Most startups can bucket their analytics into three groups:
- Experimental analytics is about testing assumptions and hypothesis. If you can remember your high school science classes—where problems take the form if this, then that—then you understand experimental analytics. Product, marketing and sales can all benefit from this approach. Throw out the subjective and let data speak for itself. Experimental analytics, specifically A/B tests, are central to the product development and marketing processes of companies like Google, Amazon and others.1
- Predictive analytics makes use of historical data (and the occasional assumption) to predict future behavior. It is the sexy side of big data; it’s also nothing new. Predictive analytics is foundational to many business models—banks loan using credit scores, insurance companies predict fraudulent claims, email service providers identify and filter spam—and you find models in diverse industries like telecom churn management (uplift modeling) and sports management (think Moneyball).2
- Archaeological analytics is retrospective and it can be one of the most time-consuming branches of analytics. Even at startups with a small user base, it’s easy to drown in data. Analyzing cause/effect in the rear view mirror requires breadth of data to control for variables and depth of understanding to distinguish a business model’s random variation from real changes. To support this process, I find it helpful to keep an event log—a shared spreadsheet to track product launches, changes in data storage, new press and any change that potentially affects your key metrics.
The common thread between these three flavors is effective storytelling. There’s a time and a place for each in startup analytics.