On September 19, 2024, Reuters reported a European court ruling on Booking.com's price restrictions. The company disagreed, maintaining that its historical clauses were necessary and proportionate.
The Court rejected the ancillary-restraint argument, without finally resolving the national dispute. Narrow parity covers a hotel's own channels; wide parity also covers rival platforms.
The business question is how a condition connecting prices changes the seller's choices. A manager might be able to enter a number in every price field, yet remain unable to alter those numbers independently. To examine that distinction, consider an invented hotel with three sales channels. The following offers and rules are a self-contained illustration, not evidence of a particular hotel's contracts or a reconstruction of Booking.com's operations.
Begin with one genuinely comparable offer
Our fictional hotel sells the same room type for the same night through platform A, platform B and its direct website D. Every quote covers two adults, includes the same services and has the same cancellation deadline. Payment timing and the treatment of taxes are also identical. The letters identify channels, not different rooms. Prices are expressed in invented money units, with no currency conversion.
This definition prevents a false comparison before any calculation begins. Suppose the direct quote included breakfast while the platform quote did not. A higher direct price could then reflect a different package, rather than the relationship between two prices for the same offer. The present example excludes that explanation. It changes only the price fields and the conditions connecting them, leaving the promised stay intact.
The comparison also takes place at one moment. A quote observed on Monday and another observed on Friday would not automatically form the pair we need, even if both covered the same stay. Here all three numbers belong to one simultaneous offer. We are not following a changing price over time, comparing completed transactions or estimating what the hotel would charge a different customer.
Write the three numbers in the fixed order A, B, D. For example, the list 100, 90, 85 means that the same stay is offered at one hundred through A, ninety through B and eighty-five directly. The order matters: rearranging the numbers changes which channel is cheapest. A report containing only the lowest and highest quote would lose information essential to the question.
A condition can restrict a relationship without setting a price
First stipulate: the direct price cannot be below A's price. In symbols, D must be at least A. The list 100, 90, 85 fails this test because eighty-five is below one hundred. The list 100, 90, 100 passes. Platform B remains at ninety in both lists, and this particular rule says nothing about its relationship with A.
The second list shows why the direction of a condition matters. A is not necessarily the cheapest place to buy the room: B is cheaper. Nor must the direct and A prices be equal. The list 100, 90, 110 would pass too. Our rule places a floor under the direct price relative to A; it does not impose a ceiling on the direct price or a single shared price across all channels.
Now hold A at one hundred and ask whether the hotel can reduce its direct quote from one hundred to ninety-five. Under the stipulated rule it cannot make that change alone. But a different combined change, moving A and D to ninety-five together, passes. The obstacle is therefore not the number ninety-five itself. It is that number appearing in the direct field while A remains at one hundred.
This is a useful distinction between choosing a value and choosing a combination. Each individual number could appear somewhere in an acceptable list. That does not mean every arrangement of individually possible numbers is acceptable. In our example, the manager needs to examine the whole proposed list, because a price that works in one arrangement fails in another.
Adding a second relationship changes a different decision
Instead require: both B and D must be at least A. The list 100, 90, 100 now fails because B is too low, even though the direct price passes its own comparison. The list 90, 100, 90 succeeds. Notice that this change does not merely tighten the direct condition. It adds a comparison involving a different channel.
Consider a proposed reduction on B while A remains at one hundred. Under the first rule, B could move from one hundred to ninety because B was outside the comparison. Under the new rules, that same proposed change fails. To include ninety at B, the hotel must also bring A to ninety or below. What was previously an independent adjustment now requires attention to another price.
The reverse adjustment is different. Starting with A, B and D all at one hundred, reducing only A to ninety satisfies both comparisons. B and D are still at least as high as A. Our two rules therefore do not freeze prices, and they do not obstruct every reduction. They distinguish which channel changes and whether the other entries remain consistent with that change.
No selling cost has entered this argument. Even if the hotel were told that B had become cheaper to operate, the two requirements would still reject the list 100, 90, 100. A change in a channel's operating cost and permission to express a different quoted price are separate objects in this example. The cost information alone cannot alter the inequalities we stipulated.
Reciprocal conditions imply less than complete uniformity
Next add a condition in the other direction: A must be at least B, and D must also be at least B. Taken together with the preceding requirements, A cannot be below B and B cannot be below A. The only way for both comparisons to hold is for A and B to be equal. Neither comparison, however, supplies their numerical level.
If A is ninety, B must be ninety. If A is one hundred, B must be one hundred. The direct price must be at least that common platform price, but can be higher. Thus 90, 90, 100 passes, as does 90, 90, 90. The list 100, 90, 100 fails. The implication is equality between the two platform entries, not compulsory equality among all three entries.
The equality follows from the fictional reciprocal conditions alone. It does not require the platforms to have identical costs, the same number of visitors or equally attractive services. None of those characteristics was included in the assumptions. Conversely, the calculation supplies no evidence that actual platforms have coordinated their conduct. It is a logical consequence of the specified restrictions, not a diagnosis of a real market.
A fourth observation helps separate a price level from a price relationship. Reducing all three entries from one hundred to ninety leaves every comparison satisfied. Increasing all three together also leaves them satisfied. The conditions constrain relative positions while permitting many absolute levels. A forecast that this system necessarily produces a particular room price would therefore require information the system does not contain.
The three tests in compact form
- The list 100, 90, 100 passes our first test but fails the second.
- The list 90, 100, 90 passes our second test but fails the reciprocal test.
- The list 90, 90, 100 passes all three tests; the direct quote remains higher.
Removing one condition need not release every price
Start from the reciprocal arrangement and remove the requirements imposed in A's direction. Retain only the conditions that A and D must be at least B. It is now possible to post 100, 90, 100. This list was previously excluded because the relationship requiring B to be at least A no longer applies. The change has created a new possible ordering of the platform prices.
Yet the hotel still cannot post 100, 90, 85 under the remaining conditions: the direct quote is below B. It would be inaccurate to describe the first removal as making all three prices independent. One set of comparisons disappeared, but another survives. The actual increase in choice is determined by what remains, not solely by naming the requirement that was removed.
There is an instructive intermediate step. Suppose we remove only B's requirement concerning the direct quote, while retaining both comparisons between A and B and A's requirement concerning D. Nothing new becomes possible. A still equals B, and D must still be at least A, so it must also be at least B. The removed direct comparison was already implied by the remaining ones.
This does not make every removal unimportant. It shows why counting deleted conditions is a poor substitute for testing the resulting combinations. In this constructed case, one deletion changes no possible price list, whereas removing a different relationship permits a previously excluded list. The distinction can be demonstrated with the remaining rules themselves; no assumption about consumer response is needed.
We can isolate that contrast using another single deletion. Return to all four reciprocal requirements and remove only A being at least B. Retain B being at least A and both direct-price comparisons. The list 90, 100, 100 now passes: B exceeds A, and D is at least each platform price. Previously it failed precisely the deleted requirement. Starting with one hundred everywhere, only A needs to change; B and D can remain unchanged. Both experiments remove one comparison, but only this experiment adds that list. This explicit witness explains the difference without attributing greater importance to a requirement merely because of its name.
A new possibility does not determine a new outcome
Imagine that all restrictions in the example disappear. The hotel could retain 100, 100, 100, choose 100, 90, 85 or select many other combinations. The first list did not become impossible merely because the second became possible. Removing a constraint enlarges the choice under these assumptions, but the model has not specified a rule for selecting one member of that larger set.
For that reason, it cannot tell us how many customers would move between channels. We have no customer preferences, booking costs, search effort or demand schedule. Two different responses are consistent with the same removal: the hotel might keep every quote unchanged, or it might change one or several. Both remain logically possible until additional evidence or explicit behavioural assumptions distinguish them.
Even observing a lower direct quote would not, by itself, fill all those gaps. Suppose the unrestricted hotel posts 100, 100, 90. A customer able to use all three channels faces a cheaper direct offer in this fictional list. We have still not established that every customer sees it, values each purchasing process equally or completes a reservation. A quoted price and a completed purchase are different observations.
This is not an argument that prices are irrelevant. It is a boundary on the conclusion supported by our example. The comparisons establish whether a proposal fits the stipulated conditions. They do not calculate the hotel's preferred proposal, the volume of reservations, the platform's continued viability or the overall benefit to travellers. Each further conclusion needs its own stated premises.
What a snapshot can and cannot identify
Now reverse the direction of enquiry. Instead of knowing the rules and checking the prices, suppose an observer sees only 90, 90, 90. That snapshot is compatible with every arrangement examined above, including no restrictions at all. It cannot identify which arrangement produced it. The ambiguity is not a problem of imprecise measurement: even exact observation of all three prices leaves several explanations open.
A different snapshot can rule out a particular set of assumptions without proving another one. The list 100, 90, 100 conflicts with our wide condition in A's direction but satisfies the first, direct-only rule. It also satisfies a completely unrestricted system. Therefore the observer could reject the former hypothetical description while still being unable to choose between the latter two.
The qualification concerning the same offer remains decisive. If the numbers describe different cancellation terms, our three-variable test no longer applies as written. Nor can a price collected at another moment automatically be inserted into the missing field. A conclusion about the stipulated system requires the observations that system defines; a superficially similar set of numbers is not enough.
An April 2024 paper studying France's 2015 change found significant price reductions for direct hotel contact, not online channels. Nonsignificant online estimates do not establish zero effects.
Keep the business claim at the level the evidence supports
The fictional hotel's problem was not that its prices had to remain high, low or permanently unchanged. The problem was that some changes could not be made independently. The examples showed exactly which comparisons obstructed each proposal, which reciprocal requirements forced platform equality and why a redundant deletion could leave every available choice intact.
That is more precise than treating every condition as a single fixed price. It also avoids the opposite mistake of assuming that several editable fields necessarily represent several independent decisions. The meaningful object is the complete proposed offer across the channels we defined. Its permissible relationships, the seller's eventual choice and the traveller's purchase remain distinct questions, even when they concern the same room.