How We Combined AI with the Human Touch to Achieve 30% YoY Growth for Our Client — Part I

AI is on everyone’s lips these days, but the point some people miss is that AI is a tool, not a replacement for the human touch. We recently leveraged the latest advances in Amazon Advertising’s reporting tools to achieve a 30% increase in YoY (year-on-year) top-line revenue for one of our clients over the T5 period — the five days between Thanksgiving and Cyber Monday. And we did this while the world was in a so-called recession.

T5, also known as Turkey 5, starts on Thanksgiving and ends on Cyber Monday. It represents the five biggest retail days of the year in terms of both traffic and sales.

We have discovered through chatting with many brands, that relying too much on AI without the human element simply does not work. In this two-part series, we’re going to show you exactly what we did, what tools we used, and how you can also leverage those tools to drive higher sales by combining their immense potential with the human sentience necessary to achieve maximum return.

In Part I, we’ll focus on the technology — the tools themselves. And in Part II, we’ll get into how we used human intervention and our experience to maximize the effectiveness of those tools.

Amazon Marketing Stream — what is it and what is it used for?

Amazon released Amazon Marketing Stream in June 2022. This tool delivers Amazon Ads campaign metrics every hour. Previously, this data was only available with a 24-hour lag. To get around this, developers often queried the existing APIs several times a day but this created other challenges.

Amazon Marketing Stream provides hourly performance data for the targeting expression by placement. It also provides data on budget consumption, traffic, conversions, and the respective data dimensions for use in analytics.

One important point to note is that marketing stream data does tend to lag by about two hours, so it definitely isn’t real-time. But it’s better than anything Amazon has previously provided.

By combining the data insights provided by Amazon Marketing Stream and PointStory’s advertising management expertise, rapid action could be taken to improve a campaign’s performance.

The tools and their features

Amazon Marketing Stream pushed data to PointStory’s data warehouse. This data was separated by day and hour, giving PointStory a large dataset of empirical data to refine bids by the hour. By the time T5 arrived, we had already painted a portrait that contained a 90-day historical view of every hour of each day of the week.

Using this insight, we leveraged existing Amazon Advertising features to automate and optimize the client’s campaigns depending on the time of day, and the day of the week.

Let’s take a look at each of these features:

Rule-Based Optimizations

The data for T5 would not be entirely reliable because Amazon Marketing Stream was only launched in June 2022 and so we didn’t have the previous year’s T5 data. But there are patterns that remain consistent, such as the fact that sales and traffic drop markedly in the very early hours of the morning when most people are asleep.

Using this data, we set up Rule-Based Optimizations to increase and decrease bids accordingly, depending on the time of day.

Competitor brands and ASINs

Amazon campaigns offer various targeting options that help get eyes on your product. One option is to target competitor brands and have your products shown next to products of that brand.

This can be as specific as a single ASIN, or more generally for products in a specific category and of a specific brand. You can also narrow your placement down further according to price range and average star-rating.

For example, if you’re selling men’s watches and your watches compete with Tag Heuer, you could bid to show ads for your watch on any Tag Heuer product detail page within a certain category. If your watch is high-end, you can narrow the price range to more closely match your watch’s price so that you improve your chances of reaching qualified shoppers.

We paired Amazon’s advertising capabilities with our experience and created a list of ASINs that closely matched our client’s products by price, feature, and color, (e.g. “stainless steel men’s watches” priced between X and Y). We then targeted these competitor ASIN product detail pages for our client’s ad to appear on them.

Finding the ideal products requires data. We used a popular industry tool called Helium 10 to find those similar price-bracketed products. Once we had that list, humans made the final call on which ASINs to target.

Competitor keyword targets

For a search-based campaign, we likewise targeted competitor brands but at a keyword level. Returning to our fictitious watch example, we might use “Tag Heuer” as a keyword to target searches for this brand name.

Placements — top of search, product pages, rest of search

In Sponsored Product campaigns, you can define a percentage adjustment for your bid so you can hit the coveted Top of Search placement. Top of Search refers to the Sponsored Products placement found on the first page of the Amazon search results page.

Let’s see what comes up when we search for “Stainless steel men’s watch.”

The Top of Search placement shouldn’t be confused with Amazon’s Headline position which sits prominently above the search results. That placement is for Sponsored Brand campaigns only. That’s the position highlighted in blue below. Top of Search is the position just below the word “Results.”

Scrolling down, we get to the “Rest of Search” results. “Rest of search” is defined as the Sponsored Product results found on page two and beyond of the search results pages.

Although the following products appear at the top of the second page, they are considered Rest of Search because they are not on the first page.

Similar to Top of Page bid adjustments, you can also adjust bids by a percentage to appear on product detail pages. Let’s click on one of the product results for the keyword “Stainless steel men’s watches” and see what happens.

At the bottom of that product page, we further ads for Sponsored Products:

Now, it’s important to know that ads can appear on a product detail page as a result of:

• Keyword targeting
• ASIN targeting
• Category/brand targeting

The auction determines which ad appears in the above list, but it’s important to organize your campaigns so that your competitor-targeting strategies are separated from your other campaigns.

This is because competitor targeting campaigns have a different overall ACoS target due to where the ad fits within your sales funnel.

By having deep discussions with our brands, we understand their risk and tolerance for new customer acquisition. And once we apply their requirements for success at the campaign level, positive results can flow in.

Many technological options for fine-grain tuning

Using the available technology, we were able to tune our client’s campaigns at a highly granular level.

Amazon provides a plethora of tools that automate much of the work. Without these tools, it wouldn’t be possible to achieve the same insight using human eyes alone. Amazon Marketing Stream data is a game-changer. The insights this provides are immensely useful. And this is just one of the many tools we use to execute a brand’s strategy on Amazon.

But despite all the tools available, there is one final element to making it all come together: Fully automating a campaign would not achieve the 30% YoY growth in top-line sales that we achieved for our client. For us to deliver those kinds of results, the human touch was required, and we executed a strategy that left no stone unturned.

Join us tomorrow when we get into the details of this in Part II.