A very fascinating paper from Amazon's Science team was surfaced recently:
"How do Advertisers React to Changes in Ad Auction Pricing?" (July 2017)
In this paper, Amazon studies how advertisers, i.e. sellers and vendors, react to changes in the ad bid algorithm. This is further proof that Amazon puts a lot of effort and brains behind the changes they make to their platform.
"We test the impact of a decrease in an ad auction’s soft reserve price which effectively reduces the cost paid by advertisers."
(The "soft reserve price" is the minimum bid required to win an ad placement. If no one is bidding at that threshold, an ad won't show. We aren't told what the soft reserve price was or how much they changed it in the experiment.)
"The experiment aims to test the hypothesis that advertisers would respond to the decrease in cost paid by advertisers by increasing bids and/or budget..." i.e. if Amazon makes ads cheaper, advertisers will bid higher and spend more (to try to increase their reach more)
Conclusions of the Research:
1️⃣ The team developed a new experimental design for this study that improved upon standard experiments
2️⃣ Amazon's goal was to " identify optimal pricing schemes to jointly maximize benefits for both customers and advertisers." (Yes, Amazon makes a ton of money off ads, and it makes the platform pay-to-play, but they still want to test what provides the best customer experience and drives the best sales for advertisers.)
3️⃣ The results showed by reducing the "soft reserve price" advertisers paid 10.67% less for their ads. In turn, advertisers increased their bids and budgets 0.51%. This led to a net *savings* of 10.06%.
Some notes:
1️⃣ This paper is from 2017, so it may have laid the foundation for some of what we experience in 2024 Amazon Advertising, but it's unlikely this is directly being implemented today (when we're all finding out about it)
2️⃣ Amazon is way smarter than me! I had ChatGPT summarize, which helped, then I read through the paper myself. I have a BS in Chemistry and took a fair amount of math (through Calc III), but I barely understood most of what they were talking about. (Shocker, I don't have a PhD in Econ!)
3️⃣ The research team included, from Amazon: two Sr. Economists, a PhD in Computer Science, a Sr. Data Scientist, and a PhD in Econ. All but one of them still work at Amazon, by the way (something unheard of on the retail side)
4️⃣ Changes like these could impact small advertisers more than large ones because it requires you to be constantly tracking your ads results and adjusting bids and budgets - if you're paying attention, you can win more
💥 My favorite line from the paper: "We may then compute the Hajek inverse propensity-weighted estimator" because it sounds so awesome and I have no idea what it means
Jeffrey Cohen give this team a high-five for doing cool research!
Spencer Millerberg will you be running some cMRD tests on keyword changes?