Amazon/iRobot: Does Amazon have an incentive to preference its own products?

We have previously posted about Amazon’s acquisition of the robot vacuum cleaner (“RVC”) brand, iRobot, for $1.7 billion. The case has now been cleared by the UK Competition and Markets Authority (“CMA”) at Phase 1, and referred to Phase 2 for an in-depth investigation by the European Commission.

We therefore have another case of diverging outcomes between London and Brussels. So what’s going on?

The CMA’s Phase 1 clearance decision

I should emphasise at the outset that the CMA’s full Phase 1 decision has not yet been published. We only have a six-page summary, but it gives a good indication of the CMA’s approach. The CMA explored three theories of harm.

First, the CMA assessed a horizontal effects theory of harm whereby the merger would remove Amazon from being a potential entrant in the manufacturing and selling of RVCs. The CMA found that there were six competitors in the market already, and iRobot was only the third largest. The removal of a potential entrant would never usually represent a competition problem on this basis, and that is what the CMA concluded in this case too.

Second, the CMA considered whether the merged parties would be able to foreclose rival smart home platforms, by limiting their access to iRobot in the future. The CMA decided that RVCs are not an important element for smart home platforms.  Also, iRobot is only one RVC product among several, so a lack of access to iRobot would not foreclose rivals.

Third, the CMA considered whether the merged parties would be able to foreclose other providers of RVCs in the UK, for which Amazon’s online store could be an important route to sell their products. This is the primary theory of harm in the case. The CMA followed the usual framework for assessing vertical theories of harm (i.e. ability, incentive, effect). It found that Amazon is an important channel for RVC products and it would be able to foreclose iRobot’s rivals by removing them from Amazon’s online store, manipulating their position in search rankings, or worsening their terms (e.g. by increasing commission rates). However, it found that Amazon would lack a commercial incentive to foreclose iRobot’s rivals – the short-term benefit in terms of increased iRobot sales would be small, and the long-term strategic benefit would also be small. Key to the CMA’s view was that the RVC market was small, was not expected to grow in size, and was “not strategically significant”.

The first two theories of harm are weak, so it was the third theory of harm that is the key to this case, and the incentives analysis is the key part of that theory. However, I am not persuaded by the summary of the CMA’s reasoning. Every product category on Amazon is “small” in its own right, but this does not mean that Amazon does not want to increase its own-brand sales in that category. The CMA is effectively saying that Amazon would never have the incentive to preference its own brands at the retail level when in fact it clearly does this every day. Just try a product search for “batteries” and see whose products appear at the top of the ranking. In fact, even if you search for “Duracell batteries”, you will still see the Amazon Basics product appear above the Duracell product! The battery category is tiny in the context of a company with revenues of half a trillion dollars last year, but its revealed incentive is nevertheless to favour its own brand. The European Commission has even accepted legally-binding commitments from Amazon to curtail its self-preferencing in certain respects, and the CMA has an open antitrust investigation into the same issue.

Further, the CMA finds that Amazon would lose advertising revenues if it foreclosed rival RVC products. This seems doubtful to me. Amazon would not completely delist rival RVC products and therefore ban them from advertising, it would more subtly disadvantage them by making sure that iRobot comes top of the product rankings. The obvious result from such a strategy is that rival RVC companies will need to pay more for advertising to ensure prominence. This is a well-known phenomenon for anyone analysing Amazon’s business and it materially raises rivals’ costs relative to Amazon’s own brands. It leads to consumers paying more for rivals’ products.

The CMA also states that consumers may switch away from Amazon’s platform, thereby defeating the foreclosure strategy. Again, this assumes that rival RVC products are completely delisted. In reality, consumers will not even know that rivals’ costs have been increased and so would not know they should switch away from Amazon. Instead, they will be gently nudged into buying Amazon’s iRobot products through Amazon’s choice architecture strategies and the fact that iRobot now has a material cost advantage. Consumers would have even less need to switch away from Amazon if RVC companies’ products are not in fact cheaper elsewhere, which is something that is often claimed, but the CMA’s analysis does not suggest it considered this issue. The CMA does not mention the fact that more than half of UK households are now effectively tied into Amazon’s online store through their Prime membership and its free delivery and therefore does not consider whether this would limit the extent of consumer switching away from Amazon.

A Phase 2 investigation could have assessed whether competition in the market for the manufacturing and supply of RVCs would risk being reduced, leading to higher prices, lower quality, and less innovation for consumers. It could have more thoroughly considered whether iRobot’s rivals will be disadvantaged post-merger.

The Commission’s decision to go to Phase 2

In contrast to the CMA, the Commission has preliminarily found that Amazon has the incentive to foreclose iRobot’s rivals by preventing them from selling RVCs on Amazon’s online store and/or by degrading their access to it through several strategies. The Commission says that this may include:

  • favouring iRobot’s RVCs in both non-paid (i.e., organic) and paid results (i.e., advertisements);
  • preventing iRobot’s rivals from buying certain advertising services; and/or
  • raising the costs of iRobot’s rivals to advertise and sell their RVCs on Amazon’s marketplace.

The Commission is also worried about two other aspects that the CMA has not investigated, namely:

  • reducing the interoperability with Alexa for rival RVC companies, which they told the Commission is an important issue; and
  • Amazon’s ability to use iRobot’s user data to gain an advantage against rival online marketplaces.

It is strange that the CMA did not investigate the Alexa interoperability issue, which seems a very plausible theory of harm to me. One assumes that RVC companies were making the same submissions to the CMA that they made to the Commission. The CMA only investigated interoperability in another direction, i.e. smart home platforms’ access to iRobot, which seems a much less plausible theory of harm as it is Alexa rather than iRobot that enjoys gatekeeper power.

I suspect the data theory will be found to be too speculative in the end, but again I agree with the Commission that it needs an in-depth analysis that can only be done at Phase 2. The Commission is right to recognise that a cutting-edge tech firm like Amazon views its business as an ecosystem whereby its separate activities are linked to each other and reinforce each other’s strengths.

The fact that the Commission has investigated theories of harm that the CMA has not suggests that the two agencies are not collaborating effectively at the moment.


So much of the commentary recently has been about the CMA having appointed itself at the world’s merger control policeman by blocking foreign-to-foreign (normally US-to-US) mergers such as Microsoft/Activision, Meta/Giphy and Illumina/PacBio. The Amazon/iRobot case paints a different picture as it is the Commission who is acting more aggressively towards the deal (and note that the Commission also issued a huge fine to Illumina this week). It also suggests that it is the Commission who is more interested in analysing Amazon’s flywheel business model, despite it being the CMA which updated its merger assessment guidelines to focus on such ecosystem theories of harm.

In Microsoft/Activision, the two agencies eventually agreed on the competition issues arising from the merger. They disagreed on whether a complex behavioural remedy was an effective solution to those issues. We can argue about who was right, but it is at least a valid and logical policy difference, based in part on the differing legal frameworks. The Amazon/iRobot case is different. In this case, the two authorities disagree on whether there is a plausible vertical theory of harm that rests on whether Amazon has the incentive to disadvantage rival companies’ products. This is not a policy difference, but a difference in the substantive assessment. They should have been applying an identical analytical framework to an identical set of facts.

I have heard conspiracy theories that the CMA is reacting to the political heat after blocking Microsoft/Activision by performatively clearing Amazon’s deal at Phase 1. After all, the CMA didn’t need to call in the merger at all if it didn’t believe the partial foreclosure story, after spending the best part of a year in pre-notification. More likely, they simply decided that the small RVC market was not important enough for an expensive Phase 2 investigation. Unfortunately, we are left with a superficial analysis of Amazon’s incentives as a result, which seems to ignore the CMA’s previous stance towards digital ecosystems and which will be unhelpful when the Digital Markets Unit considers Amazon’s designation under the new Digital Markets, Competition and Consumers Bill. We also have an endorsement of Amazon’s approach to acquisitions – unlike Microsoft, Amazon continues to buy smaller companies and reduces the merger control scrutiny it receives as a result.

Amazon’s RVC rivals will be hoping that the Commission blocks the deal at the end of Phase 2 in November this year.

One thought on “Amazon/iRobot: Does Amazon have an incentive to preference its own products?

  1. CMA analysis is reminiscent of its decision in Google/Beat that Quote, where it found little incentive to self promote, at variance with the position of the Commission in Google Shopping. Maybe the underlying point is technological – self preference is a build decision. The CMA analysis is always all about economic consequences and incentives. In reality it may be just easier for platforms to build in self preference and get on with making money?

Leave a Reply