Nobody loses attribution on purpose. It just does not survive to the payout.

Every brokerage can tell you what a lead costs. Very few can tell you what a lead was worth.

The gap between those two sentences is where most marketing budgets are decided. It is not a measurement problem in the way people usually mean. The data was captured. It simply did not survive the journey to the money.

Attribution is a chain of custody, not a field

Most systems treat source as an attribute of a lead: selected at creation, stored once, reported on later. That framing is the problem.

Source is not a property of a record. It is a claim about causation that has to be carried intact across every handover between the first enquiry and the commission payment. There are usually six or seven of those handovers. Each is an opportunity to drop it. Dropping it is the default, because carrying it requires deliberate design and losing it requires nothing at all.

The places it dies

At capture, when a person types it. A portal enquiry arrives. An agent is trying to reach the buyer inside four minutes, because that is the discipline the brokerage rightly drilled into them. They create the record and pick something from a dropdown, or leave it blank, or type the name of the portal they assume it came from. Anything a person selects under time pressure, at the moment they are trying to do something else, will be wrong at a rate nobody measures.

At the jump to WhatsApp. The enquiry lands in one channel and the relationship moves to another within minutes. From that point the conversation that actually produces the deal happens somewhere with no connection to the record. Six weeks later the record says the lead went cold. The deal closed anyway, from a thread on a phone.

At reassignment. A lead moves from one agent to another, or from a call desk to a sales team. In many operations the receiving side creates a fresh record rather than taking custody of the existing one, because creating is one click and transferring is a process. The new record's source is the internal handover. The portal that paid for the enquiry has disappeared from its own result.

At the duplicate. The same buyer enquires three times over eight months: a portal in March, a different portal in June, a referral in November when they are finally ready. That is three records. The November one closes. March and June sit in the report as dead leads. The channel that generated genuine early interest is penalised for the fact that the interest took time to mature, and the channel that caught the buyer at the end takes credit for all of it.

At deal creation. A deal is frequently created from a listing or a property rather than from a person, because that is the shape of the transaction. Whatever the lead record knew about origin is not inherited, because the deal was never designed to hold it.

At the commission. This is where it dies for good. The payout is calculated somewhere with columns for agent, split, gross, net and adjustment. There is no column for source and there was never going to be, because commission is a finance artefact and finance was not in the room when attribution was designed. Money leaves the business with no memory of what caused it.

Why the reporting still looks healthy

None of that produces an obviously broken report. It produces a confident one.

The system holds a source field, so it reports on the source field. Leads by source. Cost per lead by source. Conversion rate by source, calculated across the records that kept theirs. Every one of those numbers is defensible on its own terms, and every one measures the top of the funnel, because the top of the funnel is the only place the data is still intact.

The consequence is structural and it always runs the same direction. Channels that produce volume are measurable. Channels that produce quality are not. A portal generating four hundred enquiries and two completions out-reports a referral network generating thirty introductions and six, because the portal's contribution is legible at the point where legibility still exists.

So budget follows volume. The portal contract renews because it can evidence its lead count. The referral programme, the past-client campaign, the developer relationship and the introducer network — low volume, high conversion, largely invisible in a lead-count report — receive whatever attention is left over. Nobody decided this. The measurement decided it.

The uncomfortable version: a brokerage in this position is not being misled by its suppliers. It is funding decisions with the only number it is able to produce, and that number is systematically biased against its best channels.

What has to be true for source to survive

The requirements are not sophisticated. Each one has to hold, or the chain breaks at that link.

Origin is recorded by the system, not asserted by a person. If a channel can deliver an enquiry, it can deliver the fact of where the enquiry came from. Human entry belongs to genuine exceptions — the walk-in, the phone call, the introduction at a dinner — and those should be the only records where anyone is choosing from a list.

The same person is recognised as the same person. Identity resolution across a portal enquiry, a WhatsApp number, a walk-in and a referral, matched on something durable rather than on a name. Until this holds, duplicates will keep quietly reassigning credit to whichever record happened to close.

Records are transferred, not recreated. Reassignment has to move custody and keep origin. That is a policy decision as much as a system one, and it usually requires making the correct action easier than the incorrect one.

More than one touch is kept. Do not force the business to choose between first touch and last touch and then collapse everything into a single answer. Keep both, separately. Most real deals have more than one parent, and a model that insists on one is guaranteed to be wrong about the interesting cases.

The deal inherits from the relationship. Origin has to travel into the transaction record rather than stopping at the lead.

The commission carries a reference to origin. This is the requirement that changes the conversation, because it is the only version of attribution that survives into the numbers a managing director actually reads. Once the commission line knows where it came from, the question stops being what your leads cost and becomes which sources produced collected commission last year, against what was spent on them.

It will still be imperfect, and that is fine

Attribution is never clean. A buyer who saw a hoarding, followed the company for a year, enquired through a portal and was closed on a referral from their accountant will not resolve into one truthful source, and any system claiming otherwise is rounding.

The goal is not precision. It is to be wrong in a known and consistent direction rather than randomly, and to stop the loss happening at the exact points where the record changes hands. A brokerage that can trace collected commission back to origin, even roughly, is making a different quality of decision from one working off lead counts. It is also, incidentally, the same underlying requirement as knowing which of your producers is genuinely carrying the business.

That chain — enquiry through to the commission that pays for it — is what we mean by holding one operating record, and it is the thing the brokerage page is about.