Goals
- Develop promotion strategy for a highly competitive niche
- Assess display ads’ contribution to sales
Tools used
- Yandex.Direct / Video display ads
- Yandex.Metrica / E-commerce
- Yandex.Metrica for display ads
- Yandex.Audience / Experiments
Period: 13.06.2019 — 26.06.2019 (14 days + two-month attribution window)
Step 1. Select display ads as a promotional channel
Samsonite is one of the oldest manufacturers of suitcases and bags, founded in 1910. The brand promoted its own online store on Yandex Search, where it had to compete with larger multi-brand stores who purchased nearly all the available traffic.
Samsonite decided to focus on display advertising: they launched a brand awareness video in the Yandex Video Network that targeted users interested in traveling as well as those with high incomes.
Our goal was to increase brand awareness, especially among younger audiences of active travelers that value comfort. We also wanted to increase traffic to out site at the start of the high season.
To this end, we chose OLV (online video) campaigns.
Step 2. Set up experiments in Yandex.Audience to compare the two segments
Experiments in Yandex.Audience
Yandex.Audience Experiments is a new A/B testing tool for comparing the effectiveness of ad settings and creatives. The tool randomly distributes users in the selected audience into two non-overlapping segments, ensuring the experiment’s statistical purity. Each segment’s effectiveness can be tracked in Yandex.Metrica.
Samsonite decided to compare the behavior of two groups: users from the test group saw the video ad, and users from the control group who did not.
They created an experiment in Yandex.Audience: they entered a Yandex.Metrica tag installed on their site and divided the audience into two identical segments — a test and a control group. They then linked the experiment in the settings of the related campaign in Yandex.Direct: only half of the available users would see ads from the campaign.
Step 3. Enable Yandex.Metrica for display campaigns to track post-view results
Yandex.Metrica for display ads
Yandex.Metrica for display ads makes it possible to analyze display campaigns in Yandex.Direct based on performance metrics using data about conversions that took place after the ad impression of click on the creative.
They connected the tool in two steps:
- They added the campaign to Yandex.Metrica for display ads and filled in the data about the site: the placement period, cost, format, etc.
- They downloaded a tracking pixel and entered it in the ad settings in Yandex.Direct.
Results
- the video clip was viewed 4 million times, with 1.8 million of them 100% views
- 38,000 users visited the site
- the number of visitors peaked during the placement period
- users who could see the ad visited more often than those users who were not shown the ad.
The experiment showed the variations in revenue from purchases by users in different segments
Revenue from the segment of users who were shown ads was 4.9 million RUB higher than that of the segment of users who were not shown the video. Yandex.Metrica for display ads showed that confirmed post-view transactions alone generated revenue of 2.7 million RUB with a share of ad spend equal to 40%. The overall conversion rate from impressions to purchases amounted to 0.007%.
Sometimes two analytical tools may show different observed conversion rates due to their operating principles. If a user sees a banner on their work computer, but makes a purchase from a smartphone or home laptop, Yandex.Metrica for display ads will not link that conversion to the display ad, which may lead to an underestimation of the overall result. Experiments offer cross-device support: a user in the experimental group and saw the banner on their work computer and then made a purchase will be counted in the correct Yandex.Metrica segment, even if the purchase was made from another device or browser.
In Yandex.Metrica for display ads, a post-view analysis of the audience reach by the ad showed the following:
Age. Users aged 35-44 showed twice as many conversion as the 25-34 segment and 3× higher than the 55+ segment.
A key indicator of a placement’s effectiveness is the ratio of CPM to total revenue per impression. In the future, to maximize efficiency, we will try to run ads for segments that have shown the best revenue per impression, for example:
- bathroom furniture: income for impressions 11.5× higher than average
- tennis: 5× higher
- tiles: 3.8× higher
- computer equipment: 3.8× higher
- gifts: 3.5× higher.
Conclusion
Video advertising is a powerful branding tool that can have a visible effect of sales. Yet it isn’t always easy to see this effect, especially in numbers. Samsonite was able to track the effect of video ads by comparing the test and control groups and by taking delayed conversions into account.
You can increase the accuracy when evaluating the performance of a campaign by using experiments and Yandex.Metrica for display ads. Using these tools allowed Samsonite to solve a complex problem:
- Using experiments, they were able to measure the sales uplift in the test group compared to the control group and calculate what percentage of revenue display advertising brought.
- Using Yandex.Metrica for display ads to analyze results in more detail using post-view and create a user portrait.
We are a CPC agency, so it was important to measure the effect ads had on sales instead of paying attention to brand metrics. The effect on sales did not come immediately — the large part of the effect was observed in the second month after the ad campaign had stopped.
Video advertising in Yandex and the accompanying experiments in Yandex.Audience and Yandex.Metrica for display ads allow us to assess the contribution of video ads to conversions and find insights.
