Can I roll my own?
It’s a common question. "Can we build our own funnel analytics/forecast model in <fill in the blank with your preferred analytics engine>?"
Sure! Why not?
Got a team of statisticians and programmers? Go have at it. But keep in mind the famous Yogi Berra aphorism: “Baseball is 90 percent mental. The other half is physical.”
So what does this have to do with funnel analytics? It's 90% data prep. The other half is math.
Several years ago, the CEO of a machine learning startup remarked to me that he had built the most powerful system in the market for doing sentiment analysis on Twitter feeds. I naively asked, “Is that because you built a better ML algorithm?” “No”, he said. “The machine learning is five lines of Python code. Anyone can do that. It’s the other 50,000 lines of code we built around that to deal with the data.”
And so it is with Funnelcast (or any forecast/statistical model). Our core analytics are based on standard statistical functions available to anyone. But data preparation, cleansing, and applying the statistical functions is where the real work takes place.
Everyone likes automotive analogies, so here you go: you have the engine, why not go build a car?