An Evolution of SaaS Analytics
Over the past few years, we have seen a rapid evolution of analytics unfold before our very eyes. Since the analytics ball is still rolling, it is exciting to see how changes in SaaS analytics design swiftly adapt to technology and marketing trends. Immediate history, analysis of current affairs and prediction, all rolled up in one. Kind of like a SaaS analytics dashboard.
The Value of Telling a Story
SaaS analytics first emerged in November 2005 when Google launched Google Analytics. To this day, Google Analytics is the most commonly used descriptive analytics platform.
A word about descriptive analytics: this analytics discipline tries to answer the question what happened? Where do customers go in your SaaS? How did they get there? How long do they stay there? Where do they go afterwards? SaaS companies use these types of analytics to keep score and see how their KPIs are improving over a constant period of time. These KPIs can be anything between known number of unique visitors to anything more customizable, like sign-ups, Opportunities, specific conversion goals and so forth.
But this is raw data shaped like blunt tools. Invaluable and irreplaceable, but blunt, lacking the finesse needed for surgical scrutiny. The point where it gets interesting is the segmentation into user-specific analytics. The idea here was to look at segments of your users based on their individual properties, cohorts, etc. As opposed to seeing how all users flow through your site, the aim here is to select a specific user segment and answer the question: what really engages?
The problem with raw analytics is the sheer volume. There was too much of it to be used effectively. The amount of data stored for descriptive purposes was simply too much for the human mind to grasp in a way that made sense, so the next stage of evolution was inevitable: breaking down the information into smaller, digestible bites. Or else speeding past the warehouses of raw data, directly to the actionable stuff: predictive analytics. This concern was primarily addressed by tools like Google Analytics, Mixpanel, Kissmetrics or a proprietary dashboard developed in-house.
Predictive analytics tries to answer the question: what will happen? In order to say what is likely to happen you need to incorporate a lot of historical data and create a data model. This involves heavy lifting a lot of data and analyzing it using computing power.
In SaaS-land, we basically pay attention to 4 types of predictions:
|Immediate concern||Which trial customer is likely to buy||Which trial customer is likely to churn|
|Predictive planning||Which paying customer is likely to upsell||Which paying customer is likely to churn|
The massive amount of data gathered under the descriptive umbrella is not useless, simply too much to digest and process effectively. Consider all the metrics we have managed to track using our descriptive analytics tools. What we need is to be able to integrate it, say formulate a function that will tell us what we need to know about customer health, how engaged they are.
ƒ(m1, m2, m3… mn) = thumb up | thumb down
Getting Actionable Anything
Until recently, this was done by data analysts. But the nature of people and business is to delve deeper, find rhyme and reason and if possible –control the very future. Or at least know what it holds so we are well prepared for it. And so prescriptive analytics were born, so that we can tell what can we do about it?
This paradigm has fascinating applications in the realms of health care and natural resource management. In SaaS context, its main application is to help manage business prospecting through answering the question: what is the likelihood that a specific customer will churn and what can we do to change this ominous outcome.
Computational experts and data analysts have been slowly working their way up this new frontier for over a decade now. The one discovery that seems to be undeniable so far is that a control group is critical for applicable prescriptive analytics. Which is to say, determining which parts of our application are used by our healthy and engaged customers? This information would then allow us to communicate with our less engaged customers and drive them to start using the aforementioned areas of our application.
So what’s next?
It seems like knowing what we can do to control our business prospects is pretty much the final frontier for BI and SaaS analytics. So what is the next stage in customer engagement analytics?
One possible direction is actual modelling. Being able to come up with a generalized model, that individual SaaS applications are all instances of, with different parameters.
Another direction is the why is this happening? which may seem like a variation on prescriptive analytics, but really goes much deeper than that. Many of us would pay a lot to be able to read our customers’ minds (as far as their engagement with our software goes, in any case). Really understanding the rationale behind certain preferences and rejection patterns would take us all a long way in being able to provide the SaaS elements our customers truly need.