(verb): prod (someone) gently, typically with one’s elbow, in order to draw their attention to something.
(noun): a light touch or push.
Nudges. Your app’s full of them. You might not think about them all that often but they’re there – keeping your users on the right track. Moving them through the funnel toward some goal. That might be signing up to a mailing list, entering their credit card so they will upgrade from a free trial to paid, recovering an abandoned cart, or signing some important document so the deal your software manages will go through.
The point of nudges isn’t to harass your customers but to gently guide them when they go off track. Interestingly, nudges are a prediction against a model. You might predict that sending your customer a reminder to pay their invoice will result in the invoice getting paid. Or you predict that site visitors who see a lead magnet after reading the third blog post will enter their email address.
But how do you build these models? How do you test them? And how do you continuously optimize these models so that they reflect the current state of the product and the current customer segments you have?
Typically, you’d hire a data scientist to examine your data – web logs, SQL data, etc. and create a model that represents actions that customers take that lead to success (higher LTV or conversion from free to paid, etc.). Then, you look at behaviors that lead to anti-success: churn. But, most often predictive models that look at churn are predictive but not useful: people who never come back to your app churn. Oops, too late! So how do we keep people on the success path? We nudge them here and there.
Using these generated predictive models, data engineering embeds in your app, and your product management and development staff prioritize adding nudges. “100% of people who never connect their accounting package to our software churn.” Ok, great – let’s build a feature that nudges them to connect when they login if they haven’t done so. But, when do I do that? Immediately? After one week? A month? And, there are lots of other nudges I can do, too. How do I manage these?
Of course, too, your software and user base are constantly changing. What was predictive a year – or even a quarter – ago is most likely not predictive today. So you need to be constantly looking at each nudge and thinking about “is this the right one to pick? When? How do I pick which nudge to show today? Tomorrow?”
You need a system to manage nudges, when to show them, how to test them, and continuously optimize for an ever changing product and customer set. That’s what Kimball does. We’d be so happy to talk with you about optimizing your software, 1,000s of ways at a time. Please reach out!