Jul 19th 2021
What's New: Geeklab 2021 H1
Our users may have noticed a lot of changes and improvements to the platform over the last few months. We've been able to push at least one major update to the platform each month, introducing new features, and at least one additional smaller one with performance improvements, thanks to our continuous two-week springs.
We've already launched a new internal taxonomy, Geeklab API support, Review Profiles, Steam Product pages, fully localised pages, demographic data, and, most importantly, our completely free plan in 2021.
This post will summarize our most recent addition to the list. Completely redesigned Campaign Details page, including a new bayesian statistical model to complement the frequentist model.
New: Campaign Details Page
We've split the full campaign details view to five different tabs. We've got Overview, Traffic, Visuals, Engagement and Compare tab as separate, unique views of your campaign.
This is where you'll find your most important metrics. The tab displays the variant's unique number of Impressions, Installs, and Conversion rate. In addition, you can see the average time to install and the split between decisive and exploratory users. All of the data you see and receive from Geeklab is first-party data. We do not collect any third-party data, which results in extremely low discrepancies between the ad source and Geeklab impressions. Furthermore, we can determine whether the user is a returning visitor or is visiting the page for the first time.
We've always been fans of bayesian statistics, which is why we're so excited to finally introduce our own bayesian models to Geeklab to complement the frequentist approach. While the frequentist approach provides you with the harsh truth, absolute CVR rates, and confidence rate for that specific campaign and only that specific campaign, the bayesian models we've developed take into account the millions of people evaluated on Geeklab each year. We run around 10.000 simulations through the model using each variant's CVR distribution to get the winning odds. The bayesian model also enables us to display the potential CVR distribution for each variant.
The traffic tab provides data points that can be used to validate hypotheses. The Traffic tab displays the CVR over time chart, as well as daily install and impression cohorts. These data points aid in the detection of anomalies in the data. During any seasonal test, the CVR overtime metric is extremely exciting. You can actually get a day and time when people are sick of bunny ears during Easter by using it.
In addition to the daily cohorts, the traffic tab displays demographic data if enabled. While the free plan allows you to collect demographic data, the results are only available for subscribers and unlocked campaigns. For each variant, you can sort the results by age or gender. The overall distribution by age and gender can also be seen.
Visuals have a significant impact on the performance of your page, so they deserve their own tab. The data points from both screenshots and videos are included in the visual tab. The CVR rate for each screenshot is displayed as a chart at the very top of the page. You can see even more data points for each individual screenshot for each variant below the chart. Total views, view rate, clicks, click to install rate, installs, and conversion rate are displayed for each screenshot. We also show the decisive view, which is the view that is rendered for the viewer immediately and does not require any action on their part.
If your product page includes a video, the most important data points will be displayed below. We'll show the CVR rate per viewed second for each variant and video separately. We also have an engagement chart below it that shows when users dropped off.
The Engagement tab displays scroll heatmaps, which depict how users scrolled down the page. Each page contains multiple pixels, and whenever a user scrolls down their page, we record all of the pixels that are visible in their viewport. We'll show a powerscore for each position where the pixels are laid out. The amount of engagement on the specific position is used to calculate the power score. Each variant also received its own unique Engagement score, which is based on how evenly the powerscore is distributed.
While every scroll is important, so are clicks. Every click on the platform is recorded and displayed as a chart for each variant. It allows you to see how many people read the entire description, exit the page, and so on.
Last but not least, we have our compare tab, which allows you to quickly compare the variants. Compare all of the variants to quickly explain what the test was all about. With our most recent update, we also added the compare chart, which allows you to track the performance of your app over time. Examine the trendline and where your most recent campaigns stand in relation to historical data.
That's a good overview of the first half of 2021. We can't wait to share the big news and updates we have planned for the beginning of the fall to go along with the options that Apple provides us with over 30 customisable product pages. Until then, we will continue to update the tool twice a month, introducing highly requested features and innovative solutions that will allow all of the wonderful apps and games to learn and grow.