It’s hard to imagine Amazon as vulnerable. But after Covid, demand dropped while the costs of expensive fulfilment capacity set up during the crazy times continued. Almost a trillion dollars’ worth of turnover netted out a loss of $2bn in 2022.

With this peril in the background, the advertising business, where Amazon sells access to the top slot in the search page and other real estate on its sites, has been a lifeline. Selling ads contributed $40bn to turnover in 2022 – more than Amazon Prime – and there’s no sign of the growth slowing down.

For marketers, this has meant a whole set of new tools arriving on the desk – not literally in a cardboard box with a smiley arrow on it, but nevertheless, we’ve unpacked and had a go.

We don’t yet have old hands who’ve grown wise from experience using Amazon Ads, or meta studies that encompass a lot of trials. But econometrics companies that have evaluated the new tools are starting to see patterns.

It turns out that a lot depends on which format you use, your category and life stage, and the type of measurement used for decision making.

Generic and competitor search come out top

In evaluations carried out by my team at Magic Numbers, search ads aimed at generic and competitor keywords undoubtedly work to drive sales on Amazon. Returns per £1 spent are comparable to generic search on Google, ranging from £1 to £4 depending on category, offer, and how new the news is.

On display, we’ve seen some good effects, some not so good, and some cases where there was no discernible effect at all. The best results were for “rewarded” display where shoppers got a $5 coupon if they watched the ad to the end. The worst were where ads were placed “off Amazon” and at very low weekly weights.

We don’t know if our cases are representative, but if they are, they point to some additional watch outs. We’ve never seen brand search do more than help people navigate the now-crowded Amazon search results page. And, even despite Amazon’s stat that 90% of the effect of their ads is felt in other sales channels, we have never found an effect of Amazon Ads on sales anywhere else.

If there was a marketing effectiveness ‘crisis’, it’s now over

Category and brand strength matter for the choice of Amazon vs Google

The chart below shows data on the balance of searching across Amazon and Google for a selection of categories. Amazon search ads are more important in categories where people search Amazon more than Google. For example, where the product is bulky (home and garden), or repeatable (pet supplies).

At the other end of the scale, Amazon is less important in categories where expert reviews are helpful (consumer electrics) and where the purchase really matters to the shopper so that they aren’t necessarily going for the cheapest price (beauty, health care).

It’s also worth considering how strong your brand already is. The vast majority of branded searches take place on Google because people who are committed to buying from a particular brand head to that brand’s website via Google to buy direct.

This suggests that if you’re small and unknown, starting out on Amazon is one route to gaining enough brand awareness to later win on Google. But, on the other hand, if you’re already big, and people already search directly for you, there may be less to gain from paying to appear among the jostle of suppliers that Amazon’s algorithm serves up.

Good measurement matters

You’ll never get these kinds of insights using the last-click attribution that comes packaged with Amazon Ads. As Amazon itself would admit, it’s important to instead use measurement that can properly identify whether a sale was caused by the Amazon ad or something else earlier in the purchase journey.

The issue is that last-click wrongly attributes credit for everything that’s happened in the journey to purchase to the event that happened last, just before a customer presses “buy”.

Because Amazon very often appears at this last moment, at the end of people’s purchase journey, when they’re already intending to buy something, last-click hugely over allocates credit to Amazon. In our cases, the overclaim for ROAS is in Amazon’s last-click is dramatic and very misleading – estimates are 300-400% out.

So, it’s dangerous to test, learn and invest into Amazon Ads without independent econometrics, switch off tests, or regional hold outs.

Increased pressure is driving marketers to focus more on performance

Saving Amazon?

To emerge from their current vulnerable moment, Amazon’s senior management, along with financial analysts, are pinning part of their hopes on the advertising business.

And the advertising sales teams are certainly doing their bit. They have sponsored a number of marketing conferences, and they’re directly contacting marketing budget-holders everywhere.

But it is marketers who will decide whether the strategy ends up working.

Through experimentation and evaluation, we’ll find the right place for Amazon Ads in our media budgets. And maybe, if we’ve collectively learnt enough about the dangers of last-click, we’ll get to maturity quicker than we did with Google and Facebook.

In the meantime, Amazon would be wise to keep an eye on the core business. Google’s genius has always been in selling ads while never undermining their value to users. For Amazon, that’ll be much harder to pull off. Its core proposition is pointing shoppers to the best value item among all the options, something that is at odds with an additional proposition to point shoppers to whatever advertisers pay to support.

Just like their rainforest namesake, Amazon currently needs saving. It remains to be seen if advertising really can come to the rescue.

Grace Kite is the founder and managing director of Magic Numbers, which carries out award-winning econometrics projects to evaluate marketing, including Amazon Ads, for clients.