In our careers in consumer businesses, we have all seen business managers, product managers and analysts struggle many times with pinpointing the source of the latest metrics fire– be it why is the metric moving south, or how do I grow the business faster.
This typically takes the form of management of a company asking their teams to deep dive on why a metric is going south. And analysts start to review the set of predefined charts in a BI dashboard solution. This only works if you had all the metrics well defined and being monitored, and even then, traditional BI solutions don’t get you to underlying hotspot identification. Therefore the analysts begin to hypothesize and do slicing and dicing of the data using SQL queries and custom scripts. All of this takes a lot of energy (cost) from the analysts and associated team that includes product management, engineering and middle management.
An approach for handling this proactively is to first create a structure to your businesses? KPIs and their drivers (potentially multilevel sub-KPIs) in either a tree structure or a funnel structure. Trees are better suited for business performance KPIs e.g. sales performance drivers whereas funnels are better suited for marketing or e-commerce use cases e.g. acquisition or conversion.
Once the KPIs are well-formed, you would need to employ machines to automatically monitor the underlying customer behavior and transactions that determine the values of the KPIs over time. This is so that you can pinpoint which changes are causing the KPI deltas, and catch the changes as soon as they become observable in the data.
This is where machines are great at finding the hotspots that contribute to KPI deltas, they can be of great assistance to human analysts so that people can be more productive and spend their time on strategic aspects of the insights produced by machines.
Assuming you start with an acceptable state of the business, this will ensure any changes or deltas in the KPIs can be explained by underlying customer behavior changes or product/ marketing feature changes. This becomes insurance for your business and saves a lot of firefighting by analysts and product teams and enables you to take corrective actions.
In case the business was not already in a happy enough state, and you are looking to grow the business aggressively (e.g. growth hacking use cases), the funnel view for monitoring is more appropriate– where you would constantly want to be tackling the next biggest cause of constriction in the funnel that is preventing engagement or transactions. In this case, hotspot identification is a way to widen or stretch the funnel stages in a targeted manner by taking corrective action focusing on specific customer behavior patterns.
