How to Audit Your Hotel Mapping Accuracy: A Step-by-Step Guide

Mapping problems do not announce themselves. They hide in your inventory until something breaks in front of a customer.

A guest arrives to a different room from the one they booked. A parity tool flags a gap that is really two listings for one hotel. A rebooking engine finds a cheaper rate it cannot use, because it cannot prove the rooms match. When you find out, the damage is done.

An audit finds the problem first.

This guide is for OTA data analysts, engineers and product managers who own inventory quality. You need no new tool and no spare sprint. You need a representative sample, a SQL client and the method below.

The method covers three dimensions: property deduplication, room-type completeness and rate plan metadata coverage. Each produces one number. Each number has a threshold. Each threshold has an action.

What does a hotel mapping accuracy audit measure?

Mapping quality is not one thing. A system can match properties well and still describe rooms badly, or carry rates nobody can safely sell. So the audit scores three layers separately.

Duplication is the one most teams underestimate. Vervotech’s internal research found that a single hotel is listed an average of 9 times on a typical OTA platform. Every extra record is noise. It splits reviews, confuses ranking and puts near-identical cards in front of a guest. If you have never measured your duplication rate, assume it is higher than you think.

For background on why this happens, read why most OTAs still get hotel mapping wrong.

Step 1: How big should your audit sample be?

Pull 500 to 1,000 active properties. Five hundred is the floor for a result you can trust. One thousand lets you segment by supplier and market without each segment becoming too small to read.

Do not sample only your best hotels

Your top sellers have usually been reviewed by hand at some point, so they flatter the result. Mapping problems concentrate in tail inventory: newer suppliers, regional markets and secondary sources that have never had a manual check.

Build the sample so that at least 30% of it is tail inventory. A practical split for a 1,000 property sample:

• 500 properties from your top revenue bucket.
• 200 properties from mid-volume inventory.
• 300 properties from the tail, weighted toward suppliers onboarded in the last 12 months.

What to pull for each record

For every property, pull every record from every supplier feed. Do not deduplicate first. The duplicates are what you are measuring. For each record, capture:

1. Property name, exactly as each supplier sends it.
2. Supplier-assigned property ID.
3. Physical address, including postcode.
4. Latitude and longitude.
5. Phone number.

Keep the raw values. You will normalize copies next and want the originals for edge cases.

Step 2: How do you measure your deduplication rate?

Use three signals in a fixed order: address, then coordinates, then name. Each is weaker than the one before it.

Primary signal: normalized address

Normalize every address before you compare anything. Expand abbreviations, so “St.” becomes “Street” and “Blvd” becomes “Boulevard”. Standardize postcode format, including spacing and case. Trim whitespace and strip punctuation. Convert to one case.

Two records with the same normalized address and the same postcode are almost certainly the same physical property. Treat these as matches.

Secondary signal: coordinates within 50 meters

Some records will have missing, partial or inconsistent addresses. For those, calculate the distance between coordinates. Two records within 50 meters of each other are strong candidates for the same property.

Flag them for review. Do not merge them automatically. Hotels share buildings and sit next door to sister properties. Fifty meters catches true duplicates, and it also catches neighbors.

Tertiary signal: name similarity

Where address and coordinates are both unclear, score name similarity. Use it to send a record to review, never to confirm a match. “Grand Hotel” exists in hundreds of cities, and one chain can run three near-identical names in one district.

Calculate the rate

Count the records that refer to a property already represented by another record in the sample. Divide by total records in the sample. That is your duplication rate.

A worked example: 1,000 properties arrive as 1,180 supplier records, and review confirms 140 of those records duplicate a property already in the sample. Your duplication rate is 140 divided by 1,180, or 11.9%.

A duplication rate above 3% is common in multi-supplier OTA environments. It is also the point at which you should fix duplicates before you add another supplier.

If your rate is above 5%, sort the duplicates by supplier pair. Start remediation with the overlaps between your highest-volume suppliers. Those pairs produce the most guest-facing errors, because they serve the most searches. The cost of leaving them is covered in the hidden cost of duplicate hotel listings.

Step 3: How do you score room-type completeness?

A property can be mapped perfectly and still sell badly if its rooms are described in free text. Filters, recommendation engines and AI booking agents read fields. They do not read paragraphs.

For each property in the sample, pull every active room record. Check whether each of these five attributes exists as a structured field, not buried in a description:

• Room type category: Standard, Superior, Deluxe, Suite.
• Bed configuration: Single, Twin, Double, Queen, King.
• View type: City, Garden, Sea, Pool, None.
• Smoking policy: Smoking or Non-smoking, as a binary value.
• Maximum adult occupancy: a number.

Score each room

Divide the number of structured attributes present by the number that should be present for that room type. A room with no view still needs the field, set to “None”, so the denominator for most rooms is five.

A room scoring below 60% is effectively invisible to a structured query. A guest filtering for “sea view, king bed” will never see it, even if it is exactly what they want.

Segment by supplier

Calculate the average score, then split it by supplier feed. The split usually shows most of the gap sits in two or three feeds. Normalize those first.

Structured room data matters more as booking moves to machines. See our guide to preparing your hotel inventory for AI-driven booking and our piece on room type taxonomy for OTAs.

Step 4: How do you check rate plan metadata coverage?

Rates are where mapping quality turns into money. A rate with missing metadata cannot be safely offered in an automated flow or compared for rebooking, and it invites false parity alerts.

Pull every active rate plan for the properties in your sample. Check three fields on each:

1. Channel eligibility tag. Which distribution channels the rate is valid for: B2C, B2B, mobile, member or package.
2. Refundability flag. Refundable, non-refundable or conditional, as an explicit value.
3. Validity window. The date range in which the rate is bookable and applies.

Divide the number of rate plans with all three fields populated by the total number of active rate plans. That is your metadata coverage rate.

Treat every failure as a data gap, not a rate exception. A rate with no refundability flag is unknown, and unknown rates should not reach live inventory.

A rate plan missing one field is not a small problem. It is a rate you cannot safely sell, compare or replace.

Step 5: What do the results mean, and what should you do next?

You now have three numbers. Here is how to read them and what to do about each.

Why the order of fixes matters

Fix duplication before you add suppliers. A new feed added to a system full of duplicates does not add coverage. It multiplies the duplicates. Fix room completeness before you invest in AI search, which depends on structured fields. Fix rate metadata before you automate rebooking or parity checks, which fail quietly on incomplete inputs.

When to fix it yourself and when to bring in help

Small gaps in a few feeds can be fixed internally. If duplication is above 5% across many supplier pairs, or room completeness is low across most feeds, manual work will not keep pace with supplier changes. That is the point where a dedicated hotel mapping and room mapping layer pays for itself. Vervotech maps 3M+ properties across 400+ suppliers at 99.998% accuracy, and the hotel mapping API can be run against your audit sample so you can compare results line by line.

How often should you run a mapping audit?

Every quarter.

Supplier feeds change without notice. Hotels rename room types after renovations. New suppliers arrive with their own IDs and naming habits. A mapping that was correct in January can drift by April.

A one-time audit gives you a snapshot. A quarterly audit gives you a trend line that shows whether your fixes are holding. Keep the method fixed between rounds: same sample size, tail share, match logic and thresholds.

Three habits make the quarterly cycle cheap to run:

• Save the audit queries as scripts, so the next round takes hours, not days.
• Keep a fixed core of 300 properties in every sample, and rotate the rest.
• Run an extra audit within 30 days of onboarding any new supplier.

If you are planning for growth, our article on what happens to your mapping accuracy when you scale up covers what changes as supplier count rises.

What will the audit tell you?

Most data teams start with a feeling that the data “could be better”. An audit replaces that feeling with three numbers: what is wrong, where it sits and how large it is. It needs no new tooling, only a sample, fixed match rules and the discipline to repeat it.

Run it once and the result will almost certainly show a bigger gap than you expected. That is not bad news. It is the first accurate picture you have had of your inventory, and the first one you can act on.

Frequently asked questions