What a property matching engine owes you, and what most of them never say

Every CRM built for property sells you matching. You put in a client brief, it gives you back a list. The list is ordered. Somewhere on each row is a percentage.

Ask the software why the third one scored 78 and the fourth scored 74 and you will usually get nothing back. That silence is the problem, because the agent is the one who has to defend the recommendation to a client, not the software.

A score without a reason is not a recommendation

An agent forwarding a property is putting their own credibility behind it. If the only justification available is "the system ranked it highly", they will do one of two things: send it anyway and hope, or ignore the list and work from memory. Both are worse than having no tool.

So every match in Doment carries the reasons it scored the way it did, in plain words: within budget, right number of bedrooms, handover inside the stated window, payment plan lighter than the client asked for. Not a breakdown of internal weights, which nobody wants to read. The statements an agent can actually repeat to a client.

Some requirements eliminate. Others trade off.

There are two kinds of requirement in any brief, and treating them the same is where matching usually goes wrong.

Some things are absolute. A client with a firm ceiling of AED 2 million does not want to see a AED 3 million unit sitting mid-list because it is otherwise perfect. That is not a near miss, it is the wrong property.

Everything else is a preference, and preferences trade against each other. A project that hands over a quarter later than hoped but sits in exactly the right community may well be the best thing available. A ranking is only worth reading once the impossible options are already gone, and once the remaining trade-offs are visible rather than buried in one number.

A missing figure is not a bad one

This is the decision we would defend hardest, because it changes results more than anything else on this page.

Most Dubai off-plan projects do not publish a payment plan. Only 101 of the 1,727 projects we track do, a gap we wrote about in our first data report. If a missing payment plan is treated as a bad payment plan, then every project that simply did not advertise its terms sinks to the bottom of every list, permanently, for a reason that has nothing to do with whether it suits the client.

So an unknown is not scored as a negative. It is set aside, the project is judged on what is actually known about it, and the gap is named on the result: no published payment plan. The agent sees exactly what is missing and can go and ask the developer, which is the correct next action and one no software can take for them.

The alternative quietly removes most of the market from view and never mentions it.

What this looks like on a Tuesday afternoon

A client says three bedrooms, Dubai Marina or close to it, under AED 4 million, ready to move into within about a year.

A useful list does not simply return everything in Dubai Marina under AED 4 million. It puts the properties that satisfy the firm parts of that brief in front, tells the agent which of the softer parts each one misses, and flags the ones where something the client cares about is unknown rather than absent. The agent can then send three properties with a sentence of reasoning attached to each, instead of eleven with a percentage attached to each.

That difference is the entire product. A shorter list an agent can defend beats a longer list they have to apologise for.

The same standard everywhere

The rule underneath all of this is that software should say what it does not know. A missing number gets named, not guessed. A figure we cannot stand behind is not shown at all. A recommendation arrives with its reasoning attached, so the person putting their name to it can decide whether they agree.

It is the same standard we hold our published market figures to, and for the same reason: a number you cannot check is worth less than one you can.