds of data initiatives we should be doing and why — a meta-prescriptive system, if you will. That guidance system is the application of portfolio analysis, a selection process that serves as the unifying glue for business plans that helps to remove the pain of strategic misalignment and bias.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AIgJEu3lEJS08XsKTYZu4w.png" /><figcaption><em>Fig. 4 — Evolution gets us there, maybe, but it’s brutish and bloody. It would be nice to have some meta-prescriptive guidance.</em></figcaption></figure><p>If you are a leader in an organization wanting to capture the value of the new revolution of data science, and you care that your resources are appropriately allocated to achieve success, you will want to avoid the failure of relevance. One of the key contributors to doing that is to implement a risk-based project portfolio analysis process for your data analytics initiatives, one which specifically helps you understand…</p><ul><li>The reason an analytics initiative should be assumed;</li><li>What the potential value and risks to your organization might be for each;</li><li>How to mitigate those risks and amplify the benefits to maximize the likelihood of success;</li><li>How to prioritize and optimize the portfolio of all your analytics projects so that your capital is used in the most efficient way.</li></ul><blockquote><em>Want to learn more? </em><a href="https://calendly.com/rdbrown"><em>Schedule a free call</em></a><em> with me to discuss your data analytics initiatives and how to increase the effectiveness of your investment dollars.</em></blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2bc96cbf96fd" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Publication Announcement — Business Case Analysis with R : Simulation Tutorials to Support Complex…]]></title>
            <link>https://medium.com/@robertdbrowniii/publication-announcement-business-case-analysis-with-r-simulation-tutorials-to-support-complex-4ff21a0f7d14?source=rss-f345147f5249------2</link>
            <guid isPermaLink="false">https://medium.com/p/4ff21a0f7d14</guid>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[analytics]]></category>
            <dc:creator><![CDATA[Robert D. Brown III]]></dc:creator>
            <pubDate>Sun, 18 Mar 2018 17:03:57 GMT</pubDate>
            <atom:updated>2018-03-18T17:17:39.226Z</atom:updated>
            <content:encoded><![CDATA[<h3>Publication Announcement — Business Case Analysis with R : Simulation Tutorials to Support Complex Business Decisions</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/827/1*Xtr83vWab5kX4e-cgsUuiw.tiff" /></figure><p>I’m pleased to announce the publication of <a href="https://www.apress.com/us/book/9781484234945#">Business Case Analysis with R — Simulation Tutorials to Support Complex Business Decisions</a>. Written in a conversational style, this collection of tutorials is intended to help analysts develop an emerging interest and skill set in R and to promote a better selection of tools to perform business case analysis other than error prone spreadsheets. It’s some R and little bit more.</p><p>“Business case analysis, often conducted in spreadsheets, exposes decision makers to additional risks that arise just from the use of the spreadsheet environment. This book discusses how to use the statistical programming language R to develop a business case simulation and analysis. It presents a methodology that minimizes decision delay by focusing stakeholders on what matters most and suggests pathways for minimizing the risk in strategic and capital allocation decisions.”</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4ff21a0f7d14" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Advice to Young People Just Beginning Their Career]]></title>
            <link>https://medium.com/@robertdbrowniii/advice-to-young-people-just-beginning-their-career-4814db43f0e2?source=rss-f345147f5249------2</link>
            <guid isPermaLink="false">https://medium.com/p/4814db43f0e2</guid>
            <category><![CDATA[careers]]></category>
            <category><![CDATA[self-improvement]]></category>
            <category><![CDATA[education]]></category>
            <category><![CDATA[entrepreneurship]]></category>
            <dc:creator><![CDATA[Robert D. Brown III]]></dc:creator>
            <pubDate>Mon, 14 Nov 2016 03:57:44 GMT</pubDate>
            <atom:updated>2024-01-12T15:46:09.083Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*cM5m_mAGcN6qWm_w6LmdIQ.jpeg" /><figcaption>Benjamin Braddock (played by Dustin Hoffman) has just finished college and, back at his parents’ house, he’s trying to avoid the one question everyone keeps asking: What does he want to do with his life? (The Graduate. December 22, 1967)</figcaption></figure><p>A few years ago as I was returning to the US from a business trip to Europe, I found myself yet again waiting in the interminable line at security and customs at the ATL. To pass the time, a recent high school graduate started up a conversation about her trip to Europe and her plans for study in college. Eventually, she asked me what advice I might have for a young person just beginning her career. So, over the remaining minutes in the security line I rattled off some of the advice I wish I had received when I was in her position.</p><p>As coincidence would have it, an opportunity to reply to a similar question arose a few weeks later, and then another. After the third or fourth event of this kind, I wrote down my initial thoughts, then I kept adding to them over time. Now I have a list of eighteen items that I share whenever I’m asked, “What advice do you have for a young person just beginning their career?” The list is by no means comprehensive. It represents what I might call the meta-principles for success; i.e., the organizing principles that will help anyone develop the mindset and attitudes that lead to personal success as you define it.</p><ol><li>If you haven’t done so already, identify an area that really interests you. If you don’t know of something right now, recall your days of kindergarten. A lot of kindergarteners eat Play-Doh and glue, though, so it’s not a guaranteed solution, but it’s not a bad place to start. Do not let yourself be burdened by the fear that this interest area has to represent your career choice for the rest of your life. It’s a starting point. You can decide later whether you want to make it a lifetime career choice or a stepping stone on the way to something else.</li><li><strong>Obsessively study, play, and explore</strong> with the area of interest to learn nuances and distinctions that few others know about. Do not let your education rely on just what you get in classes. Do your own research and keep a record of interesting ideas and observations. Make your own education.</li><li>Identify an opportunity to create or serve a market with the ideas you’ve developed from exploring the area of interest. I tire of the Pursue Your Passion vs Respond to the Opportunity debate. A successful life incorporates both. But, if you’re not eventually passionate about something in your life, you’re simply marking time. Passion and market opportunity make for a really fun and meaningful life.</li><li>Get a mentor.</li><li>Actually, get two mentors. The beauty of being young is that you haven’t been spoiled by cynicism yet and haven’t yet learned what people expect you not to try or accomplish. But the problem with being young is that you don’t have a long horizon view of life, which causes you to over emphasize trivial issues. Mentors help you avoid the problem of horizon blindness. The problem with mentors, though, is that by the time they’re old enough to be a mentor half of what they know is obsolete. Therefore, two mentors (~15 yrs older and ~30 yrs older) help you overcome both the near-term and long-term perspectives of life that provide a mature understanding of current conditions as well as a mature understanding of long horizon priorities.</li><li>Publish something under the guidance of the mentors. Writing down what you know and think to be consumed by an audience clarifies your own thinking and teaches you to communicate clearly. This leads to becoming a thought leader. Don’t tell people you’re a thought leader. Demonstrate it.</li><li>Learn to network with people properly. This means learning to identify ways in which you can offer some solution to a problem or provide access to someone else for the people you meet. For the first two years of your networking efforts, do not ask someone to help you without establishing a trusted relationship with that person. A trusted relationship is one that follows through with promises and keeps strict confidences for what people tell you in private. Ultimately this means learning to be a good empathetic conversationalist, but not just for the purpose of having an entertaining evening. (Although there’s nothing wrong with having an entertaining evening.)</li><li>Learn to establish trust with people. Do not tell people that you can be trusted. Demonstrate that you are trustworthy. Never forget that the entire system of voluntary communities and exchange within those communities is based on trust. It is by no means an exaggeration to say that when you violate trust, you contribute to the erosion of the very basis of modern, enlightened, secular society. Trust is the golden nugget.</li><li>Never take the social capital of others for granted. When someone takes the time and effort to introduce you to someone else, follow up immediately and look for ways to benefit the person to whom you’ve been introduced.</li><li>Help out people even when you have no idea how you can afford to and when you have no idea if and when they can compensate you for your effort. It comes back to you when you least expect it and when you most need it. This is not self-serving or Machiavellian. It’s how enlightened people in voluntary communities support the system of a voluntary community.</li><li>However, learn to say “No” but judiciously. You can’t do everything nor help everyone solve all their problems. A burned-out you is no help to anyone. However, you can always commit to finding someone else that is currently in a better position to help.</li><li>Travel broadly. Soak up the culture. Shed your prejudices and biases as quickly as possible.</li><li>Meditate deeply, oftentimes for hours and with no intrusions from other people. This clarifies your thinking, helps you see issues deeply, and, most importantly, helps you formulate the questions that need to be resolved. At the very least, it lowers your blood pressure.</li><li>Learn to maintain healthy eating practices, especially when you travel. It’s a little more expensive, and it often irritates the people you’re with, but it helps to avoid or defer some really expensive health issues later on.</li><li>Learn to forgive and to forgive properly. You’re going to need both in life.</li><li>Say thank you. It’s cheap and makes people feel appreciated for doing what are often thankless tasks. You will learn soon enough what it feels like to be on the business end of a thankless task and why acknowledgment fuels your ability to continue when you really want to stop. The world revolves around people who don’t stop. Saying “Thank you” fuels the revolution.</li><li>Learn to compliment others when appropriate, but do so judiciously. Don’t let compliments flow easily or embroider them unnecessarily. When a person who knows you well receives a compliment from you, they will know that you gave it thoughtfully and meaningfully. Otherwise, compliments become as cheap as candy and provide as much nutritional value.</li><li>Develop the inclination to ask “How do you/I know that?” Healthy skepticism is the basis of risk management. It will eventually save you and the interests you represent money and embarrassment.</li></ol><p>What are your thoughts? What do you wish you had been guided to learn or adopt when you are just starting your career? Are you prepared to pass on that wisdom constructively when asked, “What advice do you have for a young person just beginning their career?”</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4814db43f0e2" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Seeking Beta Testers for a Web-based Sales Opportunity Portfolio Analysis Tool]]></title>
            <link>https://medium.com/@robertdbrowniii/seeking-beta-testers-for-a-web-based-sales-opportunity-portfolio-analysis-tool-41b586ba5f1d?source=rss-f345147f5249------2</link>
            <guid isPermaLink="false">https://medium.com/p/41b586ba5f1d</guid>
            <category><![CDATA[sales]]></category>
            <category><![CDATA[portfolio]]></category>
            <category><![CDATA[salesforce-productivity]]></category>
            <dc:creator><![CDATA[Robert D. Brown III]]></dc:creator>
            <pubDate>Mon, 14 Nov 2016 00:00:00 GMT</pubDate>
            <atom:updated>2016-11-17T04:12:35.416Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/400/0*7SxleRwH34Q-bP6f.jpg" /></figure><p><a href="http://www.incitedecisiontech.com/index.shtml">Incite! Decision Technologies</a> has recently developed a simple yet sophisticated web-based sales opportunity portfolio analysis tool that is ready for beta testing. Now we’re seeking parties that would be interested in participating at <strong><em>no cost and no obligation</em></strong>.</p><p>Specifically, we are looking for progressive sales managers in firms whose sales team pursues high value, low-frequency sales. Examples of target firms might be…</p><ul><li>Engineering, architecture &amp; construction firms</li><li>Professional service firms</li><li>Capital equipment manufacturers</li><li>Start-ups</li></ul><p>The purpose of the tool is to provide</p><ul><li>Improved accuracy of revenue realization and timing forecasts;</li><li>Guidance on how to allocate resources to maximize the likelihood of deal closure;</li><li>Guidance on opportunity selection and prioritization.</li></ul><p>Ultimately, you will be able to determine if the sales opportunities you are pursuing are <a href="https://medium.com/@robertdbrowniii/the-power-of-negative-thinking-how-do-we-know-this-opportunity-is-worth-the-time-and-effort-6b702ad24c86#.2c0zxem0x">worth it</a>.</p><p>If you are interested in learning more or know someone who might be, please, contact me via LinkedIn message or send me an email from our <a href="http://www.incitedecisiontech.com/contact.shtml">web form</a>.</p><p><em>Originally published at </em><a href="http://thalespress.blogspot.com/2016/11/seeking-beta-testers-for-web-based.html"><em>thalespress.blogspot.com</em></a><em> on November 14, 2016.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=41b586ba5f1d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Power of Negative Thinking: “How do we know this opportunity is worth the time and effort?”]]></title>
            <link>https://medium.com/@robertdbrowniii/the-power-of-negative-thinking-how-do-we-know-this-opportunity-is-worth-the-time-and-effort-6b702ad24c86?source=rss-f345147f5249------2</link>
            <guid isPermaLink="false">https://medium.com/p/6b702ad24c86</guid>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[sales]]></category>
            <dc:creator><![CDATA[Robert D. Brown III]]></dc:creator>
            <pubDate>Wed, 09 Nov 2016 00:00:00 GMT</pubDate>
            <atom:updated>2016-11-15T01:36:14.139Z</atom:updated>
            <content:encoded><![CDATA[<p>The sales process is an inherently risky business. It’s difficult to know if and when a deal will close, what clients really want regardless of what they have stated (i.e., the client may have failed to frame their own needs properly), and what competitors offer in price and quality of deliverables.</p><p>Compounding the external uncertainty, we often get in our own way by importing certain kinds of biases into our assessment of the value of the sales opportunities at hand. These biases can include…</p><ul><li><strong>Unwarranted optimism</strong> or <strong>wishful thinking</strong> — personal enthusiasm or a natural disposition to believe that desired outcomes will most likely occur; or, inflating initial estimates of desired outcomes to appear more effective than is warranted;</li><li><strong>Sand-bagging</strong> — under reporting potential outcomes to appear heroic when better than anticipated outcomes materialize;</li><li><strong>False precision</strong> — reporting anticipated outcomes with an unjustified level of certainty, usually as a single-point estimate rather than a range;</li><li><strong>Availability</strong> — recalling values that are memorable, easily accessible, recent, or extreme;</li><li><strong>Anchoring</strong> — using the first “best guess” as a starting point for subsequent estimating;</li><li><strong>Expert over-confidence</strong> — failure of creativity or hubris (e.g., “I know this information and can’t be wrong because I’m the expert.”);</li><li><strong>Incentives</strong> — the SME experiences some benefit or cost in relationship to the outcome of the term being measured, adjusting his estimate in the direction of the preferred outcome;</li><li><strong>Entitlement</strong> — the SME provides an estimate that reinforces his sense of personal value.</li></ul><p>Without bias-free assessments in our decisions to actively pursue sales opportunities, it’s nearly impossible to know how to allocate sales and support resources effectively to maximize the likelihood of capturing sales in a profitable and efficient manner. In short, when given the opportunity to pursue multiple opportunities with limited resources, it’s often difficult to know if any given opportunity is worth the time.</p><p>As odd as it may sound in a culture that seems to demand almost endless optimism, the Power of Negative Thinking actually helps us to overcome our biases as well as inform us how to obtain better information about the external uncertainties we face. By “negative thinking” we do not mean cynicism or toxic nay-saying. Rather, we refer to a process that asks us to consider critically the opposite of what we too easily assume (or wish) to be true. While Negative Thinking could lead us to consider the effects of unfortunate outcomes or conditions (the opposite of desired outcomes) on sales opportunities…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/521/0*4Z-T3_24L7z-Y3Z5.jpg" /><figcaption>The best laid schemes o’ Mice an’ Salesmen, Gang aft agley</figcaption></figure><p>…it could also lead us to consider the possibility of desirable outcomes or conditions (the opposite of the unfortunate) for situations that we often easily dismiss.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/634/0*p5PrVkFgl1_8dORD.jpg" /><figcaption>No, no, boy, that’s no way to make a plane. That’ll, I say, that’ll never…fly!</figcaption></figure><p>But the Power of Negative Thinking goes beyond our merely considering what can happen. We must also consider the “why” and “to what degree” those things could happen. We can account for the “what,” “why,” and “to what degree” in a process called probabilistic reasoning. But that’s the second step. The Power of Negative Thinking begins with accurately framing an opportunity, which requires that a sales team answer the following questions:</p><ul><li>What is the real opportunity?</li><li>What are our goals and objectives?</li><li>What are the client’s goals and objectives?</li><li>What are the decision boundaries and open decisions?</li><li>What are the sources of uncertainty?</li></ul><p>Answering these questions helps the team know that it has the right reasons in mind to pursue an opportunity and what constraints in their current level of knowledge limit their ability to make unambiguous decisions about what opportunities to pursue and how to go about pursuing them.</p><p>Probabilistic reasoning helps a sales team then answer these questions:</p><ul><li>What is the likely range of outcomes for the uncertainties?</li><li>What are the effects of uncertainties on sales goals, revenues, and profit?</li><li>How much risk do we face with each opportunity; i.e., how much could we lose by pursuing one opportunity over another?</li><li>What insights can we create for contingency plans or options?</li><li>How do we prioritize our set of current opportunities?</li></ul><p>The effect of taking these two steps in a structured way reveals the Power of Negative Thinking so that the sales team can recognize when an opportunity is worth pursuing…or not. Ultimately, not only does the Power of Negative Thinking give the sales team a more accurate assessment of the current state and possibilities they face, they can also develop more effective contingency plans to increase the likelihood of achieving results their organization — and their clients — desire.</p><p><em>Originally published at </em><a href="http://thalespress.blogspot.com/2016/11/the-power-of-negative-thinking-how-do_64.html"><em>thalespress.blogspot.com</em></a><em> on November 9, 2016.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6b702ad24c86" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>