How to prevent wrong listing matches in a real-estate WhatsApp assistant

The first listing a WhatsApp assistant suggests sets the tone for the whole conversation. A right listing creates the feeling of “they understood me”; a wrong one pushes the customer to rely on their own search rather than the list. Once a wrong match happens, even the following right suggestions carry less weight. So the real issue in matching is not finding more listings, but never sending the wrong one.

Rule 1: Do not start searching with incomplete information

Most wrong matches come from looking at the portfolio before the customer has said what they want. For a customer who writes “a home in Bostanlı”, is it to buy or to rent, how many rooms, what budget? Every suggestion made without knowing is a guess. That is why Realtiq does not touch the portfolio until it has collected four criteria — transaction type, area, rooms and budget. If one is missing, it is completed with one question per message. The transaction type is especially critical: showing a rental to someone looking to buy is the most basic matching error, and the analysis is not considered complete without it.

Rule 2: Stretch the meaning, not the limits

Meaning-based search is powerful because it can find phrases like “quiet street” or “close to the sea” in listing descriptions. But that flexibility should not be applied to limits the customer states explicitly. In Realtiq the transaction type and budget filter always work together with the meaning-based search: listings that fail the gates are dropped first, the rest are ranked by closeness in meaning to the intent, and those below a threshold never enter the list.

Rule 3: Read the budget as an anchor, but keep it narrow

Customers usually state a round budget; someone who says 4 million is open to talking about the right home at 4.2 million. Treating the budget as a literal ceiling misses that home, while stretching it without limit makes the customer feel unheard. In Realtiq the default tolerance is 10% and can be adjusted per office; when a listing within the tolerance is shown, the price difference is stated openly. Details in budget tolerance.

Rule 4: Measure “nearby” against official geography

To a customer who says “somewhere close is fine too”, every listing in the same city can look like a candidate. Yet the two ends of the same district can be forty minutes apart by car. The Realtiq assistant knows Türkiye's official geography and places every candidate listing on a five-level proximity scale; listings in the distant level never enter the list. Two more traps are closed: neighbourhoods that share a name (such as Bahçelievler) are pinned to your office's city, and when it cannot be sure the customer is asked; spelling variations are matched to the official name. How it works is described in region intelligence.

Rule 5: A short, honest list

A long list is the easiest way to hide inaccuracy. Realtiq sends the customer up to 5 listings in order of fit. When there is no exact match, it says so: “There's no exact match in Bostanlı right now; there are two listings next door in Atakent.” When nothing fits, instead of sending irrelevant listings it suggests criteria flexibility or a waiting list. The customer always knows what is an exact match and what is a nearby alternative.

The portfolio side: matching cannot beat data quality

Even the best rules can produce wrong results from incomplete data. Three details make the difference:

• Fill in the property type. An empty property type is assumed to be an apartment; if it is empty on a land, commercial or shop listing, the listing may be suggested to the wrong customer.
• Write detailed descriptions. Meaning-based search reads the description; details like “south-facing, in a gated complex, walking distance to a school” directly improve accuracy.
• Deactivate sold listings. The spreadsheet syncs every 30 minutes; a listing you deactivate is not suggested to customers.

All the rules for preparing the spreadsheet are in setting up your portfolio sheet.

In summary

Preventing wrong matches is not the work of one clever algorithm but of rules applied in order: complete information first, then hard limits, then flexible meaning, and finally a short and honest list. You can see how Realtiq applies these rules together on the property matching page.

← All articles