How MisMatchMaker checks your listings

Sources it compares

From a hotel name or Google Maps link, the scanner searches for the property's own listing on each supported platform and only accepts a result that is served over HTTPS on that platform's own domain:

• Booking.com
• Expedia
• Hotels.com
• TripAdvisor
• Agoda

A field is only ever judged from the sources that were successfully read for that specific scan. Listings that cannot be found or accessed are reported as such rather than assumed to agree.

Fields it compares

The scanner compares 15 fields across five categories. Each field carries a weight that sets its share of the score; the weights total 100. Impact reflects how badly a mismatch tends to hurt a guest booking.

How a scan runs

1. 1

   Discover

   A site-restricted search locates the property's listing on each supported platform. Only HTTPS URLs on the platform's own domain are accepted, and each result is labelled as a strong or weak match.

2. 2

   Scrape

   Each selected listing is fetched and its page content captured. Length limits and URL checks are applied at the boundary because listing pages are treated as untrusted input.

3. 3

   Extract

   The captured content is converted into a structured record of the supported fields, along with a self-reported extraction confidence score from 0 to 100.

4. 4

   Validate

   Every extracted record is checked against a strict runtime schema before it is used. A malformed or unexpected response is treated as a failed source, not trusted blindly.

5. 5

   Compare

   Values are normalized and compared field by field across the successfully read sources, each field is classified, and the score is calculated.

How evidence is classified

Before anything is compared, each source's contribution to a field is given one of four evidence states. The distinction between missing, unverified, and failed is deliberate: unavailable evidence is never turned into a positive result.

Value

A comparable value was extracted and is trusted — either from structured AI extraction that met the confidence threshold, or read directly from the listing text. Only these count toward a match and the score.

Unverified

A value was found on the listing, but the AI extraction's self-reported confidence was below the threshold. It is shown so you can see what the listing said, but it never counts toward a match or the score.

Missing

The source was read successfully, but no comparable value for this field was present. Missing evidence is not agreement, and it does not prove the underlying service or policy is unavailable.

Failed

The listing could not be accessed or returned malformed data. A failed source is reported as failed; it is never silently converted into a positive result.

How each field is judged

A field is only called consistent when at least two sources each produced a reliable, comparable value. Comparison happens after normalization, so semantically equivalent values — time formats, currency notation, booleans, and whitespace — are treated as equal while the original value is kept visible as evidence.

Consistent

At least two reliable sources produced comparable values and they agree after normalization.

Conflict

At least two reliable sources produced comparable values and they disagree.

Insufficient

Only one reliable source produced a comparable value, so there is nothing to compare it against.

Unavailable

No reliable source produced a comparable value for the field.

How the score is calculated

The consistency score is the share of comparable weight that is consistent:

• No score is shown unless at least two sources each contributed a comparable value.
• A source only counts toward that two-source minimum if it actually produced a comparable value — a source that was read but contained nothing usable does not.
• The report also shows evidence coverage — how much of the total field weight was populated across your sources — so a high score built on thin coverage is visible rather than hidden.

What it does not do

• It compares only the fields listed above. It does not evaluate photos, general descriptions, room or bed types, rates, availability, taxes, or full amenity inventories.
• It measures content parity, not rate parity. It is not a channel manager, PMS, booking engine, or rate-shopping tool.
• It reads public OTA listings only. It cannot see your extranet, channel manager, or internal source of truth, and it cannot confirm that an edit you made was submitted or published.
• Agreement between OTAs is not proof of correctness — every source can carry the same stale value. Always confirm a field against your own approved value.
• Results are session-only. They are held in your browser for the current session and are not stored on our servers, so refreshing or sharing the results URL will not reproduce a scan.