How we increased the conversion rate by 2.6x on credit traffic
Hi, My name is Igor Gorodko. I work as an interface developer in one cool financial affiliate program and want to share with your experience, how I independently increased the conversion in applications from our sites in 2.6 times, due to the introduction of bots from the guys from Twintwoo.
Our company has been working in the financial direction for more than 5 years, where the main task is to form the flow of customers interested in obtaining a loan. After learning about the Twintwoo project, I decided conduct an experiment and find out which bot is more effective. Our self-written or Twintwoo bot.
The company works with many different traffic sources. IN in particular, landing sites are used, which receive 80% search traffic. There were 15 such sites at the time of the case creation. An example of one of them can be found here: creditkarm.ru
A self-written bot was placed on all sites. The result of the dialogue with him were applications, in other words applying for a loan.
I have set myself the following tasks:
- compare the conversions of our bot and the Twintwoo bot
- try to increase conversion and number of applications
- test new ideas and hypotheses (run split tests)
Looking ahead, I will reveal the results before and after the experiment:
|Traffic for 30 days||46.147||46.094|
|Converting to the beginning of the dialog||7.3%||20.1%|
|Conversion to Lead||4.5%||12%|
«2.6 times growth is an excellent result that exceeded all expectations!» - Igor Gorodko
Previously, the company worked on this task 5 people (front-end, designer, back-end, analyst, tester). Of course, this was not their main task, nevertheless, it was rather long a period of time was given to our bot weekly.
The work was similar to a routine process, where the analyst periodically looked through the Webvisor, sketched variants of hypotheses in the bottlenecks of conversion drawdown, designed tasks and then returned to their paramount responsibilities. Usually updates happened once a week, sometimes, fixes were made to the same day.
First - moved the entire the logic of our bot in Twintwoo in 2 hours!
The bot assembly process resembled a game format, where you need to add components to the scene and connect them connections. Anyone with experience in Figma, Miro, Sketch will find it incredibly interesting and an intuitive process.
“It took about two hours to create the first bot. In such a short period of time, I have learned many things and understood the algorithm of creation. I think next time I will need no more than half an hour for formation of a new bot "- Igor Gorodko
Second, we carried out integrations and launched traffic!
Integration with our server was implemented very quickly and easily. Bots include a section "Automation", in which you need to select the WebHook method (You can send via Zapier or CRM, but we chose WebHook). Next, we set the settings according to the specification of our API.
Upon completion of the bot integration process, we conducted a test on one of our sites in order to identify applications. The result was not long in coming, applications began to arrive! By the end of the day, it was already clear that the conversion increased, and we immediately installed the bot on all sites.
Third - We started to squeeze maximum through splits
Now for the fun part! In the first version of the bot, after a welcome message, the bot was asked users - "does he want to get a loan?" and then two options appeared: Yes and No. I decided add this plot to the split test. Tk a large percentage of users chose the option - no. Second the branch became the option where the choice of the amount of the desired loan was immediately offered. Which was implemented through component - slider.
The second option with a slider is about 40% more efficient!
I also made a split before the user data collection form:
- first option is a call to action, “Fill out the form ...”
- the second option told how much the client can expect and also a call to action.
The second option also won, with detailed information - why fill the form. This also added ~ 15%.
Two more split tests are currently running in my bot.
- The first test is a welcome step. We sketched out some ideas with different pictures and different text.
- Second test - variant with one solid large form or variant with several forms coming in a row.
Interesting fact: Did you know that all the data that the user fills in, even if he does not come to the end, in any case, will get to us
P.S. how the current splits will end, I will definitely add the results to the case. 😎
Were there any problems?
During the development of the bot, a couple of times we encountered bugs in form fields such as:
- field "city" (there are many nuances associated with autocomplete)
- the age field.
But all our shortcomings were promptly eliminated by the support service within 40 minutes. For this separate thanks.
The conclusions are obvious
Summing up the work, it is worth noting that before meeting Twintwoo, we spent a lot of time and energy and development resources. Sometimes, going into this process, we forgot about the goals and results for which everything was conceived. Now, we are focused on new ideas and hypotheses, the process has become creative, fast and very agile. For the development of a whole direction, only one person is needed who periodically adds new ideas to the bot and monitors the growth of conversion.
Now we have plans to integrate the bot as advertising on many sites, which for us is another point of growth in traffic and requests.
Try it, experiment and you will succeed too! Simplicity and efficiency are key success factors. These are the words we would describe our collaboration with Twintwoo!