<?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 Michelle Brush on Medium]]></title>
        <description><![CDATA[Stories by Michelle Brush on Medium]]></description>
        <link>https://medium.com/@michelle.brush?source=rss-d7074eb8f1cd------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*MrWHvjuVmmtcqANklbiJZQ.jpeg</url>
            <title>Stories by Michelle Brush on Medium</title>
            <link>https://medium.com/@michelle.brush?source=rss-d7074eb8f1cd------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 08:48:19 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@michelle.brush/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[Manage Risk, Not Buffers]]></title>
            <link>https://medium.com/@michelle.brush/manage-risk-not-buffers-fa8157fd20e?source=rss-d7074eb8f1cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/fa8157fd20e</guid>
            <category><![CDATA[agile]]></category>
            <category><![CDATA[project-management]]></category>
            <dc:creator><![CDATA[Michelle Brush]]></dc:creator>
            <pubDate>Wed, 11 Apr 2018 17:28:32 GMT</pubDate>
            <atom:updated>2023-11-17T21:32:01.795Z</atom:updated>
            <content:encoded><![CDATA[<p>I used to be habitually early for flights. It drove some of my friends nuts. Conversations about carpooling to the airport would go something like this:</p><blockquote>“Do you want to carpool to the airport?”</blockquote><blockquote>“Yes, but I like to leave 3 hours before my flight.”</blockquote><blockquote>“Nevermind.”</blockquote><p>My friends are the kind of people that like to time their travel such that they arrive at the gate just as boarding starts. I think my friends take on too much risk. They think I plan too much for it. They are probably right.</p><p>Now, I don’t take the same excessive risk management approach to an 8:00AM meeting. I don’t show up at 6:30AM just to be safe. I leave at a time that will allow me to manage 75% of the local traffic scenarios and still get to work on time. If there is some outlier event, I have my phone so that I can call people and tell them I’m late.[1]</p><p>There are many reasons I treat these two scenarios differently. One is that the impact is different. Missing a flight is a huge inconvenience that can disrupt the rest of the trip. Being 15 minutes late to a meeting is an inconvenience that can be managed by an apology and a schedule change.</p><p>Another reason I treat them differently is that there are more things that can go wrong on in a trip to the airport. Yes, as with the meeting, a traffic incident could delay me, but also, the parking lot near the terminal could be full. The airport shuttle could be late. There could be long lines at security. I could be randomly selected to be searched.</p><p>We can agree it would be silly for me to treat both scenarios with the same arbitrary rule like, “Be sure to leave 3 hours before the event,” or “Always double the travel time.”</p><p>The reason I’m talking about all of this is to set the stage to talk about something I see a lot in software project management. I want to talk about estimation buffers.</p><p>When I talk about buffers, I mean the rules that some engineering folks use to add arbitrary time to their work estimates.[2] One of my former colleagues doubled any estimate he would give his boss. I also knew a manager that added a flat 4 weeks to any number she heard from her team. Another colleague only plans 8 weeks of work for every 12 weeks of time.</p><p>All these buffers are meant to address estimation error. The idea is that if something goes wrong, you’ve got room in your schedule to recover. However, schedule slips are not usually the result of small estimation error. They are typically the consequence of events that revealed our model for thinking about the work was incomplete or ill-fitting. [3]</p><p>I like to think of it in terms of the <a href="https://medium.com/lightspeed-venture-partners/knowns-vs-unknowns-78b0da5ca887">knowns and unknowns</a> of a project. People are padding to leave room for a “known unknown” to actualize and significantly impact the timeline, but padding only helps with known unknowns. It doesn’t account for unknown unknowns. The <a href="https://en.wikipedia.org/wiki/Ninety-ninety_rule">Ninety-ninety Rule</a> captures our feeling that this occurs a lot:</p><blockquote>The first 90 percent of the code accounts for the first 90 percent of the development time. The remaining 10 percent of the code accounts for the other 90 percent of the development time.</blockquote><blockquote>— <a href="https://en.wikipedia.org/w/index.php?title=Tom_Cargill&amp;action=edit&amp;redlink=1">Tom Cargill</a>, <a href="https://en.wikipedia.org/wiki/Bell_Labs">Bell Labs</a></blockquote><p>Buffers are meant to give us wiggle room to deal with all flavors of risk but I rarely see a project that given a buffer, still gets done on time, with the full original scope, within the normal definition of quality, without late nights and long weekends, and within budget.</p><p>This shouldn’t be a surprise. It was captured by <a href="https://en.wikipedia.org/wiki/Hofstadter%27s_law">Hofstadtler’s Law</a>.</p><blockquote>Hofstadter’s Law: It always takes longer than you expect, even when you take into account Hofstadter’s Law.</blockquote><blockquote>— <a href="https://en.wikipedia.org/wiki/Douglas_Hofstadter">Douglas Hofstadter</a>, <a href="https://en.wikipedia.org/wiki/G%C3%B6del,_Escher,_Bach:_An_Eternal_Golden_Braid"><em>Gödel, Escher, Bach: An Eternal Golden Braid</em></a></blockquote><p>So why is that? Well, for one, the buffers themselves are usually victim to the same bad assumptions that triggered the original estimate. We assume a linear impact to effort for something that is actually polynomial. We discount the likelihood of an event.</p><p>However, there’s something else going on as well. I genuinely believe the mere presence of the buffer gives people unwritten permission to bleed into it, and in some cases, past it. The buffer becomes a self-fulfilling prophesy.</p><p>This is because of 3 things that happen when buffers are added to project plans:</p><ol><li>During the project, people don’t estimate the impact of a newly discovered risk. They just assume the buffer covers it.</li><li>People double-dip the buffer. Multiple things go wrong, but people assume the buffer covers all of them.</li><li>People lose their sense of urgency or cadence to the work because they perceive they have more time than they need.</li></ol><p>The third point is what worries me the most. There’s another old adage, <a href="https://en.wikipedia.org/wiki/Parkinson%27s_law">Parkinson’s Law</a>, that says, “Work expands to fill the time allotted.” It captures that when we have too much time or resources to do something, we sometimes take on additional lower priority work. This lower priority work might be rewriting something that doesn’t need to be rewritten right now. It might be sneaking in a few delighters that weren’t planned. It might be agreeing to do a bunch of side projects in parallel with the project. No one plans this work. It just happens. Unfortunately, the unplanned work can easily disrupt the whole plan.</p><p>Getting back to my airport habits, sitting at the airport is a pretty low priority for me. I’d rather be sleeping. I’d rather be hanging out with my family or my friends. I’d rather be grabbing a nice lunch or site seeing.</p><p>I’ve been trying to get better. I’ve been trying to make decisions just in time. That means at the extremely early time I used to leave for the airport instead I check online for traffic delays, security lines, and parking lot capacity. I factor all of that information into deciding a new time to leave. If the parking lot is 95% full, I leave a littler earlier. If it’s only 50% full, I leave a little later. If security lines are short, I leave even later. If things are going well, I get to stay home and do higher priority things. I’m actually managing the risk.</p><p>If and when I do add extra time to my travels, I know why I’m doing it.</p><p>This is how we should manage risk in software projects as well. We should have a range of possible timelines that are updated as we evaluate and resolve some risks.</p><p>Some projects are predictable like early morning commutes. You can be optimistic as long as you have a mitigation plan should things go wrong. Some projects are like trips to the airport, there are a lot of variables that could disrupt the plan. We need to favor a more pessimistic plan, and front load development activities that will inform us of the challenges. In both cases, as we work, we monitor and respond to risks as they are realized.</p><p>We shouldn’t add arbitrary buffers and pretend we’ve solved for things we don’t understand.</p><p>[1]I have to give a <a href="https://twitter.com/BPoiry">friend</a> credit for this. She was the one I first heard draw an analogy between project planning and driving to work.</p><p>[2] Arbitrary is a key word here. If time is added to a project to account for a likely event that has a known impact, that is good planning. The time still might not be needed, but you’ve at least tagged it with a purpose.</p><p>[3] This can be quite wrong in the case of <a href="https://en.wikipedia.org/wiki/Black_swan_theory">Black Swan type events</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fa8157fd20e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Curse of Knowledge]]></title>
            <link>https://medium.com/@michelle.brush/the-curse-of-knowledge-e44e7bc91fc8?source=rss-d7074eb8f1cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/e44e7bc91fc8</guid>
            <category><![CDATA[mathematics]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[management-and-leadership]]></category>
            <dc:creator><![CDATA[Michelle Brush]]></dc:creator>
            <pubDate>Sun, 25 Feb 2018 13:22:39 GMT</pubDate>
            <atom:updated>2018-02-25T13:22:39.787Z</atom:updated>
            <content:encoded><![CDATA[<p>Recently, as I was explaining how to multiply negative numbers to my son, I realized I was explaining it exactly the way it had been explained to me. By that I mean, I wasn’t really explaining it. I was listing rules.</p><p>The experience reminded me of a time in high school when I was tutoring a classmate on the same set of rules, and he was struggling to get it. We worked for an hour on problems with me reminding him which rule applied to his situation. “A negative times a negative is positive.” I declared success when he finally memorized and applied the rules without my direction. Unfortunately, as a side effect of this experience, I probably taught him that math was a series of arbitrary rules he would apply but never understand. (I’m sorry for not being a better tutor, Matt.)</p><p>I decided I was going to do better for my son. However, to do better, I had to remember why exactly the rules made so much sense to me. I had to get around the Curse of Knowledge.</p><p>The Curse of Knowledge captures the situation when the more expertise someone has with a concept, the harder it is for them to explain it to someone else. This happens because the expert makes an incorrect assumption about what prior understanding the listener has.</p><h3><strong>Experts are not necessarily the best teachers.</strong></h3><p>Another way to reason about this is the Stages of Competence. You start out not even realizing you don’t know how to do something. You are unconsciously incompetent. Then you recognize your gaps in understanding, but don’t know how to fill them — conscious incompetence. Next you learn the right way to approach the problem, but you still have to reason about it as you are doing it. Finally, you internalize the understanding and no longer have to think about it. You’ve achieved unconscious competence.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lMg70NwpAkZPYRDF9jj1gg.png" /></figure><p>While “unconscious competence” is the peak of the triangle, people at the “conscious competence” stage tend to be better teachers. They know how to do something, but are still at the stage that they are conscious of what they are doing. They can explain it to others.</p><p>That being said, there are experts that are good teachers. I wonder if there needs to be a level above “unconscious competence” for them. It would capture the people that have both internalized what they’ve learned, but also understand it to the level that they can explain it. That’s where I needed to be to help my son with multiplication of negative numbers.</p><p>I needed to roll back some of my unconscious competence. I decided I would try to derive the behavior from things I knew about multiplication and negative numbers separately.</p><ol><li>A negative number is a number that when added to the positive version of itself, yields zero as the sum. 3 + -3 = 0</li><li>Adding a negative number is the same as subtracting the positive form of it.</li><li>Multiplication is the operation that tells you to sum a number a number of times. A x B means A added B times. 3 x 2 = 3 + 3</li><li>Multiplication is commutative, so A x B is the same as B x A.</li><li>Multiplication is distributive. A (B+C) is equal to A x B + A x C.</li><li>If you are multiplying a negative number, you still add it, but the negative is distributed. You can think of it as flipping the addition to subtraction. -3 x 2 = -1(3x2) = -1(3 + 3) = -3 - 3</li><li>If you have 2 negative numbers, you are distributing the negative twice. You are now subtracting negative numbers. -3 x -2 = (-1)(-1)(3 x 2) = (-1) (-1)(3 + 3) = -(-3) -(-3)</li></ol><p>After this, it’s now obvious that multiplication of two negative numbers would yield a positive one. Multiplication of a negative times a positive would be negative, and so on. The rules make sense.</p><h3><strong>How many other rules do we cite without thinking?</strong></h3><p>Often when working on a design of a new API or new system, we have an intuition when something is good or bad. We attempt to codify that intuition into rules like “Coupling is bad. Cohesion is good.” Sometimes we use more complicated rules like SOLID. However, if we’re honest, we didn’t arrive at our judgement through rules. There’s a deeper understanding that we are unable to put into words. The rules are an attempt to retrofit an explanation onto our feelings.</p><p>I’ve taken to calling that intuition “taste.” That’s a cop out on my part. It allows me to pretend you either have it or you don’t. It relieves me of having to explain it. However, to be an effective technical leader, we have to teach others how to see the system in the same way we do. We need to explain it. That may mean revisiting our own understanding of the system, and that’s hard work.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e44e7bc91fc8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Death of Cannot Reproduce]]></title>
            <link>https://medium.com/@michelle.brush/the-death-of-cannot-reproduce-d3793033caca?source=rss-d7074eb8f1cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/d3793033caca</guid>
            <category><![CDATA[software-testing]]></category>
            <category><![CDATA[reproducibility]]></category>
            <category><![CDATA[software-architecture]]></category>
            <dc:creator><![CDATA[Michelle Brush]]></dc:creator>
            <pubDate>Wed, 18 Oct 2017 15:42:24 GMT</pubDate>
            <atom:updated>2023-11-17T21:26:02.241Z</atom:updated>
            <content:encoded><![CDATA[<p>I used to write code for automotive navigation devices. In one of my most epic troubleshooting and debugging moments of my career, I sat outside the house of my CEO. I had the device mounted on the dashboard, one hand on the wheel, and the other hovering over the “Calculate” button. I floored the car’s accelerator while simultaneously hitting the button as fast as I could, to prove that I could, in fact, reproduce the issue he reported, if and only if, I drove like a madman.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*J1kEX710Eja4pWPofL-HbQ.png" /></figure><p>It worked.</p><p>Once I reproduced the issue, I knew I could fix it because I had a model in my head of what had to be happening in the system. My experiment proved that model was correct.</p><p>The navigation device in question was a comparatively simple system. For starters, it was a single computer, and therefore, not distributed. It was not a connected device. On top of that, there were limited inputs which I could iterate through in a loop; tightly controlled configurations and environment variables. The whole system was updated en masse, eliminating risk of permutations of dependencies that had never been tested before. The hardware was fixed and manufactured by the company itself. It was controlled.</p><p>Rarely do we get this level of control of our systems. This is why if you search something like the Apache Software Foundation’s issue tracking system, you find over 12,000 were closed as “Cannot Reproduce” as of the writing of this blog. It is a sign that the engineer has waved the white flag and conceded the issue was either woefully documented by the reporter or completely and totally dependent on some system state or configuration that is beyond comprehension. We don’t know what to do.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PupjATkYtLgQKDPyEmasng.png" /></figure><p>It is frustrating to be on either side of the “Cannot Reproduce” conversation. I remember once being told by the corporate IT department that they could not reproduce the issue I was having with the corporate-approved VPN software. They told me it must be something wrong with my machine. I ended up systematically uninstalling, reinstalling software until I found that that the version of a PDF viewer I had on machine conflicted in some strange way. It was monotonous and infuriating experience, but I prevailed.</p><p>Most developers have experienced this at some point. If you work with commercial desktop software, you may have no idea what version of operating system, office productivity software, messaging applications, and sadly, malware will be installed on your customer’s device. If you work on web applications, you might worry about browser settings, versions, ad blockers, plugins, and bandwidth issues. If you work in the cloud, well, you might not know how much memory the machine on which your software is running really has or where it is located.</p><p>Early in my career, one of the challenges to reproducing the issue was knowing the steps to recreate it. We had a separate QA organization that would execute test plans manually and report the issues found. Bug reports would come in missing essential information. The first hours or days of troubleshooting were delayed in the back and forth of asking for more information.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zoUzhAkcA2TZZMZv_Hj0Qw.png" /></figure><p>Then Kent Beck and Erich Gama wrote JUnit on a plane to Zurich. While automated testing surely existed before that point, with JUnit testing became this immediately, efficient and repeatable aspect of our software development world. We no longer had to worry about missed testing steps, missed test cases, or wholly missed tests. The computer did it for us sometimes at night while we sleep. Now when we engage with Open Source projects, we can offer up automated unit tests alongside the bug reports. We can guarantee the issue will reproduced.</p><p>That was just one step.</p><p>Let’s talk about repeatable builds. One of the first things we teach engineers fresh out of school is the need to have good source code management, dependency management, and automated builds. We show them how to have the build system run their test. We teach them to stop treating their development devices as the source of truth for whether their code works. “It works on my machine,” or “It builds locally,” are things we have to break them of the habit of saying.</p><p>We get them plugged into automated integration tests.</p><p>However with case-driven or example-based tests, our tests are only as good as our imagination and the level of effort we’re willing to invest in generating test cases and scenarios. Until we take that our of our hands as well. With Property-Based testing, we describe the system in terms of the behaviors we expect. We declare the rules we want to be true, then we let the system generate test cases for us and report back which ones didn’t work.</p><p>This requires us to pause and be thoughtful about the system behavior and our expectations of the system — which we should have been doing anyway. Now our tests as good as the amount of effort we’re willing to take to make them painless to run, upkeep, and review.</p><p>Years ago, working on an embedded system, we wanted to make sure that no matter how many buttons the user pushed, no matter the order, and no matter the frequency, the system would not fail. Now obviously we couldn’t fully guarantee this. No one can prove their code is bug free. However, someone had the great idea to build a harness that would randomly fire off UI events sometimes simultaneously and sometimes in quick succession. The harness was stupidly simple. It had a registry of all possible events, multiple threads, and a random number generator. It was glorious. It discovered all sorts of hidden bugs in the system. It also taught us all the places in the software where we were blind to the state and metrics we needed to troubleshoot complex issues. Just as much as it helped us find bugs, it helped us find our blindspots in observability.</p><p>It made what would have been rare events, immediately available to us. Now that I’m dealing with thousands of commodity machines, I’ve recognized rare events aren’t rare. When we’re dealing with millions of people interacting with a system at the same time, it’s the <a href="https://en.wikipedia.org/wiki/Infinite_monkey_theorem">Infinite Monkey Theorem</a> but instead of the monkeys producing Hamlet, the odds are that that one decision you made to ignore a potential vulnerability or put off an outlier failure case will be triggered by someone. The system will melt down in a spectacular way.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*F4xhwsbmxtRsmB6rcYjTvQ.png" /></figure><p>I wrote a bug that was a 4 in 70 million chance of happening. It was in consumer electronics. I failed to handle a complicated edge case in how a graph might be constructed. When we tested this part of the system, we did it by randomly generating hundreds of thousands of test scenarios in the graph. Just before the product was available for sale but after the devices had all been boxed up and shipped to the stores, our full-scale automated tests triggered the scenario. We fixed the issue immediately and released an update. We changed our tests to run all 70 million scenarios rather than a sampling of 100,000. We had the update available before it was even possible for a user to use the software. However even with all of that, weeks after the product release, I saw a review on Amazon complaining about my bug. It was a 4 in 70 million scenario. I now live in a world where we’re controlling software version as a variable in our experiment.</p><p>This leads to environmental issues. Where I work, the phrase, “Environmental Issue” has been banned. We’re not allowed to absolve ourselves of troubleshooting because the issue seemed to go away on its own. Every environmental issue is an opportunity to reevaluate the resiliency and observability of the system. We revisit the whole of the architecture in the context of the potential for failure.</p><p>Once we’ve designed that architecture with failure in mind, we have to test that our mitigations will and do work. As that scales to the whole architecture, we need <a href="https://medium.com/netflix-techblog/chaos-engineering-upgraded-878d341f15fa">Chaos Engineering</a> which Netflix named and made famous. This allows us to forecast and simultate all sorts of outlier or anomalous events and see how the system would respond.</p><p>We’re losing the right to assume rare events will be rare. We need to build our architectures such that we can test our assumptions. Testability becomes an attribute of an architecture not just the code.</p><p>So what’s left to control?</p><ul><li>Infrastructure</li><li>Configuration</li><li>Time</li></ul><p>I used to work on a massive 3-tier architecture written predominantly in C++. It was a badge of honor that our services would run and serve requests for months at a time. It meant we wrote code without memory leaks, without memory corruptions, and with robust error handling. At this point, we weren’t hosting the services, so every client that bought our software would run it themselves using recommended operating system versions — recommended configuration. I spent a lot of hours trying to figure out what recommendations a client may have ignored thus creating the circumstances under which the hard to reproduce bug happened.</p><p>Now, the systems I build are run on thousands of commodity machines. Ensuring consistency is not a recommendation. It is essential. It is so essential, we took it out of the hands of humans. <strong>It’s now a badge of honor how quickly you can tear down and stand up your infrastructure.</strong></p><p>In science, when it comes to experimentation, there is the concept of reproduceability. It is an understanding of the repeatability of research. “How likely is it that another person can reproduce the experiment and see the same results?” It is precipitated on how well the controls can be controlled, how well the experiment was documented, and how good is the manual.</p><p>You don’t need a manual if there are no steps to take. You don’t need a manual if you can rebuild the entire system from scratch completely hands-free. You don’t have to worry about controlling inputs if you can try all possible values within seconds.</p><p>As the industry has grown (and will continue to grow) we’re approaching in a place where there should be nothing outside our control. Yes, hardware does fail. Bits can flip. Networks can experience interference. The power can go out. I recognize we’ll never be able to control everything, but we have more control than we did before. The secret has been and will continue to be, to take the management of the variables in our world out of the hands of the humans. Give it to the machines.</p><p>To do this we must think about reproducibility as an attribute of our systems and therefore our architectures. That is my call to action here. Think of your architectures as something that should offer you reproducibility.</p><p>As we get closer and closer to a world where everything we do in software is repeatable, as we treat true reproducibility as an attribute we expect of our systems and architectures, I hope, one day, no user will ever have to hear, “We cannot reproduce your issue,” especially not over 12,000 times.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/576/1*34eIfXBpQHX5WBBgGcn2-Q.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d3793033caca" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Cult of Knowing]]></title>
            <link>https://medium.com/@michelle.brush/the-cult-of-knowing-2991906407ec?source=rss-d7074eb8f1cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/2991906407ec</guid>
            <category><![CDATA[management]]></category>
            <category><![CDATA[communication]]></category>
            <dc:creator><![CDATA[Michelle Brush]]></dc:creator>
            <pubDate>Mon, 16 Oct 2017 10:08:17 GMT</pubDate>
            <atom:updated>2017-10-16T10:26:01.890Z</atom:updated>
            <content:encoded><![CDATA[<p>Early in my career, I attended a leadership class that focused on recognition. The thing that stuck with me is that we naively assume people want to be recognized by money and authority, and that is not true for the majority of people. The class had us reflect on the different ways we’d want to be recognized. You can be recognized by being given increasingly challenging problems to solve. You can get excited about having more autonomy to solve a problem. You can feel appreciated by getting the opportunity to share what you learn with others. You can be given public recognition for your efforts.</p><p><strong>It’s not a bad thing to want to be recognized</strong>[1]<strong>, and recognition is something that works best when personalized</strong>[2].</p><p>Something I had never considered before that class was that some people want to be recognized by being in the know. They feel appreciated when they feel they have transparent leadership. They feel recognized when they have information when others have it. That’s not a bad thing. It should be relatively inexpensive and easy form of recognition to provide in a transparent company.</p><p>What if the company is not transparent? What if knowledge becomes a political currency in the organization? You’ll see the rise of knowledge cliques. There will seem to be groups of people that know things — the inner circles. A <a href="https://twitter.com/BPoiry">friend of mine</a>, who is a former English teacher, calls this the “<a href="https://www.youtube.com/watch?v=vZpt1lHYTNE">cult of knowing</a>.” How do you know if the cult of knowing is happening in your organization? Ask yourself a few questions:</p><ul><li>Do you find that you have to leverage the grapevine to discover context?</li><li>Are you told not to share context with others you feel would benefit from it?</li><li>Is it hard to reason about who knows what, when, and why?</li></ul><p>If you answered “yes” to the above questions, it may be worthwhile to honor <a href="https://en.wikipedia.org/wiki/Hanlon%27s_razor">Hanlon’s Razor</a>, “Don’t assume malice for can easily be explained through incompetence.” It could mean your organization is bad at communication. I once worked for an organization that was so bad at communication, they sent out an email announcing an reorganization to employees with the line, “If you have any questions about this change, discuss it with your manager.” The problem was the only communication to the managers had been that same email, nothing else. They had no means to answer any questions their employees would have.</p><p>There’s good news for small, colocated organizations. They don’t have to work hard to be good at communication. They can communicate via a series of incidental interactions: hallway chats, people popping by each others’ desks, and casual conversations in the spaces outside meetings. <strong>Small, colocated organizations can manage through proximity.</strong> They can just sit near each other and share the right information just in time.</p><p>The same tricks don’t work for large organizations. Large organizations have to be intentional about communication. They have to identify audiences. They have to understand how those audiences should receive information. They have to send the same message through multiple channels. My English teacher friend says, “This is why no one airs a commercial just once.”</p><p>However, this competent and intentional approach to communication is what often leads to a cult of knowing. The people planning the communication have the knowledge. <strong>They don’t know what it feels like not to be in the know.</strong> On top of that, it is their job to decide who knows what, when, and how. Once that’s their job, they take it seriously. They make careful decisions. They control the flow of information. With that serious care and control, comes a loss of transparency.</p><p>I started this blog talking about recognition. The loss of transparency through the rise of the cult of knowing doesn’t just hurt your ability to make people feel appreciated. It hurts your organization’s ability to act fast at lower levels of the company. <strong>When people are missing context, whether through organizational incompetence, bureaucracy, or malice, they either make bad decisions or stop making them altogether. </strong>Then you cease to have a reason to recognize them.</p><p>[1] I’ve heard some people claim wanting credit or recognition for your contributions is a bad thing. I think that stance fundamentally ignores a basic human need for appreciation. Not giving others credit or showing appreciation for their work is what I would consider the dysfunctional behavior.</p><p>[2] In case you are curious, I prefer recognition to be in the form of increasingly challenging problems for which I get a bare bones “Thank you” from leadership for solving.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2991906407ec" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[When Ideas Are the Problem]]></title>
            <link>https://medium.com/@michelle.brush/when-ideas-are-the-problem-bdc394920667?source=rss-d7074eb8f1cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/bdc394920667</guid>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Michelle Brush]]></dc:creator>
            <pubDate>Fri, 15 Sep 2017 20:39:11 GMT</pubDate>
            <atom:updated>2017-09-19T01:58:58.800Z</atom:updated>
            <content:encoded><![CDATA[<p>A few years ago I found myself working more with people whose backgrounds were in marketing. I decided to take a <a href="https://www.coursera.org/learn/wharton-marketing">marketing class</a> to better understand their space and better communicate about problems and solutions. I loved the class. It gave me a new perspective on how to approach software design and development.</p><p>It’s hard to distill a great class down to one simple lesson, but if I did, I would say that too many people think marketing is selling a product after you’ve built it. Marketing starts before you build the product. It starts with identifying whether you even have a market and if so, on what axis are you going to distinguish your product in that market.</p><p>That last part was the most powerful for me. I had worked in a lot of environments where the implicit expectation was that the product would be the best at everything: highest quality, lowest cost, greatest number of features, prettiest and most usable interface, and so on. The class taught me that it doesn’t make sense to try to be all the things to everyone. You have to have a distinguishing attribute. <a href="https://digitalintelligencetoday.com/the-jam-study-strikes-back-when-less-choice-does-mean-more-sales/">This is because people are bad at making choices when there are too many variables to consider and everything looks roughly the same.</a></p><p>I learned you need a product vision that allows you to say no to some investments, some ideas. If you try to be everything to everyone, you’ll always be behind the market on any given axis. You only have so much capital or capacity for work, and whatever you don’t spend your resources on, a competitor will.</p><p>This means a good product vision has to not just tell you what to do, but what not to do. This isn’t to say it should be negative: “Our software will never…” You just need it to create dividing line so you know what is essential for that vision and what is not.</p><p>A simple example would be something like, “Banking where you are when you need it.” This implies the software will favor accessibility, uptime, and mobile availability. It won’t emphasize sophisticated budgeting and reporting features. It won’t fall victim to <a href="http://www.catb.org/jargon/html/Z/Zawinskis-Law.html">Zawinski’s Law</a> and start reading email.</p><p>If your product vision isn’t allowing you to say no to some features, it probably needs some work.</p><p>Even with a great vision that allows people to focus on core aspects of the product, teams may find there are still way too many ideas when compared to the capacity of the team. How do they deal with the volume?</p><p>I’m going to reframe the problem from a systems thinking perspective. In systems thinking you can have flows or faucets that are producing things. You can have stocks that accumulate things, and you can have feedback loops or controls that affect those things.</p><p>Let’s say we have a flow of ideas. These ideas may come from product managers, customers, business analysts, or any where. I’m going to abstract their titles to a single role: idea generator.</p><p>Then you have software engineers, architects, system engineers, and so on. I’m going to call them idea implementors. (I’ll acknowledge the idea generators can be idea implementors and vice versa.) They consume the flow of ideas.</p><p>An idea generator’s rate of ideas — meaning how fast they can put an idea in the flow — is limited by their ability to form an idea and the effort cost of expressing the idea.</p><p>The idea implementors rate of production is limited by their ability to create the software and the definition of done for the idea. The definition of done informs the effort cost of implementation.</p><p>If the implementors’ rate of output is lower than the rate of the generators, you get a backlog. If the implementors are much, much slower than the generators, you may overflow the backlog. Great ideas are being lost in the noise. This is not to say the change tracking system can’t scale, but rather that the humans curating and triaging the backlog can’t.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4D98q4he0pcCBhkbhEftZQ.png" /></figure><p>How do you keep the rate of generation in line with the rate of implementation? You can scale the implementors by adding more engineers. You can change the definition of done — sacrificing quality to favor volume of features and time to market. You can invest in platforms, tools, or SDKs that dramatically improve the productivity of the implementors. If you believe in <a href="http://www.effectiveengineer.com/blog/how-to-become-a-10x-engineer">10x engineers</a>, you can recruit and pay for them.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*klEzg3j3pI9XoSClDjgSqA.png" /></figure><p>However, this assumes the problem is on the implementation side. It assumes most the ideas in the backlog are good and should be implemented. Are they good? Do they all fit the product vision?</p><p>Probably not.</p><p>So let’s address the flow on the idea generator side. What can you do? You can change the definition of done for an idea. You can raise the bar on idea quality.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3zaTffTdKJ7nbT0nyZib3g.png" /></figure><p>I worked for a company where ideas were sent to the engineering teams as single sentences via email. It cost the idea generators two minutes to submit an idea. There was constant friction between the implementors and the generators. The generators perceived the implementors as indifferent and slow. The generators were perceived as unreasonable. The implementors were losing productivity responding to escalations and timeline questions for way too many ideas. Sometimes competing escalations would cause the team to thrash, losing even more productivity. Fewer and fewer good ideas got implemented.</p><p>Lower productivity was the natural consequence of the bar for idea submission being too low.</p><p>In organizations that value product management as a discipline, ideas have a higher bar. Ideas are not done without market analysis that identifies and describes the target audience and why they would care. They are not done without clear evidence of how the idea advances the product vision. Ideas are not done without documented ROI. Changing the definition of done to include these things not only slows the rate, but it also ensures the organization has all the context needed to understand the idea enough to evaluate it.</p><p>Sadly, I’ve seen organizations go in the wrong direction in solving the overflowing backlog of ideas problem. There’s an antipattern where the idea generators’ response to the perceived slow throughput of implementation is not to craft their ideas more carefully, but to take on some of the implementation work. They try to offer armchair designs. They skip the idea evaluation step and describe the idea in terms of the software architecture and not the market. “Why don’t you just…?”</p><p>It’s easy to see how this happens. People love their ideas. Analyzing a great idea for value seems like a waste when you can just roll up your sleeves and get started. If the implementors are perceived as stalling, the generators think they can show them how easy it can be.</p><p>However, if the idea generators are describing the system, who is making sure the organization is solving the right problem at the right time? Who is doing the analysis to make sure the idea really has value?</p><p>And what are the implementors supposed to do when the best part of their job is done for them?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bdc394920667" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[We’re not even working on the same plane.]]></title>
            <link>https://medium.com/@michelle.brush/were-not-even-working-on-the-same-plane-a3ee55399713?source=rss-d7074eb8f1cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/a3ee55399713</guid>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[software]]></category>
            <dc:creator><![CDATA[Michelle Brush]]></dc:creator>
            <pubDate>Sun, 10 Sep 2017 23:48:18 GMT</pubDate>
            <atom:updated>2017-09-10T23:48:18.446Z</atom:updated>
            <content:encoded><![CDATA[<p>In geometry a plane divides a 3-dimensional space the same way a line divides a 2D one. It severs the space into two regions with points falling to either side. In software engineering, we have <a href="https://en.wikipedia.org/wiki/Project_management_triangle">The Project Management Triangle</a> as one such 3-dimensional space in which we work.</p><p>The PMT says, a project can be fast, big, or cheap but you can only have two out of the three. (It assumes you agree that you want quality.) Pick two. That’s the simplest form. We call it a triangle, but It’s more like a 3D space in that you are optimizing across the dimensions. Advancement in one dimension will make you pull away from one or both of the others. You just can’t have it all. The two that you pick form a plane.</p><p>What happens when people on a team or large organization are not aligned on where a project should fall in that space? What if they’re working on different planes? Dysfunction.</p><p>Let’s say we have two teams contributing to a project. One team cares a lot about time to market and low cost. One cares about delivering on all the scope with low cost. They all agree all things are important, but they have differing emphasis. The time to market team’s desire to move quickly will antagonize the scope team. The scope team will be perceived as being obstinate in their assessment of what is minimal and viable for the product.</p><p>This assumes we agree on the definition of quality. What if we don’t? At this point, the triangle becomes a diamond. We can now disagree on the priority of four dimensions.</p><p>Recently, my team was asked to essentially copy and paste some code from one part of a project to another to support a new capability. The person asking wanted fast time to market and low cost. They wanted the scope controlled to only the immediate need. I was concerned with code maintenance. I commented that refactoring was in order. It wouldn’t be as fast as originally thought. To be clear, the refactoring wouldn’t be slow. It just wouldn’t be as fast.</p><p>I explained the reasoning for spending a little more time to refactor. I brought up the potential for future reuse in other areas in addition to improving the code for the current need. In the course of the conversation, it came up that there was going to be a near-future request to replicate the behavior to a third part of the system. I stressed that the fact that we were discussing copying and pasting at all was a design “smell” and we should spend an extra week to invest in making the next request unnecessary.</p><p>The response I got back, “I don’t see why you can’t just do it.” At first I was frustrated. I thought my opinion was being summarily dismissed. Then I realized, the person said that, meant it. He couldn’t understand where I was coming from. There was an immediate need. It had an immediate solution. Any discussion of even a minor delay to favor code quality or future scope was ridiculous in his mind.</p><p>You may be wondering how this resolved itself. Well, my team was responsible for the code, so when we do work, we do it by our value system. This did lead to frustration that we tried to smooth over by being completely transparent about our goals with the work while also being sympathetic to the immediate need.</p><p>I’m not going to pretend my team always maximizes for software quality. We have found ourselves on the other side of the plane. We have worked with teams that favor abstraction and investment in design at levels above us. We’ve experienced our own frustrations.</p><p>The sad thing about these conversations — whichever side of the plane we’ve found ourselves on — is that we almost never feel open to admitting that it’s all a compromise. As engineers, we often moralize our definitions of quality, cost, scope, or timeliness. We have pride in the dividing planes we’ve drawn about what we value and what we don’t.</p><p><strong>“I don’t see why…” is a moral judgement.</strong></p><p>Instead of standing our ground and refusing to compromise, we need to force ourselves to brainstorm options and map them into the space collectively. We need to structure the conversation as having no right answer. Ultimately, it is an optimization problem, and we have to try to see it that way.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a3ee55399713" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Tyranny of the Middling Solution]]></title>
            <link>https://medium.com/@michelle.brush/the-tyranny-of-the-middling-solution-8bcc1d377470?source=rss-d7074eb8f1cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/8bcc1d377470</guid>
            <category><![CDATA[programming]]></category>
            <dc:creator><![CDATA[Michelle Brush]]></dc:creator>
            <pubDate>Sun, 30 Jul 2017 23:46:10 GMT</pubDate>
            <atom:updated>2017-07-30T23:46:10.656Z</atom:updated>
            <content:encoded><![CDATA[<p>I never really got Goldilocks. Was it a lesson saying you should settle so you can eat and sleep before the bears come? Was the lesson to not steal from bears? Most people take away the lesson against seeking perfection.</p><blockquote>“Perfect is the enemy of good.” -people paraphrasing Voltaire</blockquote><p>I was working at an embedded systems company when my manager asked that I speed up an underperforming legacy algorithm. He set a performance target and asked for a plan to meet it. Since the algorithm’s author had long left the company and no one had profiled the things in years, I proposed essentially the following plan:</p><ol><li>profile it</li><li>fix the obvious bottlenecks</li><li>profile it</li><li>if no remaining obvious bottlenecks exist, alter algorithm</li><li>otherwise, GOTO 2</li></ol><p>I estimated we’d need a month of profiling and fixing bottlenecks before we’d know if we’d need to revisit the essence of the algorithm. If we did have to revisit it, I estimated 3 months of research and development would be required.</p><p>My manager didn’t like my plan. He thought the profiling was a waste of time because he felt we knew what the problem was. We knew where I/O-bound reading data needed to inform the algorithm. He wanted to just cache a subset of that data in memory and move on. He felt the cache work would take a couple of months, beating my plan. I disagreed.</p><p>For one, I felt he was wrong about which subset of data to cache, and I need to profile the algorithm to show him this was the case. Second, the way our users used the algorithm, I didn’t expect the cache to have a high hit rate. I held my ground.</p><p>The great thing about this manager is even when we disagreed technically, he still gave me complete autonomy. He didn’t take a command and control approach and order me to follow his direction. That isn’t to say he bought into my plan. He just let me run with it. To make himself feel better, he would stop by my office more frequently to hear how I was progressing. He wanted to make sure he could steer me away from my plan if there was evidence he was right, but he didn’t push me.</p><p>I was right. Profiling showed we had tons of performance problems unrelated to reading the data. It was clear the I/O in question was a smaller percentage of our performance than we had assumed. We would not get us where we needed to be with caching.</p><p>Why was the cache idea so appealing? I think partly because it landed in the middle of two options.</p><p>A) Profile &amp; tune: 1 month</p><p>B) Cache: 2 months</p><p>C) Rewrite: 3 months</p><p>It was so easy as to be suspicious of option A. It was too cheap. My manager didn’t believe we could buy any significant performance from simple low-level tuning of the code. However, we were able to shave off 30% of the algorithm running time. (Donald Knuth would have been proud as the problem code was a collection of poor attempts at premature optimization that had actually made things worse.)</p><p>Option C, rewriting the algorithm, seemed risky and expensive. No one wanted to do it. As a result, caching <em>seemed</em> “just right.” But data showed it wasn’t. We did end up rewriting the algorithm after all.</p><p>I’ve moved away from embedded systems and now work on large distributed ones. But the goldilocks situations still present themselves. For example, recently, my team realized the YAML-based configuration in our system was unwieldy and prone to mistakes. This meant that a small number of experts on the team were the only ones trusted to change anything which was a bad situation. The cheap fix was to create configuration templates that people would copy and paste with comments of what to touch and what not to touch. The medium cost fix was to refactor the configuration such that when performing operations people would interact with a smaller subset of the configuration. The right fix was revisit all the configuration, determine what was essential, remove what wasn’t, and build domain-driven services that exposed the configuration more intuitively.</p><p>We went with the refactoring approach thinking it was good enough to get us out of the mess. We didn’t get out of the mess. People still fat-fingered things in the smaller subset we exposed and triggered incidents in the system. We wished we had argued for the expensive option. The problem is, it felt hard to justify now. We had spent some development effort on the refactoring, and the result of that effort was a reduction in the number of incidents. From an ROI perspective, you could argue it was a success. You could argue it was good enough, not perfect.</p><p>Like the wines in the middle of the wine list, the middling solution is tempting. No one will be accused of skimping on the problem, and the risk of failure is smaller. Why not take the middle option?</p><p>There’s a tyranny in the middling solution. It’s just good enough to work that it will be hard to justify it’s eventual replacement.</p><p>The good news is that we’ve decided not to stick with good enough. We are embarking on the domain-driven services work now. This is again, due to having leadership that is willing to trust the team when we say we need to do something different. However, going back, I wished we’d taken a different approach initially. I wish we had created similar plan to the algorithm performance work I described. That was, I wish that instead of picking the middle option first, we had done the cheap one to buy us just enough time to do the expensive one.</p><p>You may ask, “Why not just to do the right one upfront?” We were in operations hell. Part of the appeal of the middling solution is its lower cost. When the team is feeling pain whether from operations or market pressure, faster always feels better. People don’t feel they have time to do it right. We can say in hindsight, “Well, did they have time to do it twice?” but that ignores the emotional component to the decision-making.</p><p>I’m arguing you may have to do it twice. Once, as cheap as you can to buy some time and then twice to do it right. Goldilocks didn’t go straight to the right bed, the right porridge. Even she had to try a couple of options first.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8bcc1d377470" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Opportunity Is Not a Zero Sum Game]]></title>
            <link>https://medium.com/@michelle.brush/opportunity-is-not-a-zero-sum-game-692759276db6?source=rss-d7074eb8f1cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/692759276db6</guid>
            <category><![CDATA[management-and-leadership]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[management]]></category>
            <category><![CDATA[organizational-culture]]></category>
            <dc:creator><![CDATA[Michelle Brush]]></dc:creator>
            <pubDate>Sun, 16 Jul 2017 23:52:32 GMT</pubDate>
            <atom:updated>2017-07-16T23:52:32.399Z</atom:updated>
            <content:encoded><![CDATA[<p>I’m going to start with the following premise:</p><p><em>Most organizations have a large number of unsolved problems.</em></p><p>In fact, most organizations have more problems than they have people with the focus, resources, and skills to solve. Hard problems create opportunity for people to be creative and demonstrate their skills. The hardest problems tend to require people from multiple parts of the organization buying into a solution. This means problems can create opportunity for visibility and career development. <strong>Problems have value.</strong></p><p>I’ve been in multiple organizations where I’ve seen these hard, cross-organization problems attempted to be solved one of two ways:</p><ol><li><em>Form a committee that is representative of all the people impacted by this problem and come to an agreement through meeting after meeting. (This is “death by committee.”)</em></li><li><em>Identify one single individual and assign him responsibility for the problem. Tell him to solve it as a “stretch” or “reach” opportunity to demonstrate his potential for career development. (This is “sink or swim.”)</em></li></ol><p>I’m going to propose that the way hard problems in large organizations actually get solved is secret option 3:</p><p><em>Identify a small but diversely skilled group of people passionate about the problem space and let them have </em><strong><em>focused</em></strong><em> time to work on a solution. Sponsor and champion their work. Give them access to stakeholders and a feedback loop, and help them sell the solution.</em></p><h3>What’s wrong with the wrong approaches</h3><p>I’m not going to discuss the “death by committee” approach in detail. You would think it would be obvious to organizations that large committees fall victim to the <a href="https://en.wikipedia.org/wiki/Diffusion_of_responsibility">diffusion of responsibility</a>. The only reason I can fathom for why a company would take this approach is that the quality of the solution is less important than the perception of consensus in the organization. Basically, the organization values individual representation over contribution.</p><p>If we ignore option 1, the interesting question is “Why do companies do option 2?” Why is “sink or swim” still prevalent?</p><p>I was once in a meeting where I was asked to give feedback on a proposed solution to a problem. This problem had been given to a lone individual as a “stretch” opportunity. The individual wasn’t in the room when the solution was presented. (His manager was presenting it for him.) I pointed to some gaps in the approach and suggested more aspects of the problem needed to be considered. Later, his manager told me he hadn’t wanted to provide my feedback to the individual because it wouldn’t be giving this individual an opportunity to grow. The manager was concerned too much feedback would stifle the person’s development. He wanted him to figure it out on his own.</p><p>Separately, I’ve been in other conversations where people have said, “We need to give others in the organization opportunities so I’d rather we didn’t have so-and-so participate,” and “We keep going to the same 10 people. It’s someone else’s turn.”</p><p>Based on these comments and several experiences, I would argue the “sink or swim” approach survives because of two faulty beliefs and one sad reality:</p><h4>Faulty Beliefs</h4><ol><li>Opportunity is a scarce resource that must be doled out on an individual by individual basis via some sort of resource management algorithm.</li><li>The way someone grows or develops is by being given a really big challenge with little guidance.</li></ol><p>Both are false. More than that, both are harmful.</p><h4>Sad Reality</h4><ol><li>It’s really hard to align problems with the people ready to solve them.</li></ol><p>This one is true.</p><h3>Opportunity is not a scarce resource.</h3><p>The sentence, “We need to give others in the organization opportunities,” implies an opportunity can only be owned by one person. The sentence also implies that there is a sense of unfairness with the same individual getting multiple opportunities. It makes the company sound like a zero-sum game, not a functioning organization aligned on values and behaviors.</p><p>I started with the premise that most organizations have more problems than people that can solve them. If problems are opportunities, this means opportunity is not scarce. If it is not scarce, there should be no problem with the same person solving multiple problems.</p><p>Recently, I was talking about the challenges of delegating effectively with a leader in a large organization. He complained no one had context to solve the problems he needed solved and he didn’t have the time to share context. As a result, he ends up either ignoring the problem or solving it himself. The supply problem was on the people side, not the problem side.</p><p>The real underlying problem is that the market for opportunity is inefficient due to lack of information. People with the right skills and passion aren’t aware of the opportunity or they aren’t visible to leadership. To deal with this market inefficiency, organizations should not implement some arbitrary <a href="https://en.wikipedia.org/wiki/Round-robin_scheduling">Round Robin</a> algorithm for managing opportunity.</p><p>They should describe the qualities needed in a person to solve a particular problem and do a broad search for people that have those qualities. In the delegation example mentioned above, I suggested that the leader look for people that are good context-seekers.</p><p>Unsurprisingly, an approach focused on qualities <a href="http://firstround.com/review/atlassian-boosted-its-female-technical-hires-by-80-percent-heres-how/">can yield more diversity</a> in the distribution of opportunity as leaders are making decisions not based on who they know, but what they need.</p><p>A single person may not have all the qualities needed. This is why it is likely that a small, focused team will be required. It is more important to find the right team and provide guidance and empowerment than to worry about counts of opportunities. <strong>People can share an opportunity, and a person can have more than one opportunity in their career.</strong></p><h3>Opportunity without guidance is not growth.</h3><p>When a person is given an opportunity without guidance or feedback, the organization is not growing her. They are testing her. “<a href="https://hbr.org/2007/07/do-you-work-for-a-sinkorswim-o-1">Sink or swim</a>” is not a strategy for career development. It’s a brute force algorithm for identification of people that already have the skills and qualities. It follows a <a href="https://hbr.org/2014/11/how-companies-can-profit-from-a-growth-mindset">fixed mindset</a>.</p><p>Real growth requires investment. Investment means aligning someone with a mentor to assist him in the work. The mentor needs to be an effective coach that doesn’t just point out flaws in the work, but is able to “think out loud” on how she might approach the problem. People learn how to reason about complex problems by being walked through how others reason about them.</p><p>If the organization wants people to grow, it does have to give them room to try new ideas but it also needs to set up basic guard rails. Guard rails can be in the form of clearly communicated and preferably documented, guiding principles or constraints. The guard rails can be early and often feedback sessions. They should not be in the form of <a href="https://en.wikipedia.org/wiki/Seagull_manager">seagull management</a>.</p><p>Finally, and most importantly, when someone is offered a growth opportunity, failure has to be an option. If solving a big problem is really meant to give someone a chance to develop new skills, it has to be recognized they may not succeed. Regardless of the outcome, the effort still has to be recognized for the value it did provide the organization:</p><ul><li>The person learned new approaches, built a new network, and/or developed new skills.</li><li>The organization learned the problem was harder than they thought.</li></ul><p>If these things are not valued and failure is punished, the organization will have created an environment where individuals will be discouraged from taking on hard challenges.</p><h3>A proposed solution to the market inefficiency</h3><p>I’ve argued the real challenge in most organizations is there’s not enough people with the skills needed to solve all the problems, so it’s likely you will need to give people multiple opportunities to solve problems.</p><p>In <em>How Google Works, </em>Eric Schmidt argued the old saying, “If you want something done, give it to a busy person.” While on the surface, this is an argument to pile more and more work on a small pool of people, it is also another form of the idea that the greatest predictor of what people will do is what they have done. The challenge is how do leaders find <em>all</em> of the people in the organization that have solved and can solve problems. I argue it’s no different than other selection problems:</p><ul><li>Create a bigger pipeline by publishing opportunities.</li><li>Publish a list of the qualities desired for each opportunity.</li><li>Allow self-selection or application.</li><li>If there are multiple responses, interview based on behavioral questions aligned with the desired qualities.</li></ul><h4>One final parting shot at “death by committee”</h4><p>“Death by committee” can be a backdoor way to identify the small team of passionate individuals needed for a problem. With any group effort, there’s a chance that a few passionate people will step up and do the work while the rest remain bystanders or in the best case, reviewers. While this may solve the problem, it comes at a cost. There’s a huge inefficiency in having a roomful of people pretending to solve a problem. Even worse is the opportunity cost of all the problems those people could be solving.</p><p><em>Imagine all the problems that could be solved if we had the right people working on them.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=692759276db6" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>