At the beginning of 2017, Compareitor started to have problems regarding the cost of acquiring traffic, in 2 months the cost was multiplied by 3 and the CEO ordered me to look for alternatives to normal acquisition channels. He intuited that we can move part of the budget in AdWords to promote the app and keep the acquisition costs low.

So I started to collect information about it, there were many pieces in the puzzle and I needed to arrange them so that everything made sense. These were the pieces

  1. In mobility user’s tests we see that our user archetype wants to solve the problem of insurance renewal in less than 3 min (we call it “micromoments”) while the web users are willing to spend 6,3 min.
  2. In June 2017 for the first time in the history of the company 60% of our traffic came from smartphones. All Compareitor products are desktop minded.
  3. It is not the same to invest money in AdWords than in app stores. The evaluation of the app is key for the conversion rate. If you below 4.4, the conversion will decrease by 10% for each decimal that you lose in valuation. Therefore if you put € 10,000 in AppStores and your app does not provide enough value to the user in the medium term you will lose much more. There is a multiplier effect on this if you don’t perform well.
  4. The reports on the use of apps in smartphones in Spain tell us that the time the user uses his device is distributed as follows:
  • 64% Messenger App (WhatsApp, Facebook Messenger & Telegram)
  • 23% Social Networks (Facebook & Instagram)
  • 10% E-commerce (Amazon, Zalando, Just Eat, ….)
  • 3% Other apps

So we are fighting for 3% of user time, so my thinking was: Why do not we take Compareitor to where the user is? Let’s adapt the user experience conversational channels

I started doing one quick research, I clustered the Compareitor users of the last year and created an artificial but representative profile of each cluster. I focused on the most profitable profile for the insurance vertical and started designing for it. I also talk to the customer care dept. and we did 5 interviews with real profiles that came out of the studio. I tested my hypothesis with the archetype found and it seems to be true. The user would be willing to interact with Compareitor through Facebook or WhatsApp and the UX is good.

So I designed a plan to do an MVP for the auto-insurance product. We discarded launching it on WhatsApp because it’s business version was not yet in production and we focused on Facebook Messenger. I evaluated 4 providers of these technologies but finally decided to develop it internally.

There was a great constraint for the project, Compareitor core service did not have an API. So the first thing I had to do was convince the CTO to develop it and define its specifications. Once the issue was unlocked it took 3 months for the team to have the API running in production.

I defined the following architecture with 4 layers

  1. Facebook Messenger as user layer.
  2. Chatfuel as chatbot editor.
  3. Chatbot engine to connect Chatfuel to Compareitor.
  4. API Compareitor as insurance price provider.

The customer journey on which we work was this:

We capture the user by Facebook campaign, we take him to facebook messenger and there we start an automatic conversation with user that reproduces the questions of the web form. Once the user completes the journey we show him prices and transfer it to the insurer he has chosen.

The design of the user experience and its compliance with the Compareitor policies was a great challenge. The integration and adaptation to the requirements of Facebook also demanded a great effort both of this challenges would be susceptible to a separate case study.

We validated the hypothesis and we managed to reduce the time it took to collect the user’s information, the goal was to go from 6 min to 3 min and we stayed at 3’5. Doing so we also reduce the churn rate, we went from 43% in the web to 32% in the Facebook Messenger.

We found two “easter eggs” that helped the project to work well.

Acquisition cost. We discover that Facebook wanted to promote this type of integrations and gave us more visibility to our acquisition campaign. If the user acquisition con in around  1€, we expected it to be 0.50€ and finally it was 0.35€

Abandoned Conversations. We found that the Facebook Messenger framework has a native mechanism to remind you that the conversation has not ended, so this retargeting that is normally done through email was integrated here and was one of the causes of the improvement of the churn rate.

Although the funnel data of this experiment was very good in the final part, the user was not as good qualified as on the web. If the Quote (user see price page)  to Lead (user transferred) on the web is 42% this new channel was around 33%. Even so, being cheaper, it compensated for the fall of the quote to lead and above all it was an MVP with many things to optimize.