My Data-Driven Aha Moment 💡

There seems to be no company nowadays that doesn’t state they make data-driven decisions. It seems so trivial to not guess your next steps, but simply rely on data to show you the way.

The problem with data is that it is useless in its raw state:

• You don’t know what to measure.
• You don’t know how to filter through the clutter.
• You don’t know which parts are significant
• It isn’t knowledge, it’s just data…

Just like the rest, in my startup, Kumba, an online house-call veterinarian service, data-driven decisions were part of our core values. I truly believed this concept but didn’t know how to go about it.

After several pivots (and hard failures), it finally hit me. Below is a (rather intuitive) plan on becoming a data-driven company.

1. Prepare a funnel (User Journey) — This is a fancy word for “what the user is doing”. This means that a potential customer goes through different stages during his lifecycle before he is willing to pay. The most common model is the Pirate Metrics Model (AARRR). It includes the following steps:

Moving from one step to the other down the funnel is called a conversion. You need to deeply understand and measure each step with what exactly the user is doing and what counts as a conversion.

At Kumba, we wanted to measure how many users are subscribing on their own. We created the following funnel in Facebook Analytics (don’t mind the mock numbers 😁):

Kumba Self Service Users Funnel

The steps are:

• User Engaged — User scrolling to the bottom of our website.
• User Converted to Pets — User converted from our website to our app
• Lead — Created a user (email & password)
• Complete registration — created a pet and completed the profile
• Purchase — one-time vet visit/yearly subscription

2. Track conversion rates — After creating the funnel and making sure all data is properly collected, now is the time to optimize 🎉 Calculating the conversion rate is easy. For example, in the funnel above, the conversion rate from our website to the app (User Converted to Pets) is 11/54=20%. This means that out of all the users who engage with our site, 20% percent of them will click one of our Call to Action buttons and redirect to our app.

This is serious! Say we could enhance this to a 25% conversion rate, in big numbers this could greatly improve our bottom line.

Now it’s time to start our DIET — Data, ideas, experiment, testing

3. Ideas — Looking at our funnel, we need to start raising ideas on each step of the funnel and what could increase our conversion rate. For example, if we want to increase site engagement, we could change the tagline of the site into ‘Modern Vetcare’ or something like that.

Each idea should have a complexity and perceived value attached to it. Text changes, for example, are the easiest to change (mostly without coding) while adding UI components require development resources.

After getting the final score of all ideas, they should be prioritized (somewhat like scrum).

4. Experiment & Testing — Taking the first idea from the backlog, we should define the experiment. Let’s take the tagline experiment. Our base condition is the old tagline and the other variant is a tagline of ‘Modern Vetcare’. The way we measure the experiment is via a goal we define. The variant getting the best results in terms of goal completions is the better one. This is called A/B Testing.

The easiest (and free) way to create A/B tests is with Google Optimize. You can easily import Google Analytics goals and have them as the experiment goals. The best thing is — you can do a lot of UI and text changes without coding!

Things to note:

• Statistical significance — You need to have many (at least a few hundred) experiments for it to be statistically significant. Having too few observations may reduce the confidence level of the results.
• One experiment at a time — You can’t have multiple experiments on the same page. This could pollute your results and you won’t know which experiment caused the results.
• It’s okay (and even great) to fail — A lot of ideas will probably not give you the results you thought they would. Just be encouraged that it cost you the bare minimum to verify it.

5. Repeat — Make sure you keep doing this for each new feature. Always think of the least effort required to conduct the experiment and how to measure success. Conducting a successful experiment is a satisfying experience.

In summary, you can use this step-by-step guide to quickly get you on the data-driven mindset. It will change the culture of your company and improve the decision making processes.

Feel free to ask me anything or suggest improvements 😊