How to build a multivoron using the example of Mailchimp and Google Analytics and optimize it

We all know about A / B tests and other conversion optimization techniques, we constantly use them. However, it is not always clear to what aspect to pay more attention and what change will give the best result. It’s clear that if the landing page is tangled, registration is broken or activation in the product requires huge labor costs from the user, then testing the text of the letter and its call-to-action will not bring tangible fruits.

What to do if the situation is not at such an extreme point, and the problems are blurred throughout the user's path (from the letter to the product interface)? Of course, you can simply make up several variants of the letter, vote with the team, even conduct an A / B test, send it to users and calm down, but why, if as a result we’re ineffectively emptying the channel?
')
Any business or startup is a complex mechanism that cannot be managed without feedback. To get the maximum effect, we will compose the text of the letter, send out the first batch of users, measure the performance at each stage of the funnel (Letter - Landing Page - Product), gather the team to analyze the results, find bottlenecks and generate ideas, implement the ideas and repeat all over again. And then again. And further. Rushed?

Here is one of our cases. During the closed beta, we actively sent out invites, but a number of users did not receive them. This is the audience with which the product is interesting, and therefore it is worth reminding them that we graduated and we have a free tariff.

I will adjust the funnel and experiment with the help of our hopox service.

It all starts with an idea. Here’s how it sounds:



If you noticed, I immediately put down the expected result. This is a kind of forecast and at the same time the minimum threshold for the success of an idea. If there are fewer orders, the cost of implementing the idea is unlikely to pay off and scaling is out of the question here.

This idea, like any other, requires a number of preliminary works for implementation:



So, the tasks we have described, distributed to colleagues, now is the time to write a letter. The original version contained a couple of buttons, links to the manual (a little about HADI cycles and the ideology of data-driven marketing), and a link to the registration in the service. In general, enough to test.

In order to track the user's path from the letter to the product, I marked the link to the landing page with a UTM tag. The landing page is configured in such a way that it translates these tags into a product. Thanks to this, now I can configure the metric in the hypothesis so that the service considers purchases only by those users who were attracted by our email campaign:



After the end of the testing period for the hypothesis, we received 0 purchases. The hypothesis failed. Why did this happen? Obviously, something prevented users from passing through the funnel from start to finish. In order to understand exactly what, we need to build a funnel and analyze the bottlenecks.

For work of a cyclical nature, as well as for work with key business metrics in hopox, there is a “Summary” section. This is the very same business management panel that links the team's experiments with their results in a quantitative form. In this panel, you can always see how the implementation of a certain idea influenced key indicators, how quickly a team moves, how many ideas it generates, how focused, and where a business or startup flies.

The funnel is such a kind of conveyor belt, on which the user moves, and in our interests - to hold it to the end. I want to note that hopox obtains all data from various analytics sources, and, accordingly, an important value is the ability to build multi-gamers, where each stage can represent data from different sources (After all, the user's path is often not covered by a single analytics system, as in our example).

As a result, our funnel should look like this:

• Number of letters read (source: Mailchimp)
• The number of transitions to the landing page of the letter (source: Mailchimp)
• The number of users who clicked on “Get Access” on the landing page (source: Google Analytics)
• Number of users who clicked “Pay” in the service (source: Google Analytics)
• Number of users who hit the final payment page (source: Google Analytics)

Thus, with the help of one funnel, we link the statistics that Mailchimp and Google Analytics give us, which allows us to see the whole picture:



The naked eye can see that the main problem is that people do not follow the link in the letter. As mentioned above, the letter contained several links (to the manual, to the survey and to the landing page), which not only blurred attention, but also simply distracted from the main purpose of the letter.

We sent the next version of users the next version of the letter, in which there was a short reminder of why the service is needed (after all, the person received the previous letter more than two months ago) and an explicit button that leads to the landing page.

The conversion of the first stage of the funnel increased to 17%. Understandably, if a person receives a letter from a service that he has subscribed for a long time, you should not overload him with information. We recorded this case in the conclusions of the hypothesis and are ready for the next iteration.

Assessing the next stage (the number of people who switched to the service), at the next growth rally, we described about fifteen ideas related to landing, which could affect the conversion of the transition to the service. Assessing the deadlines for implementation and the expected effect of each of them, we stopped at the fact that we need to critically reduce the amount of information on the landing page, conduct A / B test of value proposition, button position, remove the extra “Name” field and replace the static image in the header On video with the service interface.

After all the changes, our funnel looked like this:



Having worked with the remaining two steps, we increased the conversion of the last two to 12% and 83%, respectively. At each growth rally, we continue to generate ideas for increasing the overall conversion of this funnel.

Every campaign, every channel, every user segment needs constant optimization. The found patterns and the fired hypotheses sometimes turn out to be a revelation, and sometimes a banal truth about which they simply forgot. Experiment every day, try new ideas, get feedback from your business in the form of metrics and invent new ways to grow.

We continue to work on the hopox service, which helps to quickly check ideas and focus the team on growth. And yet, we got a free rate of unlimited time.

Thank you for your attention, hopox team.