Sponsored Brands video helps increase sales and click-through rates

By: Zee Shah, Senior Media and Analytics Manager, German Schnaidt, Applied Scientist and Ashton Brown, Technical Writer

In this dual-method study we found that both sales and click-through rates increased when combining Sponsored Brands video with campaigns already benefiting from a “better together” approach of Sponsored Products and Sponsored Brands.

Story highlights:

Over the past few years, we have researched the effectiveness of different ad products. We began by testing the individual effectiveness of Sponsored Products and Sponsored Brands. Next, we moved onto testing the combined effectiveness of Sponsored Brands + Sponsored Products. Our analysis found that these two programs work better together. In this analysis, we take that research a step further by testing whether sales (year-over-year) and click-through rates increase, decrease, or remain the same when combining these two programs and adding a third program, Sponsored Brands video.

To test the causal impact of adding video to the initial two campaigns, we control for other campaign attributes (e.g. total sales, units sold, average selling price and total ad spend), meaning, we created two categories of advertisers, those who used the original combination and those who added video. We then use those pairs to estimate the causal impact of adopting video.

To conduct our analysis, we selected brands within the United States and Europe (France, Germany, Italy, Spain, United Kingdom) that were using Sponsored Products + Sponsored Brands, but had not yet adopted Sponsored Brands video. Once brands were identified, we used machine learning and modelling to conduct two causal analysis tests:

  • Short-term (between December 2019 and November 2020): Examined the causal impact of adding Sponsored Brands video to ad campaigns. This shorter-term analysis examined the month-over-month impact of brands that combined Sponsored Brands video with Sponsored Products + Sponsored Brands campaigns.
  • Long-term (between January 2019 and December 2020): Examined the year-over-year (YoY) sales impact of brands that used Sponsored Brands video + Sponsored Products + Sponsored Brands versus brands that only used Sponsored Products + Sponsored Brands (while controlling for other variables such as selling price, total ad spend, etc.

For more on how we performed this study, see the Methodology section at the end of this article.

In the short term, brands that adopted Sponsored Brands video for the first time saw a 21% sales increase the next month, compared to those that didn’t

When examining a causal relationship, it is important to establish if and when results occur. To test if or when results were seen when adding video for the first time, we ran a short-term study that analyzed the next-month impact of adding Sponsored Brands video. We found that there was a 21% sales increase in the month after for brands that adopted and combined Sponsored Brands video to pre-existing Sponsored Products + Sponsored Brands campaigns.

Next month sales % increase when combining Sponsored Brands video with Sponsored Brands + Sponsored Products campaigns

21%

In the long term, brands that added Sponsored Brands video to Sponsored Products and Sponsored Brands campaigns increased sales by 10% and CTR by 25%

To determine the impact Sponsored Brands video has on campaigns, we found and compared brands who used all three products for 12 months to brands who only used Sponsored Products + Sponsored Brands. We found that brands that included Sponsored Brands video into their mix had 10% more sales YoY and a 25% higher click-through rate YoY.

Long-term impact of adding Sponsored Brands video to campaigns

10%

Long-term sales % increase when adding Sponsored Brands video to campaigns already using Sponsored Products + Sponsored Brands

25%

Long-term click-through rate % increase when adding Sponsored Brands video to campaigns already using Sponsored Products + Sponsored Brands

Conclusion

Using a dual-method approach, we tested if the “better together” results (regarding sales and click-through rates) of Sponsored Products + Sponsored Brands would increase, decrease, or remain the same when combined with Sponsored Brands video. Our analysis revealed that brands that adopted Sponsored Brands video saw a positive, causal impact in both the short term and long term.

Methodology

To perform this study, we used a dual-method approach consisting of a shorter-term casual analysis focused on determining if brands that combined Sponsored Brands video with Sponsored Products + Sponsored Brands would increase sales or CTR (the next month), and a longer-term causal analysis focused on YoY growth of sales and CTR.

Both methods are detailed below.

Short-term causal methodology

To measure the causal impact of advertisers who adopted Sponsored Brands video for the first time, we employed a machine learning causal inference methodology inspired by techniques [1], [2], [3] to determine the effect of taking an action on advertiser performance in a shorter term of one month. Our current methodology follows a method called 2-stage GP (2-stage Gaussian Process) that shows improved performance on various causal performance metrics as compared to existing methodologies such as Double Machine Learning [1] and Causal Forests [2] when applied within the context of adv

For this study we selected over 78,000 advertisers in the US marketplace and matched 25,000 of them using this methodology. 78,000 advertisers were in the input dataset for the evaluation, and 25,000 samples (treated and non-treated) have been used for the propensity score.

Long-term causal methodology

To measure sales and click-through rate (CTR) impact in the longer-term, we employed causal analysis techniques to determine the impact of taking an action on advertiser performance in a longer term of 12 months. First, we created two bins. In bin one, all advertisers used Sponsored Products + Sponsored Brands. In bin two, advertisers combined Sponsored Brands video with Sponsored Products + Sponsored Brands. To control for other campaign attributes, we ensured that brands were similar in campaign attributes such as: total sales, units sold, average selling price and total ad spend.

This allowed us to compare sets of bins with similar probabilities to adopt Sponsored Brands video. To do this, we used machine learning to measure the propensity scores for each brand based on their ad spend, total sales, total units sold, total impressions, total clicks and average selling price.

Attributes used in propensity score calculation: Natural logarithm of total sales 2020, total units sold 2019, average selling price 2020, total impressions 2019, total clicks 2019, total ad spend in 2019, total ad spend 2020 and total sales 2020.

Response variable: Natural logarithm of CTR, natural logarithm of total sales growth rate in 2020.

Sources

  • Alaa, A. M. and van der Schaar, M. “Bayesian nonparametric causal inference: Information rates and learning algorithms.” IEEE Journal of Selected Topics in Signal Processing, 12(5):1031–1046, 2018.
  • Hill, J.L. “Bayesian non-parametric modelling for causal inference.” Journal of Computational and Graphical Statistics, 20(1):217–240, 2011.
  • Pauwels, K., M. Caddeo and G. Schnaidt. 2022. Causal impact of digital display ads on advertiser performance. In: Proceedings of the European Marketing Academy, 51st (108183): EMAC. http://proceedings.emac-online.org/pdfs/A2022-108183.pdf
  • Van der Schaar, M. and Alaa, A. “Bayesian inference of individualized treatment effects using multi-task gaussian processes.” NIPS, 2017.