Skip to content
Blog

How many hotel API providers do you need? Count the mapping

Tekravel Editorial
Written with the help of AI

The hotel search went live last quarter on one supplier, and for a while nobody complained. Then sales forwards a screenshot. A client was planning Istanbul, found a four-star hotel on their phone, and asked why your site did not show it at all. Not a worse price. No listing. The next message is addressed to you: should we add another provider?

Probably. But how many hotel API providers do you need is not a coverage question with a coverage answer. Every supplier you add widens the inventory and also sends you hotels you already have, under a different name, a different ID and slightly different coordinates. The second integration is mostly code. Everything after it is mostly deduplication, and that is the cost people leave out when they count suppliers.

Each supplier is two feeds, not one

A hotel supplier hands you two very different things, and treating them as one is the root of most early mistakes.

The first is static content: property names, addresses, coordinates, star rating, photos, room descriptions, amenities. It changes slowly. You pull it in bulk, store it, and refresh it on a schedule measured in days. The second is live rates: what is available for these dates, for this occupancy, at what price, under which cancellation policy. That changes between the search and the booking. You cache it for minutes at most, and you re-check it at prebook no matter what the search said.

So two cache lifetimes, two failure modes. Stale static content shows the wrong photo. A stale rate sells a room at a price the supplier will not honour. When you add a provider, you add both feeds, and the static one is where the real work hides, because it is the one you have to reconcile against everything you already hold.

Why the same hotel arrives three times

Connect three suppliers in the same city and search. You will see something like this: one supplier calls it Grand Plaza Hotel Taksim, another Grand Plaza Taksim, a third GRAND PLAZA HTL - ISTANBUL. Three internal codes. Three sets of coordinates a short walk apart. Three photo sets, one of them from before the renovation. To your traveller, it looks like a broken site.

The fix is a mapping layer, and the field you dedupe on is not the name. It is your own canonical property ID. Each supplier's hotel code points to it through a mapping table: supplier, supplier hotel code, canonical ID. The supplier code never becomes the primary key of your catalogue, because the moment you add a second supplier that decision becomes very expensive to undo.

How you decide that two codes are one hotel is a matching problem, and it is rarely clean:

  • A shared external mapping code, when suppliers carry one, is the strongest signal. Not all do, and not for every property.
  • Coordinates within a small radius, plus a normalised name (strip "hotel", the city, punctuation, case) catch most of the rest.
  • Address and phone number settle the close calls. Two branches of the same chain on one street are exactly the case where name and distance agree and the hotels are different.
  • Whatever is left goes to a human review queue. It does not go live on a guess.

Rooms are harder than hotels

Once the property is mapped, the rooms still disagree. "Deluxe King", "Superior Room, 1 King Bed" and "Double Room - City View" may be the same room or three different ones. Most teams compare room offers at the property level, by board type, cancellation terms and price, rather than trying to map every room name across suppliers. That is a defensible choice. Just make it on purpose, and show the traveller enough of the supplier's own room description to know what they are booking.

One supplier, several, or an aggregator

Three routes, and the honest comparison is about who carries the mapping and the parity risk, not about headline coverage.

 CoverageRate parity riskIntegration effortWho handles mapping
One supplier, directStrong where that supplier is strong, thin elsewhereHigh: when its price is off, you have no second quoteOne integration, both feedsNobody needs to; one code per hotel
Several suppliers, directWidest you can controlLower: you can pick the better quote per stayOne per supplier, plus the mapping layer, plus its upkeepYou, permanently
An aggregatorWide, decided by the aggregator's sourcesMixed: you see the price it chooses to pass onOne integrationThe aggregator, with your judgment on what slips through

You can disagree with the middle row, and plenty of teams do: owning the mapping is also owning the quality of your results, and some would never hand that over. The point is that it is a staffing decision as much as a technical one. The mapping table is not a project with an end date.

Deciding the number

Count markets, not suppliers. Start with one provider that is strong where you sell most. Keep a log of the searches that came back empty or thin, and of the screenshots sales forwards. Add the second provider when you can name the cities and properties the first one is missing, and check that the candidate actually covers them before you sign anything.

Each addition should close a gap you can list. If you cannot list it, the new supplier mostly sends you duplicates of hotels you already sell. We wrote separately about how to judge travel APIs on their axes rather than their brochures, and that framework applies to every candidate here.

Not integrating everything yourself

There is a fourth route, which is to integrate once against a platform that already holds supplier connections. On this platform supplier content is shared through supplier groups rather than one-to-one deals, and a tenant sees the suppliers its groups grant it. For a developer, that moves the question "which suppliers stand behind my hotel search" from code to configuration: you build against one hotel API, and the supplier mix is a commercial arrangement, not another integration sprint.

It does not remove the questions in this article. Ask how mapping is handled for the suppliers you would be granted, and test it in the cities you sell, not the demo city. Access to the flight and hotel APIs starts as a conversation rather than a self-serve key, which is the right moment to ask exactly that.

What under-integrating costs you

The missing hotel in the screenshot is not one lost booking. It is a client who now checks their phone before they message you, and after a few more of those, instead of messaging you. The agency did nothing wrong on service. Its search simply did not know the property existed.

Over-integrating costs differently: three rows of the same hotel in your results, three cancellation policies to explain at the counter, three contracts and three sets of credentials to rotate. Neither failure shows up in a supplier count. Both show up in what a traveller sees on the results page.

Build the live-rate side so it survives slow suppliers; we covered that in designing a hotel availability client. Then let the gaps you can name decide how many providers sit behind it.

Tekravel Editorial

Travel technology desk

The Tekravel travel-technology desk writes for the trade: agency owners, consolidators and the developers who integrate them. Every article is checked against the platform it describes before it is published.