Walter Gretzky is credited with the quote: “Go to where the puck is going, not where it has been.” Walter used socratic questioning to teach his son, Wayne, hockey strategy. Here’s the full context from Wayne’s perspective: Him: “Where do you skate?” Me: “To where the puck is going, not where it’s been.” Him: “Where’s the last place a guy looks before he passes it?” Me: “The guy he’s passing to.” Him: “Which means…” Me: “Get over there and intercept it.” Him: “If you get cut off, what are you gonna do?” Me: “Peel.” Him: “Which way?” Me: “Away from the guy, not towards him.” (Gretzy, Reilly, Gretzky: An Autobiography p. 88) Puck On To win a game of ice[…]

If you need a tool to break down complex scenarios, this approach, a tool using decision forests, might be right for you. By the end of this post, you’ll be able to use the gentler, forward, variant of dancing in a decision forest. This is a post is intended for a curious audience. Decision You’re an extraordinary assembly of chemical gradients. At any given moment you have the opportunity to make millions of decisions. The crudest segmentation, the roughest way I can impose order on all of this complexity is divide them up in two types of decisions: to act, or to not act. Further, when you decide to act, there are two broad types of actions: explore and exploit.[…]

Suppose the following scenario: Series A or B; A data science firm (narrow machine intelligence, applied machine intelligence, general machine intelligence, predictive or prescriptive analytics, software or hardware); Technical CEO / Co-Founder; Chief Marketing Officer (CMO) just hired; What might the CEO-CMO relationship look like? The relationship could be great. If there’s one stereotype about data science CEO’s, it’s that they like incentives to be aligned. The CMO would likely be brought on to focus on growth. If revenue grows, valuation grows, and collective comp would grow. There might be points of friction. From the CMO’s Perspective: Why is the CEO constantly at me about metrics all the time? Why is the CEO always on about non-working dollars? (Why don’t[…]

It seems like a lot of people value certainty. People buy a lot of products and stories for certainty. Insurance. Investment advice. Forecasts. Indulgences.Many entrepreneurs, in particular those in data science, sell certainty. What else is an F1 score other than a measure of certainty on some level? Given some inputs, our machine transforms them some way, which produces some statement about the past, present, or future, with some quantifiable amount of certainty, so that you can do something with confidence (or feel more secure). We sell certainty. And yet isn’t it curious about how much insecurity we’re creating while we do so? It has always been easier to sample data from the past, pull a heuristic from it, and[…]

Data scientists spend so much time focused on learning: both machine learning and human learning. A machine can learn. A data scientist spends a lot of time just trying to persuade a machine to learn. It just takes a lot of labelled data. What about collections of people? Organizations can learn too. It’s just that the data isn’t all labelled well. Why Organizational Learning is Important I was so impressed with Carl Anderson’s synthesis two years ago, about Data Driven Cultures, that I unpacked it and applied it to startups and strategy. Coming back to it now, in 2018, a lot of what he was saying is purely about learning. Carl Anderson, 2015, described a data driven culture as on that:[…]

Ben Thompson calls culture the accumulation of decisions. Assume that it’s true. How do decisions at a tech startup come into being in the first place? A startup can be instantiated with the business plan. And if you take a Beinhocker (2006, The Origin of Wealth) approach to it, you may believe that there’s a Library of Smith which contains every single business plan that’s possible. There are trillions upon trillions of potential business plans. And management is pretty much reduced to a machine that is able to execute the plan to generate wealth. Everything that has potential is possible at the beginning and assume competent management. (Image related – a bit esoteric*). In the context of a startup, a[…]

A great mind in public policy told me, just this last September, that people are really bad at judging the rate of technological change and when it’ll affect them. It’s like standing on a railway. You can see the train out there. Some people assume that the train is going to hit them very soon. They get off the tracks. Then, when the train is getting very close, others misjudge the speed and assume that it’s still a far way. And then they get hit. It’s a great analogy because it combines prediction with decision. The rate of technological change is actually quite difficult to predict. If it was easy there’d be a lot more successful startups. One Heuristic Start[…]

Some work is very clearly product work. It’s work on things where the success and failure is dependent on the users of the thing. Your users pay you. Their satisfaction matters above all else. Optimizing for the satisfaction of end users is a distinct activity. Hypotheses have to be assessed and then tested – because it’s very likely that you’re going to be wrong. There’s technology that has to be set up such that it’s reliable and robust for the intermediate to long run. It’s designed to be effective and persistent, with all of the instrumentation that goes along with that. That might include manual A/B testing, user-focused analytics, and extra special attention on the optimization objective. Clear product work is[…]

Why does it seem like all the unimportant, easy stuff gets done first? Look up The Urgency Bias. Employing simplified games and real-life consequential choices, we provide evidence for “urgency bias”, showing that people prefer working on urgent (vs. important) tasks that have shorter (vs. longer) completion window however involving smaller (vs. bigger) outcomes, even when task difficulty, goal gradient, outcome scarcity and task interdependence are held constant.- Zhu, Yeng, Hsee (2014) Even when task difficulty, goal gradient, outcome scarcity AND task interdependence is held constant, urgency wins. Even when it would be more beneficial to do something important instead of something urgent, even when you’re painfully made aware of those incentives, you still gravitate towards doing the urgent. There’s[…]

In general, information retrieval from analytics systems becomes harder with the degree of customization (It gets harder to find things over time). That customization is frequently an expression of the values of a culture over time. The inertia of the technical debt caused by early customization is greater than the inertia of a data driven culture. There are no silver bullets. The rest of this post unpacks that paragraph. Information retrieval from analytics systems becomes harder with the degree of customization Assume a vanilla implementation of Open Web Analytics. Or Google Analytics. Or Adobe Analytics. It’ll tell you a lot about a web system on its own. The optimization objective that is at the core of the business will typically[…]