Don't miss interesting news
Tier-3 often seems to be a quick way to get a lot of traffic. Indeed, there is a lot of audience here, the rates are lower, and the opportunities for scaling are open. However, as in any other direction, there is another side to the coin. In the case of Tier-3, it is technical instability, atypical user behavior, weak devices, and slow internet. All of this creates additional barriers in the funnel.
In an environment where even a second’s freeze can disrupt the entire interaction, not only the product itself but also the infrastructure on which it runs becomes key. Third-party services, which are often relied upon by websites, do provide a quick start and flexibility. But at the same time, they often become the point of ROI loss.
If at Tier-1 it can still be covered by volumes, then at Tier-3, every percentage of conversion is critical, and every unstable link in the system directly affects the final result.
This was one of the reasons why the DEVILS team decided to test its own PWA product in real Tier-3 conditions. Prior to that, DEVILS PWA worked stably on Tier-1, but it had not yet been tested in more complex markets.
To eliminate the influence of extraneous variables and get pure analytics, the launch was conducted in the format of a split test: the same creatives, budget, and target. The test lasted three months and covered three consecutive iterations, each with its own dynamics, mistakes, and growth points. GEO – Bangladesh
How it happened – in stages, further in the case study.
Iteration 1 – technical failures and drawdowns at the start
During the first launch, the results were lower than expected. Click2Install was only 5-6%, which was enough to call into question the profitability of the entire test. This figure was twice as low as the average rate for this GEO (10-14%) and, accordingly, made it impossible to drive traffic to the plus side with the existing bids and costs.
To find out the reasons for the drawdown, the team immediately went deep: technical logs, manual testing, and a full check of the entire interaction chain. And at this stage, it became obvious that the funnel had been failing from the very beginning.
The first problem is the redirect.
Due to the specifics of devices, the transition to Chrome simply did not work for some users, which led to a loss of traffic at the very start – literally at the level of the first click.
The second is the opening speed.
Slow loading of the landing page, caused by the weak Internet on GEO and an unoptimized server, forced users to close the page before it could fully open.
The third is native.
Even those who got to Chrome often faced with incorrect display of elements on the screen – and this instantly reduced the trust and desire to install PWA.
As a result, up to 50% of users got lost even before installation. It was this drop-off at the entrance that pulled the conversion down the entire funnel. It became clear that technical corrections were not enough – the interaction logic itself needed to be revised. And this is where the second iteration began.
Iteration 2 – localization
At this stage, the DEVILS team focused on the real user journey in the context of this GEO. The focus is on the details that build trust and influence the willingness to go further in the funnel.
In order to reduce the dropout rate, a number of point changes were made:
As a result, Click2Install grew by 10-12%. This was already enough to stabilize the entrance and launch traffic without critical losses at the start. However, at the deeper stages of the funnel, the situation remained difficult: EPC continued to sink by 8-10%, which clearly signaled that losses appeared after the installation.
Stage 3 – optimization of deep conversions: push strategy as a decisive factor
After the technical optimization, the funnel started working stably – the dropouts at the start were removed. However, at the deeper stages, inst2reg and reg2dep, the drawdown was recorded again.
Since technical factors no longer influenced the result, and only two variables remained within the split test – offers and pushes – the logical next step was to work with the push strategy.
The main assumption was that users were not motivated enough to return and complete the target action. That’s why the team decided to completely restart the strategy – from triggers and timing to texts and visuals.
After a separate analysis of the local behavior of the BD audience, the following changes were implemented:
In the new version of the strategy, the push notifications have become more personalized, precise in timing, and visually noticeable. This made it possible to partially relieve the load from inst2reg and reg2dep and bring the user to the target action.
As a result: EPC grew by 10%, and DEVILS PWA showed ≈$2.2 versus ≈$2.0 in the control group that worked on a third-party service. It was this advantage that ensured the final positive ROI of the campaign.
DeVILS PWA on BD case study results
Given the specifics of this GEO, namely, working with a large volume of traffic and impressions, the starting metrics simply did not fit into the economy and did not allow the campaign to be profitable.
But thanks to in-depth analysis, systematic work with user experience and adaptation to local conditions, we managed to level the funnel and achieve consistently better results.
This case clearly demonstrates: Tier-3 is not only about volatility and potential, but also about the willingness to rebuild the funnel when the basic scenario does not work. And what is even more important is not to be afraid to implement radical changes and at the same time work systematically, working through every detail.
It is a comprehensive restructuring, not point fixes, that gives stable growth and measurable results. In this case, it is the growth of key metrics and the final positive ROI that can be scaled.
