Why are so many, so hesitant, to make a claim of causality? This symbol, the honking red arrow, is the most powerful and important one in analytics. The arrow represents a claim that one variable causes another. To say that X causes Y involves judgement. Statisticians are quite right to say that correlation doesn’t prove causality. Statisticians have tests that attempt to rule out causality. But to assert that a relationship is causal requires judgement. Statisticians still have a few issues to work out with Father Time and Mother Positivism. To a certain extent, digital analysts have inherited a few of these issues. Is there a deficit of judgement in digital analytics? Probably not. Leading digital analysts generate loads of[…]
Author: Christopher Berry
My first experience with product management was a course called ‘software engineering’. Of the fifteen teams of students who were using the full throttle software engineering method, complete with UML and a short burst of requirements gathering in the up-front, eight teams failed to deliver a product. Of the three teams I knew that succeeded, they each had one person that did all the software development, while the rest of the team was responsible for the documentation and working the waterfall. Certainly, this was quantitative evidence that software engineering, as it was conceived of back then, was really ineffective. How could mechanical engineers hang a VW Bug from a sculpture, reliably and safely, using their engineering principles, while, a group of software[…]
Most figures I found, for the month of October 2012 (Including Mobile): Google’s search market share around 86 to 90% in the United States and 89% globally. Bing is ~7% in the US and ~5% globally. Yahoo 3% US / Baidu 3% globally (China). Search, as a design pattern requires, at minimum, a box into which you enter words or numbers, and a medium to display results, what is today called a Search Engine Results Page (SERP). (It doesn’t really have to be a page at all, which is why I use the word medium. Siri is a good example. Glass is another. That sort of thing.) The more places Google can put that box, the better it is for[…]
Decision Orientations pose their own special challenges for folks used to delivering situational awareness artefacts. Typically bounded by time, involve prospection, and contain a weighing of preferences, Concretely: Time Bounded: Planning for a major strategy roll out may take quarters. Deciding to discount a SKU may take only a few seconds. Time horizons vary. And they matter. Propsection: thinking about outcomes in the future. How much do you expect to gain from a decision? How much do you expect to lose? Preference of the people making them and those around them: Is that future something that you believe is desirable? Optimistically, you can view these facts as opportunities. Pessimistically, you can view these facts as constraints. Let’s view them as[…]
Both US campaigns make use of analytics and good practices from Operations Research. According to an article from ARS technica, the Republican machine, ORCA, didn’t do so well. Summary: The Romney campaigns Get Out The Vote mobile app / engine was called ORCA. There were severe deployment failures, including using a single server to power the mobile app, and made a complete mess of secure sign on and credentialization. It cost the Republican ticket many resources, most notably, the time of 30,000 of their greatest supporters many hours of frustration on voting day. Editorial: The ORCA’s fail whale has the potential to provide comparative operations researchers a great chance to compare and contrast IT and analytics practices. Good will come[…]
Nate Silver, who ran forecasts for the Five Thirty Eight blog, called the 2012 Presidential Election right. The #datascience crew tweeted #mathwins Nate mixed a lot of time tested methods and ported a novel one over into political science. He talked about the future. He demonstrated the power of predictive analytics. He accomplished a lot. And he did a lot for predictive analytics. I thank for him for that. He took a good risk with his career and with analytics more generally. He won. Pundits lost. And, he’s totally getting laid today. Nate’s work was a beautiful piece of predictive analytics, and, had more work was done in D3 and productization, it would have qualified as full on #datascience. The[…]
Most discussions of statistical bias, in the world of sampling, revolve around the actual randomness of the sampling. Is there a systematic bias in the way the sample is collected: either in terms of those who have been selected to participate, those who opt to participate, and those who chose to answer specific questions. It’s commonly argued that if the sample is biased, you have to throw away the whole data set, because the sample is not representative of the overall population. And, in general, we confine our discussions of bias to the nature of the sampling, or, how summary statistics vary against what is expected. There’s another type of bias that revolves around inferring causality. Statisticians generally don’t enjoy[…]
There appears to be a belief that facts have two sides to them. It makes the: marketing scientist in me smile public policy quant in me rage scientist in me flip the table Jimmies status: rustled. Stories may have two sides Models may have two sides Ideologies have two sides Facts do not have two sides. And yet, there’s been a few folks coming out of the woodwork lately: Jack Welch and his tussle with the BLS Unskewed Polls in the Presidential Race And many, many others. A fact is defined as something that actually happened, exists, or is reality. We can experience facts first hand, or observe them though instrumentation. In analytics, there are multiple types of instrumentation that[…]
Has Pinterest topped out? You may be familiar with the Bass Diffusion Model. In short, there’s a predictable function that is very effective at forecasting the adoption curves of new products. The trickiest part of using the Bass Diffusion Model is estimating at what point saturation occurs. Saturation just means where the number of adopters levels out. In the image below, growth started decelerating around 4 years in and certainly flattened out at year 7. I don’t have access to Pinterests own analytics tools. So, like you, I have to rely on public, third party, estimations of what’s going on. Alexa is reporting a flattening reach curve. You can see that below. Most web traffic follows what digital analysts call ‘an[…]
We sat in a nice pub as Sandy rolled in to Toronto last night, just a few data scientists and I. “It’s very difficult to express, succinctly, how dashboards aren’t what people have been told they are.” “Well, you know what the problem with dashboards is.. that they’re not presented in 1080p High Definition. We need to get our dashboards in High Definition.” “Or retina display!” “No. In 3D!” “Pie charts will COME, RIGHT, AT, YOU!” And after laughing for about five minutes, our shoulders bowed upon realizing that somewhere out there, regardless of how ridiculous the concept is, some group of finger-clickers are discussing the same idea right now. Only that they’re serious about it. And that retina high[…]