<?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 Josh Jackson on Medium]]></title>
        <description><![CDATA[Stories by Josh Jackson on Medium]]></description>
        <link>https://medium.com/@josh_0000001?source=rss-1b990a510c08------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*dmbNkD5D-u45r44go_cf0g.png</url>
            <title>Stories by Josh Jackson on Medium</title>
            <link>https://medium.com/@josh_0000001?source=rss-1b990a510c08------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 20:43:12 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@josh_0000001/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[The Collapse of Software, Advertising, and Everything Else: What Agentic AI Just Made Obvious to Me]]></title>
            <link>https://medium.com/@josh_0000001/the-collapse-of-software-advertising-and-everything-else-what-agentic-ai-just-made-obvious-to-me-78fd6d905bd1?source=rss-1b990a510c08------2</link>
            <guid isPermaLink="false">https://medium.com/p/78fd6d905bd1</guid>
            <dc:creator><![CDATA[Josh Jackson]]></dc:creator>
            <pubDate>Sun, 18 May 2025 22:28:57 GMT</pubDate>
            <atom:updated>2025-05-19T02:29:50.412Z</atom:updated>
            <content:encoded><![CDATA[<p><em>First, a pre-article tangent: Why didn’t we call it ‘Aigentic’? I’m assuming there was a VERY good reason, because it was right there.</em></p><p>Until recently, I wasn’t sure how far AI could really go. I saw it writing marketing copy, helping with customer support, generating decent-looking images. Useful, sure, but not transformational. I had previously been waiting to see if AI can actually do a job without humans. And how well it could do said job. With every task at work, I ask my self how AI can help or replace me. Honestly, I didn’t see much it could replace.</p><p>Yesterday I watched a demo of OpenAI’s new agents. What I saw didn’t just change my mind. It confirmed a suspicion I had: AI can do any job, given the right training. And that single realization changes everything about how we understand work, value, and power.</p><h3>The First Real Training: Why Coding Came First</h3><p>The most advanced AI systems today are surprisingly good at writing and understanding code. But it’s not because code is easy or because software is some special domain like you often hear. It’s because the people building AI are engineers. They trained these systems using the real, messy, day-to-day knowledge that only practitioners have. This was the point that was validated by this video. I have already written about how AI isn’t going to touch analytics until someone actually trains it in REAL LIFE analytics. (I still haven’t seen that.)</p><p>So this first wave of AI is shaped by the context and competence of its trainers. Specifically, AI can’t really learn how to do something by looking at the output and the popular culture perception on how a job is done. With coding, it’s actually learning the hidden rules of a job. The workflows. The judgment calls. The unspoken expectations.</p><p>That tells us something important: AI isn’t limited by capability. It’s limited by access to expert-level training in real life.</p><p>So if you apply the same kind of real-world training to other domains — medicine, law, teaching, logistics — the result is obvious. AI can do those jobs too.</p><h3>Software Is Done. Hardware Is Everything.</h3><p>If AI can already write a decent app, debug it, and deploy it to a cloud platform, then the real bottleneck isn’t software anymore. It’s hardware. More specifically, it’s clean, flexible, open hardware that can run whatever the AI builds.</p><p>This flips the traditional hierarchy. For years, software companies were dominant because they owned the systems. Now, anyone can ask an AI to create a new operating system. You can convert apps into whatever native format you want. If you can describe it, you can generate it.</p><p>That means we’ll start seeing new operating systems, even new phones, not tied to Apple or Google. In fact, I think the real big show of force of AI will be either a Windows phone or a Meta phone. For a bit here, people with a huge staff of software engineers will still have an advantage. One of these companies can crank out a ground up phone OS in the next year and also build a tool that allows for instant app compiling in their new phone OS. This isn’t a great move, but these legacy software companies will probably make it because it is all they know. I’d be surprised if it wasn’t already happening.</p><p>But in the near future, having a large professional staff won’t matter. We’re quickly moving towards a world where the hardware just needs to be good enough. The OS doesn’t need to be branded, it only needs to work.</p><p>The winners will be the companies that make fast, reliable, modular hardware that can run AI-generated interfaces. We will be basically moving back to the pre 90s where you buy hardware and do what you want with it. A TV wasn’t locked to any given content. Soon, a phone will be sold without an operating system, or maybe just a common open source one.</p><h3>Proprietary Software Can’t Keep Up</h3><p>As stated, big software companies still have an advantage. They have the infrastructure and the teams to integrate AI quickly. But over time, they won’t be able to keep up with the scale and speed of open source.</p><p>AI changes what it means to be a solo developer. One person, with help from an AI assistant, can build polished, production-grade software in a fraction of the time. Features that used to require full teams: testing, internationalization, accessibility, responsive design can now be tackled during the day while a developer is at his or her “real” job.</p><p>Open source used to be the slower, scrappier alternative. Now it becomes the innovation engine. Thousands of ideas will emerge from people who were never “developers” before. The interface, the polish, and the onboarding can be generated, refined, and tested by AI.</p><p>The result is a flood of real products, not just tools. Proprietary platforms simply can’t iterate fast enough to compete. We’ll finally get email clients with tabs and holding areas for different daily tasks just in time to not need email anymore.</p><h3>Advertising Stops Working</h3><p>The ripple effects don’t stop at code. AI will also destroy traditional advertising. Soon, we won’t make buying decisions alone.</p><p>Imagine shopping with an assistant that knows everything about your health, budget, preferences, routines, ethics, and tastes. It doesn’t care about packaging. It doesn’t respond to emotional manipulation. It just finds the best match.</p><p>You won’t search for “best shampoo.” You’ll just ask, “What should I buy this month?” And your AI assistant will handle it.</p><p>At that point, marketing tricks collapse. Flashy branding, influencer endorsements, even mind reading pop-ups don’t matter if the decision-maker is an AI acting on your behalf.</p><p>Could companies try to inject ads into those systems? Maybe. But no one will use an assistant that’s obviously biased. And from what we saw with Grok regarding South Africa, it isn’t clear that it is even possible to inject a bias not present in the full model. And if an AI starts pushing soda over water for no good reason, people will move to one that doesn’t.</p><p>In this world, advertising becomes a kind of metadata. A product launch is just a data point in a field of thousands. The only way to win is to make something good and describe why it is good in a way that AI can match it with their user. The AI will find it, and your customer will buy it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tUEbs8VvyVs3Tsiu9n1oYA.png" /><figcaption>AI image of what an ad for AI might look like. Nothing about how you will use it, just the facts so the AI can see if it will fit in your life.</figcaption></figure><h3>Still, Even AI Will Be Commoditized</h3><p>Today, AI seems massive. It runs on powerful servers and costs real money to operate. But that won’t last forever.</p><p>Thanks to open models and rapid optimization, AI will eventually run on personal devices. Just like once-heavy operating systems fit in our pockets, so will AI. Whether through Moore’s Law or better efficiency of design, we’ll get there.</p><p>Some of this is already happening. Open-source models like Mistral and LLaMA are getting smaller and more efficient. Hardware accelerators are improving. The knowledge is no longer locked behind corporate firewalls.</p><p>Eventually, governments may even release public models. The same way GPS went from military tech to a standard smartphone feature, AI will become part of the infrastructure.</p><p>In the end, AI itself becomes a layer. Just a utility. The real value will come from how we use it, what we train it on, what we build with it, and what we choose to leave behind.</p><h3>The World That Comes After</h3><p>I really don’t know. What I do know is that advertising and big software companies are already done. Big AI companies are also just providing a service of creating the AI that will replace them.</p><p>Software is no longer a moat. Ads are no longer persuasive. Even the models themselves will become interchangeable. What’s left are ideas, quality, trust, and tools that <em>actually work</em>.</p><p>It will be chaotic at first. But also exciting. A world rebuilt not by legacy institutions, but by people who finally have the tools to build what they’ve always wanted.</p><p>We’re not just automating work. We’re automating the system that decided who got to participate in the first place.</p><h3>Did ChatGPT Write This?</h3><p>The majority of the article is written by AI, yes. I’ve gone through and clarified some of the ideas it watered down. I also took out many of the tells. I specifically told it not to do the “It’s not X — It’s super X” which is an obvious tell. But as you can see, it decided to end it that way, and I left it in. We laughed about it together and it said, “Haha, I’m glad you caught that — couldn’t resist sneaking in a gentle nod to the thing you told me not to do. Consider it a self-aware Easter egg. 😄” I see you Chat GPT.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=78fd6d905bd1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Build a Sales Pipeline Dashboard]]></title>
            <link>https://medium.com/@josh_0000001/how-to-build-a-sales-pipeline-dashboard-6e916f183f9b?source=rss-1b990a510c08------2</link>
            <guid isPermaLink="false">https://medium.com/p/6e916f183f9b</guid>
            <category><![CDATA[sales-pipeline]]></category>
            <category><![CDATA[data-visualization]]></category>
            <category><![CDATA[dashboard]]></category>
            <category><![CDATA[analytics]]></category>
            <dc:creator><![CDATA[Josh Jackson]]></dc:creator>
            <pubDate>Sun, 13 Oct 2024 23:36:43 GMT</pubDate>
            <atom:updated>2024-10-17T20:40:46.988Z</atom:updated>
            <content:encoded><![CDATA[<p>I’m going to take a different approach on this article. Slightly different from what you might expect in a “how to.” In fact, there will not be technical details on how to actually build the dashboard. Instead, I hope to convey the untold parts about a build like this. The actual concepts and steps that might need to be understood to be successful in building any dashboard.</p><p>If you are looking to build something like this out, I am assuming you know how to use your BI tool and know how to shape data. If you are just looking to understand how analytics products get built from a non-developer standpoint, you also don’t need to know the technical details about how everything is done here.</p><p>Without further ado, this is the dashboard that I will be discussing.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9gFXdjq2KMaHMy4lHiZA4A.png" /></figure><p>These are the types of products I really enjoy building. This whole dashboard is centered around one fairly simple concept, but the act of making data around this concept is a very big challenge that involves a lot of data trickery and more importantly, human understanding of intent behind the data. It’s been almost half a decade since I first built this, but it is still a favorite of mine.</p><p>The scenario here from the client is “we want to be able to keep an eye on our sales pipeline.” This is a Tableau dashboard pulling data directly out of Salesforce. At the time, Salesforce was just a very poorly documented, inconsistent data source to me. This build just cemented my beliefs around it.</p><p>This concept of the sales pipeline is basic functionality of Salesforce. In SF, opportunities are the entities that go through the pipeline. A company has contacts, some of those contacts are opportunities to become accounts. An account is another entity that the business sells products to. the journey from opportunity to account is the sales pipeline. It is different for most companies, but it usually involves getting your foot in the door, talking to people that matter, getting things approved through the bureaucracy, then actually having them buy your product.</p><p>Salesforce allows users to name these steps, track where they are, and add notes and contact details along the way. That is very useful. What salesforce does not do is provide anything like the dashboard above. And there are reasons for that that will be covered throughout this article.</p><p>First, I just want to clarify that the sales steps for the original dashboard are covered with more generic steps that should be easier to understand than the proprietary requirements of the client this was originally built for.</p><p>Let’s get started with the build. the first request was for a chart that showed how many opportunities had been in each step of the sales pipeline. This is a nuanced concept.</p><p>Why would anybody want this? Well, this client was very data driven. They always were interested in seeing more in the data, and this example was no different. The reason you would want this concept is that it tells you how much effort you are putting into each step. as an opportunity moves through the pipeline, and that status changes, it is still interesting to know how many opportunities you’ve put through the legal approval process, as an example.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/935/1*cq-phsBCXBjVxU9jnh5dpQ.png" /><figcaption>The original ask for this dashboard was to see how many opportunities had gone through each stage/step</figcaption></figure><p>Unfortunately, as you may already know there isn’t a data set built-out for this concept. If you weren’t aware, the issue is that the opportunity needs to exist in all of the steps it’s been through, not just the step it is on. The chart doesn’t just “know” that because an opportunity is on step 4, that it went through step 1,2, and 3 as well. This dashboard was built before Tableau Prep was really an option, so this was all SQL and Level of Detail (LOD)s off the raw tables in Salesforce. Luckily, the data needed to get this was pretty easy. It will be covered a bit later, but this wasn’t really a big hurdle to get over.</p><p>The first real data issue was that there wasn’t anything preventing sales agents from going backwards on the pipeline. They could go steps 1,2,3, then back to 2. So, I had to ask the client about that. Was the pipeline truly a step-by-step process or was it more like ingredients that can be added at any time? The client insisted that this was a linear path, that any backwards steps were a mistake. I generally find this a bit suspect, as I tend to think that reality is more fluid than these simple constructs can hold, but they insisted. Either way, that choice did make it easier to work with the data.</p><p>That brought up the next question, if a sales agent went back, do we call the current status the status, or do we count the furthest achieved status as the status? If the opportunity went to step 3, but is currently on step 2, how should that be displayed? The client decided that based on the premise that pipeline was linear and that out of order steps were a mistake, that the current status would be the rightest information to display. Basically, we would assume that they accidently advanced the steps, and they meant to be on the step they left it on.</p><p>All of that means that built into this data set is the concept of all of the steps, which order they should be in, and flags to detect steps that should be ignored based on current step. The two money maker tables being used here from Salesforce are an opportunity current status table and an audit like history table to get timestamps when the statuses changed. There are of course the multitude of other tables needed to get names from the keys in these tables and provide meaningful context on the opportunities and sales agents. That’s about as technical as we are getting with this article, but I did want to acknowledge that the data was always going to be a little complex even if these were just joined tables.</p><p>There is one more data issue that I want to cover here. The skipped step. There were instances when people would go from step 1 directly to 3. Again, this is something that needed to be gone over with the client. Especially for a chart like this. After all, we are specifically looking at how many opportunities were ever in each step. If they can skip steps, then maybe they shouldn’t count for those steps. Once again, the client said this was a linear timeline and they had to do each step. If they skipped it, it was a mistake.</p><p>Well sadly, that wasn’t the easy data answer. It meant that even if there wasn’t a row of data for each step and opportunity, there would need to be one added. This concept is called padding data. Adding in data that is “missing” only because data doesn’t work the same way we do. Those new rows of data also needed to have made up timestamps to “fit” between the steps that were actually tracked. That added a couple more joins and flags to the SQL.</p><p>This one chart has all of these data concepts addressed and accounted for. It’s what allows this chart to be both a vertical and horizontal bar chart. Vertically it’s showing how many opportunities went through each step, and horizontally it should which steps an opportunity has gone through. any of the individual bar pieces can be clicked on and that will highlight the entire row and all the other data representing that opportunity.</p><p>By now, you may have guessed why Salesforce has nothing like this in their product. Each one of those decisions is up to the client. If the client choses differently, then this chart doesn’t even really work. This is custom concept for this client that has this specific scenario. The choices this client made may not be uncommon, but they certainly aren’t the only options. The questions had to be asked and the product needed to adapt to the answers.</p><p>That’s the build of the primary chart for this dashboard. All the other elements on the dashboard use the same data source, but since they are simpler concepts, we won’t need to dive too deep into how they work with the data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/309/1*g0yIo7FJo56ZsT8mdX7ObA.png" /><figcaption>Shows how many opportunities are currently in each stage</figcaption></figure><p>The next chart just to the left of the main, is the basic status chart called “Accounts in Stage.” Though this dashboard says accounts, most people would generally call this the opportunities because these are not yet accounts that have gone through the pipeline. This client called that concept “Accounts in play.”</p><p>This is the chart your junior analyst sends you after copy pasting the current status data into Excel. But its simplicity is also its necessity. It was obvious as soon as I finished building the primary pipeline chart. The primary chart doesn’t easily convey how many opportunities are in each stage and that is the very first question to pop into your mind when talking about pipeline steps. That question is so foundational that it should actually be displayed to the left of the primary chart, so that it is read first. This often happens in complex analytics. Once you find a way to visualize an interesting set of data, it becomes clear that the audience is going to need to be grounded first. Sometimes this is numbers across the top of a dashboard, and sometimes it is a summary to the left of main analytic.</p><p>As I went back and forth with the client, the other charts came to life. They wanted to see the following:</p><p>· How long was each opportunity being worked</p><p>· How many opportunities was each sales agent handling</p><p>· How long was each step taking</p><p>· Row level of each opportunity, and key dates around it.</p><p>Initially, the chart showing how long an opportunity had been in the pipeline was much more prominent. If I remember correctly, it was a bar chart with different relevant time periods like 1–5 days, 5–20 days, etc. But as time went on, it was clear that it wasn’t as important to have that spread as it was to have the ability to quickly know individual and average times. The big jitter plot across the middle shows how many days each opportunity has been in the pipeline.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*myYUJmpA0BS42v5FwPGsMg.png" /><figcaption>This shows how many days each opportunity has been in the sales pipeline.</figcaption></figure><p>There is one quirk here to mention. This isn’t just a date difference from the start date. Well, it is that as long as the opportunity is in the pipeline. Once it is out of the pipeline, then it is the difference between the start and close dates. Again, this is a simple concept, but it is almost always missed by people new to analytics. Without this, opportunities will continue to gain days as time goes on and this will throw off averages. it is easy to catch in this chart, but a simple average days number will obscure that detail.</p><p>It’s important to note that since this is Tableau, everything is clickable and filterable. If you click on a sales agent, you filter to just that sales agent and have a custom dashboard just for them. If you click on an individual square, that opportunity is highlighted so you can see all of the relevant information for that opportunity. So, if you click on the opportunity that has taken the most time, you can quickly see which agent is handling that one, which step it is on, how long each step took and how that compares to the other opportunities that went through those steps. Knowing when to filter or highlight, and when to move the average lines based on interaction is another key skill in making useful analytics products.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/412/1*8TM5OETOP3wjwj29unview.png" /><figcaption>Jitter plot showing the number of days each opportunity spent in each stage/step</figcaption></figure><p>The next chart to discuss is the time in each step section. Again, this is a jitter plot with average because the client really wanted to know the average, but they also wanted to know how each opportunity was going. These jitter charts may be a bit advanced for some audiences. This client was very interested in squeezing everything they could out of the data they had. But really, I think a jitter like these two charts is something that can easily be educated and provides much more value in the same screen space when compared to a simple bar chart with averages. I would always try to push this concept if the client is at all interested in the individual data points. Audience, data set, and relevance really matter here, but I like to lean towards pushing people if it is helpful.</p><p>The quirk here goes back to those skipped steps. The skipped steps show as having 0 days. Rather than just using that as part of the average, I once again, pinged the client. Did they want to include the 0 days as part of the average or did they just want to average the non zero days. Even if the steps aren’t being outright skipped, a zero in an average usually doesn’t help much. We’re really usually interested in how long something takes when it actually takes some time. The screenshot here does include the zeros in the average, but later they were removed to provide a better picture of how long those steps would take.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/435/1*KK72CI-3Erm-1roPnC0Jvg.png" /><figcaption>Shows how many opportunities each sales agent group is working</figcaption></figure><p>To the right of the “days in each stage” chart is the sales agent chart. It is a unit stacked bar, meaning that the bars are built on the individual data rows, instead of just an aggregate of the total. This concept is similar to the main pipeline chart and serves a similar purpose. We not only get to see how many opportunities each sales agent has, but we can also see what step they are on. As mentioned earlier, this allows for highlighting to see exactly what is going on with each of the opportunities. A simple bar chart would cut out some very useful functionality. This chart could have also been a time based jitter, like the others, but that isn’t really needed. If you select an agent, it will filter the rest of the charts, and specifically show the timeline for the opportunities in the two other jitter plots. This unit bar chart is a perfect blend of complexity and simplicity to keep users grounded in basic chart types that everyone should be familiar with plus the additional functionality.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/426/1*FTIN9DGzhB8WYjt7zs44SA.png" /><figcaption>Shows the detailed status for each opportunity</figcaption></figure><p>Finally, we have the classic “can I just see the data” request. I do like showing a good data table in a lot of instances. It builds confidence and empowers the audience. This is very basic table of opportunities that just provides information that is already available on the dashboard in tool tips, but it is very useful for screenshots, and it allows the users to download a nice table that is going to work well in something like Excel.</p><p>It’s worth noting that this table is still running off of the highly shaped and padded data that was mentioned earlier. As such, even a simple table has a lot of work behind it to make it look like the very simple table that most people imagine drives a data product like this.</p><p>Well, that’s it. That’s how you build this dashboard, assuming you already know data and your favorite BI tool.</p><p>The point here is that there is a lot of interaction, assumptions, and data manipulation that goes into any good data product. These are core concepts that most people that build these products are very aware of. But for whatever reason, everyone else in the industry, from the vendors to the clients to the stake holders doesn’t understand. It’s the concepts in this article that are the reason that you might see a vacant stare from people when they are told “AI will take your perfect data and make everything you want.” Because even perfect data, which doesn’t exist, still will need to be framed for the stakeholders’ ideas of what the data really means.</p><p>This single concept dashboard took a few meetings and many more emails to get it done how the client wanted. I have actual intelligence, and I couldn’t just guess what the data means or what the client wants to do with every exception. Inquiring with your client/stakeholder is required for quality analytics.</p><p>All of this is why I never understood why vendors start with a completed data product as an AI panacea to analytics. I would be much more impressed with an AI product if it started out questions and explaining the consequences of different paths. And I don’t doubt that that is possible. It’s just never what is built.</p><p>Thank you for reading this article. It’s something I’ve been meaning to put together for a long time. I hope you enjoyed the insights into building analytics, or were able to commiserate the hidden complexity with me on a “simple” build.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6e916f183f9b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[2024 Tableau Feelings]]></title>
            <link>https://medium.com/@josh_0000001/2024-tableau-feelings-945298b6d35b?source=rss-1b990a510c08------2</link>
            <guid isPermaLink="false">https://medium.com/p/945298b6d35b</guid>
            <category><![CDATA[tableau]]></category>
            <category><![CDATA[interpretation]]></category>
            <category><![CDATA[data-analytics]]></category>
            <category><![CDATA[analogy]]></category>
            <category><![CDATA[analtyics]]></category>
            <dc:creator><![CDATA[Josh Jackson]]></dc:creator>
            <pubDate>Sun, 26 May 2024 18:51:58 GMT</pubDate>
            <atom:updated>2024-05-26T18:56:24.757Z</atom:updated>
            <content:encoded><![CDATA[<h3>2024 Tableau Feelings: Who’s the tool here?</h3><h3>Author’s Notes:</h3><p>This article is a recap my feelings about the 2024 Tableau Conference in San Diego. There will be little tactical or even strategic discussion. This will be more of a philosophical or metaphorical recap of my feelings up to and including the conference. They center around one incident. While the incidents was very minor in the scheme of things, I felt like it highlights my general feeling about everything Tableau today.</p><p>Before we start, I do want to add a caveat. My words here might come across as harsh, but they aren’t intended that way. I look at the situation very clinically and am not implying anybody is better or worse than anybody else. Have fun.</p><h3>The Article:</h3><p>When Ann, my wife and business partner, went to the Tableau Conference this year, it wasn’t out of necessity. We’ve both put aside our data careers for a bit to see and understand the world and reality. The main reason we were there was to connect with people. But as for the conference, I imagined it was like being on an away mission in Star Trek. We were here to observe a late-stage acquisition of a company.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*64_rVMB1Xs8A_gXRf6kc1g.jpeg" /><figcaption>Observing the architectural beauty of the San Diego Convention Center.</figcaption></figure><p>Full Disclosure: I’ve never liked Salesforce. I was involved in a few migrations to Salesforce in a call center role. The roll outs were very rocky and many of the promises fell through. Later, much deeper in to my data career, I’ve had clients on Salesforce. The data is so troubling and undocumented that it is almost impossible to pull anything useful. Basic insights required elaborate workflows to clean up the data model. So, I’m not a fan.</p><p>In my more recent dealings, I haven’t felt any better about the product. Salesforce is basically an easy-to-use database. Or that is what it tries to be. That means, that at its core, it tries to put things into boxes. Things in reality can’t ever really truly fit into boxes. It’s a concept common in tech, but it is the foundation of Salesforce.</p><p>Conversely their marketing is all natural and rounded, I feel, in a way to try and deny the forced boxing that underpins the product. It’s like a criminal trying to become a police man, only to find out that he can’t fight his true nature. It’s also why, I think, the Tableau community rejected it. The contradiction and self-deception feels weird in the brain.</p><p>But I digress. Let’s get back to the story/metaphor that underpins many of my issues with Tableau.</p><p>During the “Candor” session towards the end of the conference, there was some mention about branding strategy from a person that runs his own consultancy. The gist is that he didn’t like the idea of getting rid of the mission statement “see and understand your data” as a core concept of Tableau. The panel explained that the feared changes would not be happening to the relief of the crowd. But there was a telling misstep. The person representing marketing made it clear that she did marketing and we were data people. Implying, if not outright saying, that we should stay in our lane and leave the marketing to the professionals.</p><p>To be clear, we are data people. Well, we are analytics people. It doesn’t roll off the tongue as well, but it is what we mean when we say we are data people. We don’t actually “do” data. We are forced to because we are the people that USE data to DO analytics. As such, we are in the best position to do data as well. But we generally don’t build databases or organize things into boxes. We actually take the things out of those boxes and do our best to translate them back into reality so that humans can understand them. It’s really the opposite of Salesforce. And as such, our jobs and skills are the opposite of Salesforce implementers and users.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9aIgypwkqhFFlOtb_j4xog.jpeg" /><figcaption>Me wearing a lot of “Data” gear. “Analytics” wouldn’t fit as nicely on the merch.</figcaption></figure><p>Additionally, a CRM is a requirement for any company that hits a certain size. At some point you can’t keep track of things in spreadsheets, and you need to go to a real CRM. It’s just a question of which one. Salesforce is a big and pretty company that offers this product. Companies must pick one and Salesforce just needs to look like the best option.</p><p>Analytics, on the other hand, is a luxury. Nobody needs analytics. We say they do, and we want them to, but it is not really needed. Businesses do fine without seeing or understanding their data. Because of this, every analytics project is an up sell. That means analytics people are always messaging and selling analytics, no matter if they are freelancers or deep in corporate org charts. We must continuously show that the tools and the headcount are “worth it.”</p><p>Basically, everyone needs somewhere to put their stuff (CRM). Nobody really needs to display their stuff, but the people that care will seek out people that can display their stuff in a nice way (analytics). A quick analogy is that Salesforce implementors are similar to warehouse workers, while Tableau implementors are more like museum curators. If you know nothing about either of those things, they can be described in a way that sounds the same, (a place where things are stored and can be accessed easily) but they are actually very very different.</p><p>Now back to the “stay in your lane” esq slight that was dropped. Some of us “data” people run our own consultancies. When I was walking around with Ann’s friends as she talked with person after person at the conference, I kept saying, “oh yes, this is another person with a small consultancy, we know each other and have a certain respect… etc.” but why do we have a certain respect? Mainly because we have all the skills needed to run a business. We all do our own sales and marketing as well as the tech stuff. The truth is, as stated earlier, we always had to sell our product, to our bosses, to stake holders, to everyone. It’s part of the job. Additionally, we all know how lucrative this job is. Ann and I are currently taking a year or 3 off because we can. Interworks literally parked a camper in the middle of the conference. That’s just a guy that has a couple dozen people that can do Tableau working for him dropping untold thousands for a pretty small marketing opportunity. And for the owner, it’s money out of pocket, not a corporate budget. It’s actually more of a flex than a marketing move. But it’s a flex that I can give a nod to.</p><p>And that’s the issue. The odd guy in the audience asking a question about the mission statement. He runs a consultancy. His team is in a very large company that I am sure brings pride to Tableau sales and executives. But HIS team is why Tableau is there. His team is the talent. He is the marketer that keeps Tableau at that company. I’m being coy because I’m not sure if there is an NDA in place, but the company is probably the first one to pop into your mind. You might even say to yourself, “no way it’s THAT company” but it IS that company.</p><p>So here I am, sitting in an audience watching one of my fellow successful entrepreneurs getting dressed down about how marketing works. His marketing skills are visible on the Tableau bottom line. And he said nothing out of respect for the situation. Immediately, I begin typing up a question about this. I have trouble speaking in public, so I needed it written down, and by the time that was done, the show was over. Luckily, I have a word processor to capture these ideas.</p><p>Let’s digress again. Notice how I have a word processor and I use it to capture and share my ideas? Imagine if a famous author were at a word processor conference and the head of marketing of the word processor told the author to stay in their lane. That’s basically what happened, except maybe it was a popular ghost writer, but still, I think everyone would be taken aback. No one was taken aback.</p><p>And now we get to the real point of this whole thing. Nobody really understands what analytics is, and that includes most of the people at Tableau for the last 6–10 years. And absolutely includes the Salesforce staff that replaced the Tableau staff.</p><p>Analytics, is a practice where professionals use tools to interrogate data and make it understandable for the people that couldn’t understand the data in its raw format. It’s a very human and creative story telling job that involved empathizing with people that know less than you about a subject, and packaging things for simplicity and usability. It’s more art than science. And it’s so new that the science isn’t even there yet.</p><p>Tableau is a tool in the tool-belt of the analytics professionals that want to use it. Tableau is not the talent or the solution, we are. Tableau is a 5-burner stove, and we are the chefs. Tableau is a word processor, and we are the authors. Tableau is a line of power tools at The Home Depot, and we are the doers getting things done. Tableau is a Louisville Slugger bat, and we are the ones hitting the ball out of the park. Tableau is Fender, and we are the rock stars. Of course, we are all at different levels of capability and success. But to be clear, Tableau, and those working at Tableau are not in the same game we are in. Tableau and Salesforce make the tools. We use the tools. Only in a field as nascent as analytics is it even fathomable that the tool maker would be acting the way Tableau does. The whole product strategy depends on the majority of the population not understanding what analytics and data visualization actually is, or how it works.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2sO3tSbgwYyI1O3LeA4oHw.jpeg" /><figcaption>Petco Park. Nobody showed up here to see the sports equipment.</figcaption></figure><p>The remaining questions at the session were about containers, colors, and other usability issues in the tool. You can’t help but to feel a little bit eye roll-y listening to someone explain container issues to the heavy b-school c-suite that was there. But allow me to translate what they were saying: They are asking Tableau to improve the tool so we can do our work faster, and better for the people that rely on our work and pay the license fees. We need a product that we can sell as useful to our bosses. One where we don’t need to make excuses for why we can’t do the things the boss wants.</p><p>It will be clear that the chart building tool is up to spec when all the charts in the keynote for the chart making tool are made in the chart making tool. That’s what our stakeholders want. They want beautiful, accurate, and quickly made charts that fit their very distinct and precise needs. Again, we always have to sell our worth, and we sell that worth with good analytics. And that depends on good, modern, tools.</p><p>It feels like Tableau forgot that they make the tool. This is a company that made the LoD. A very deep analytics concept. Set actions, parameter actions, these were the headlines of the last decade. These are things that we can show to our stakeholders in the products that we make for them. They don’t need to know how they work, they just want to get analytical insights. What exactly are we presenting to our teams when we get back from this conference? A sales pitch for Data Cloud Einstein the Wave Surfing Forest Creature? A 50% off coupon code?</p><p>To be fair, of all the tangential products that Tableau has released that never get used like Metrics and Ask Data; Pulse is the one that seems like it might be deep enough to have some legs. But I went back through my book of clients and thought to myself, “if I had pulse would it answer this ask.” Hundreds of products across dozens of clients, and maybe a few could have used a Pulse box for some of it. Which isn’t even up for debate. Tableau themselves had a slide up for Pulse that said it intends to cover the 20% of requests that are asked 80% of the time. Meaning that 80% of asks will still have to use the chart and dashboard maker. A tool with a future that is very ambiguous. Is this a tool on my computer? Is it in the cloud? If it is in the cloud, how am I applying custom JSON files to it for formatting? Is this a standalone product, or is it a tab in the Salesforce eco system with square pills? (Credit to Ann for catching that slide in the keynote) It was really hard to tell from what we saw, but one thing I know is that square pills are hard to swallow.</p><p>In the end, the core problem might actually be the mission statement at Tableau. It is not Tableau’s job to help us see and understand our data. It’s their job to make a TOOL that lets US help our stakeholders see and understand their data. If that mouthful had been the mission statement at Tableau this whole time, we might not be in the mess of weird half-automated tools that rely on fictional perfect data sets, and an abandoned chart maker/data explorer.</p><p>Heck, Salesforce might not have purchased Tableau if they had realized that you don’t get the Yankees when you buy Louisville Slugger.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=945298b6d35b" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>