Moving the needle
Data and analytics can go a long way to improve outcomes. Once a baseline is established and a model is running in production, you may start to question how to make more significant changes. Sometimes you’ve thrown all existing data points at a problem and they aren’t explaining the story fully enough. The fact is, slight incremental changes don’t always improve things, and you may start to imagine other ways to move the needle.
An example problem could be identifying signals in user behavior within your product. The usage data you have doesn’t distinguish the patterns well enough to rely on the model outcome, and you’ve explored all existing variables thoroughly. For this problem, moving the needle requires new data points, and there are different ways to tackle it. Other than buying third party data, you can attempt to build it yourself.
PRODUCT FEATURE
One of the best ways to introduce signal is to add a hand-raising feature within the product. Instead of trying to infer something based on usage patterns, imagine how your customers can offer the information themselves, preferably with minor impact to a product team’s ongoing initiatives. Is it communication through a sharing option? Or prompting a selection to choose from that highlights clear distinctions? There could be a number of ways to add novel insights. It’s possible the Product team isn’t able to make changes easily to support new features though, and in that case there’s another option.
CHECKOUT PROCESS
Requiring mandatory field completion when purchasing a service online is a situation we’ve all experienced. Typically, companies want to know their customer demographics and reasons for purchasing. Without making the process too cumbersome, it might be easier and faster to request the information up front. A common one is requesting a company name, even if you’re purchasing a product as an individual. The quality of the data becomes questionable with free text, but it can certainly be useful.
In every data project, we should be asking ourselves how to move the needle to improve business outcomes. It’s a creative aspect of the profession to imagine such possibilities. It requires cross-functional collaboration to implement outside of your data realm, but it’s rewarding and potentially impactful to the business, in ways that existing data might be insufficient to produce.