Algorithms are powerful new agents of double jeopardy
The marketing playing field has never been level. Effective marketing is easier for big companies, it’s a well-established fact. The theory of why – double jeopardy – comes from respected academics including Byron Sharp and Jenni Romaniuk of the Ehrenberg-Bass Institute, and the evidence is also indisputable.
But what isn’t well known is how the tilt of the pitch has changed as technology has transformed the way consumers and marketers do business.
The pendulum had two swings. Starting in the 1990s, the Internet made it easier for start-ups to make their products available, to market, and to grow. But since then, the direction of travel has increasingly been in the opposite direction, with an ever-increasing advantage for the big players.
The reason for this is that the algorithms are trading more and more buys and they are prioritizing things that are already popular. This means companies that are already big. Small businesses are served less often, pay more for their digital ads, and have to spend more money defending their existing sales.
Both swings result in a situation where it’s easier than ever for small businesses to get started, but also where there’s a limit to the scope of their growth. Where challengers cannot challenge as much, and large companies remain at least partly isolated from the competition.
It’s better to be big
The graph below plots advertising ROI in 343 UK econometric studies between 2015 and 2019 by size of advertiser’s business. It’s clear, the return on investment is higher if you are a large company.
The explanation is that the products sold by large companies are more readily available, to more customers, than the products sold by small companies.
If Tesco puts an advert on TV it is much easier for someone who lives in Birmingham to respond than if the advert was for a corner store in London.
The table above is an overview – in many different categories. But more detailed data shows that it’s better to be big too. The chart below shows the double jeopardy pattern – something that has been seen across a large number of categories around the world.
It is almost always observed that larger companies – illustrated here by larger bubbles – have both more buyers (higher household penetration) and more loyal buyers (higher purchase frequency).
The explanation again relates to availability. Customers split their purchases more or less evenly between the options they know about and can get. And the products that big companies sell are more mentally and physically available.
Algorithms as new arbiters of availability
When e-commerce first became a realistic possibility in the 1990s, being available no longer necessarily meant an expensive physical store. Small businesses could create a website and customers could find them, alongside larger competitors.
Economists have rightly said that it would be good for consumers, who would have a lot more choice, and good for the economy through healthier competition as well.
Then, over time, consumers increasingly turned to algorithms to help them navigate all the new options. The calculations in the Google, Amazon and Facebook machines increasingly decided what would be made available, especially for the 70% of people who never look at the second search page.
And all the important algorithms prioritize things that are already popular. This means companies that are already big.
Google’s organic search algorithm is the first important case. It moves businesses to the top of the page if they have low “long click times”. That is, if they are clicked often and users linger when they land. The first part of this criterion – which is often clicked on – favors large companies, so it’s much more often large companies at the top of the page.
Unfettered competition between small and large businesses keeps the economy efficient, ensures progress, and allows consumers to get the best possible deal.
And the big business advantage widened as Google, Facebook and others monetized their search pages. Because they again used popularity-driven algorithms to determine how companies would pay for positions on the page.
Paid search auctions offer ads at a lower cost per click if the algorithm expects an ad to be clicked on a lot. And larger companies have this higher click-through rate, and, to use Google’s setup as an example, that means a higher “quality score”, and in turn, the ability to be at the top of the page for a lower price.
Many large companies are using this advantage to attack the increased availability to smaller competitors that economists were counting on by bidding on the brand names of smaller competitors.
Again, there is an inequality. A large company bidding on the brand name of a smaller competitor will get a better “quality score” than a small company bidding on the name of a large competitor. This means smaller brands are disproportionately under attack and must disproportionately defend their availability by paying for ads against their own name.
Merchant media is the next big frontier, as this same configuration increasingly plays out in retailers’ results algorithms. Amazon is a particularly big case because during the pandemic it became the largest single search engine for products with 53% in 2020, according to data from eMarketer.
The top search on Amazon is increasingly available for purchase, and advertisers have proven to be very willing to pay. The graph above shows, in orange, Amazon’s revenue from paid search advertising. Last year, that figure was more than YouTube’s total ad revenue.
Meritocracy would have been better for all of us
It is unlikely that the engineers of Google, Amazon and Facebook intended to jeopardize the functioning of the algorithms. Undoubtedly, they were much more focused on the challenge of finding a way to deliver millions of good recommendations with minimal computing power and latency.
Crowd-sourcing solved that problem and produced a nifty way to prioritize what made sense. If a lot of people are already buying it, that must be good, right?
But that wasn’t the only possible solution, and in fact, Amazon’s organic algorithm is a good example of that. In the absence of any paid influence, it is both meritocratic and populist. It examines a match between the entered text and the product, price, reviews and out of stock products.
More meritocratic algorithms in e-commerce would have been better for smaller companies with a good product. And it would have been better for the global economy. Unfettered competition between small and large businesses keeps the economy efficient, ensures progress, and allows consumers to get the best possible deal.
Something that would have been better for all of us.