The study shows that multi-sided markets pose difficulties when the relevant product market is to be defined. These difficulties pertain to two questions. The first question is whether one or several markets should be defined. In this regard, it is not easy to extract a coherent method from the cases examined. Instead, the methods applied give the impression of ad hoc-solutions, where similar circumstances result in dissimilar outcomes. Indeed, it is hard to reconcile the different market definitions in Visa International MIF and MasterCard MIF. The uncertainties are not limited to these two cases, as the methods applied in Google Shopping too give rise to ambiguities. Why was the market for general search engine platforms separated into two distinct product markets, but the market for comparison shopping services encompassed both sides? Unfortunately, this discussion was not present in the decision. This only serves to reinforce the impression that these questions are solved on an ad hoc-basis. A clear method of approaching multi-sided markets is desirable, not least because the enforcement of competition rules must be characterized by consistency and foreseeability. Hopefully, the judgement in Google Shopping will bring further clarity to this. Nonetheless, the conclusion is that one market should be defined when differences between competitive constraints on the two sides are absent. E contrario, this means that two markets should be defined when such differences are present. This is true regardless of the market in question being a transaction or a non-transaction market. This method seems preferable to strictly adhering to the division of multi-sided markets into transaction or non-transaction markets. If the Commission’s analysis is correct in that there are no differences in competitive constraints on the two sides of comparison shopping services, in combination with crossing network effects, the platform’s multi-sidedness is a necessary trait for both sides. This means that a substitute has to be multi-sided in order for it to be included on the relevant product market, which minimizes the risk for false negatives. Vice versa, the definition of two markets allows for one-sided products to be included on the relevant product market, which minimizes the risk for false positives. This is important not only for the binary finding of dominance or non-dominance, but also the degree of dominance. As concluded above, incorrectly defining one market may artificially inflate the degree of dominance into false super-dominance, and incorrectly defining several markets may artificially dilute the degree of dominance. The second question is how substitutability should be measured. It is obvious from the cases examined that qualitative measures are used and not quantitative measures. The products’ characteristics, intended use, purpose, functionalities, users’ perceptions of the product, etc. were given much attention. The SSNIP test was not applied in any of the cases. The first conclusion to be drawn from the examination above is therefore that the difficulties regarding measuring substitutability on multi-sided markets mainly concern quantitative measures. The arguments against applying a SSNIP test related to the cellophane fallacy (in two different forms, one of which was deceivingly similar to the reverse cellophane fallacy) and differences in price sensitivities between the two sides. Network effects present an additional difficulty, which may lead to exaggerated results when measuring substitutability. The second conclusion to be drawn is that there exists a reluctance to apply a SSNIP test in a way that is tailored for multi-sided markets. One method that has been proposed is to apply the test on the total sum paid by both sides, while allowing the intermediary to adjust the increase in price in accordance with its price structure. The categorical dismissal of applying the test in this way suggests that adapted versions have some time to wait before being introduced into case law and decisional practice. If they, as their proponents argue, are a robust way of broadening the evidence of possible substitutability, this is unfortunate. The risk of defining the market overly narrow or overly broad is of course present in this regard as well. A broader spectrum of evidence therefore minimizes the risk of incorrectly finding both dominance and non-dominance.