Attribution in real estate investing does not work the way the marketing industry describes it, and pretending otherwise produces reports that are precise and wrong.
A seller got a letter in February and threw it away. In April they saw a sign. In May a neighbor mentioned you. In June they searched your name, read a page, and called the number from a second letter. Which channel gets the deal? Any single answer is false, and most attribution software will confidently give you one.
What You Are Actually Trying to Decide
The useful reframe: you are not trying to assign credit fairly. You are trying to decide where next quarter's money goes.
Those are different questions, and the second one is much easier. You do not need to know that direct mail deserves sixty percent of the credit for a closing. You need to know whether closings happen when mail is running and stop when it is not.
Once the goal is decision support rather than accounting, the standard problems mostly dissolve. You stop needing a model and start needing a record of what sellers touched.
The Methods That Are Worth the Effort
Ask, and record it as text. The single highest-value thing, and the most commonly done badly. Ask every lead how they heard about you. Record their actual words rather than forcing a dropdown, because "I got a letter a while back and then saw your sign on Route 9" contains two channels and a timeline, and a dropdown destroys all of it. Read these monthly. Patterns appear quickly.
A distinct phone number per channel. The most reliable mechanical signal available, and cheap. A number on the mail, a different one on signs, a different one on the website. Calls are then attributed without anyone remembering anything, and you get call volumes by channel as a side benefit.
A distinct URL or landing page per channel. Same logic for anything that produces web traffic, and it doubles as message match, which improves conversion independently. That argument is in message match.
Campaign tagging on links. Standard tracking parameters on ad and email links, so digital sources arrive labeled. Worth doing, and be aware it only covers the fraction of your traffic that arrives by clicking.
First contact date, recorded once and never overwritten. Not attribution in the usual sense and it is what lets you connect a closing to the campaign that actually started it, months earlier. Without it, delayed deals get credited to whatever touched them last, which is the systematic error described in what a seller lead is actually worth.
A source field on every lead record, filled at creation. Not reconstructed later. A lead whose source was not captured at the moment it arrived is permanently unattributable, and this is where most investor tracking actually fails.
The Methods That Are Not Worth It
Multi-touch attribution models. Linear, time decay, position based. These require enough conversions for the model to mean anything, and you have four this quarter. They produce a precise-looking split from almost no data, which is worse than admitting uncertainty.
Last-click attribution. The default in most analytics, and it systematically credits the channel closest to the conversion, which in this business is usually search for your own name. That makes branded search look like your best channel when it is actually a record of everything else working.
Pixel-based cross-device tracking. Increasingly unreliable, increasingly restricted, and aimed at a problem that matters more for e-commerce than for a business where the conversion happens by phone.
Anything requiring the seller to do something. Codes to mention, QR codes to scan, forms to identify themselves. Response rates on these are low enough that the data is a biased sample of unusually compliant people.
The Holdout Test
The most underused method, and for channels that resist tracking it is the only honest one.
Turn a channel off for a defined period and watch what happens to total deal flow. Not leads, deals. If nothing changes over a full cycle, the channel was not doing what you thought. If flow drops, you have causal evidence that no attribution model can give you.
This is also the only method that survives the objection that your channels influence each other. A holdout measures the total effect of removing something, including whatever it was contributing to other channels, which no per-channel number can capture. That interaction is why cost per deal by channel is a guide rather than a verdict, as noted in cost per lead versus cost per deal.
Two cautions. The window has to exceed your sales cycle, so a ninety day cycle means a test measured over at least a quarter and probably two. And it works best on channels large enough that their absence would be visible, since turning off something that produces one deal a quarter tells you nothing.
The variant that avoids risking deal flow: hold out a geography rather than a channel. Run mail in two counties and not in a third comparable one, then compare. This is more work to set up and it gives you a real answer, which is more than most attribution produces.
Reading Self-Reported Data Properly
What people say is unreliable in specific, predictable ways, and knowing the direction of the bias makes the data usable.
People report the last thing they remember, so the most recent touch is over-credited and anything from months back is under-credited. Direct mail is systematically under-reported because a letter that prompted a search does not feel like the cause. Word of mouth is under-reported because people forget who mentioned you. Online is over-reported because "I found you online" is the easy answer.
Which means: when a seller says they found you online, ask what made them search. That one follow-up question surfaces most of the missing first touch, and it is the highest-return question in the whole intake conversation.
Also record how long ago. "I got your letter about six months ago" tells you something your monthly report cannot, which is that your mail is working on a delay your reporting window never sees.
What Good Enough Looks Like
A realistic target for an investor doing a handful of deals a month.
None of it requires attribution software, and all of it requires the lead record to be the single place this lives, which is the case made in the guide to investor CRMs.
You know spend by channel by month. You know roughly how many leads each channel produced. You have self-reported source text on most leads, and phone numbers or URLs giving a mechanical signal on the rest. You can say, for each closed deal, which channels were involved, even when the answer is more than one. And you review it quarterly rather than weekly, because quarterly is the shortest window where closings mean anything.
That is not a model and it will not produce a percentage split. It will tell you which channels appear in the paths that end in money, which is the decision you were trying to make.
The Intake Questions Worth Standardizing
Attribution quality is decided in the first ninety seconds of the first conversation, and most investors leave it to whoever picks up.
Four questions, asked the same way every time, produce most of the available signal.
How did you hear about us? Recorded as their words, not categorized on the spot. Categorizing happens later, when you can see patterns across many answers.
What made you reach out today, specifically? This separates the trigger from the source. Someone may have known about you for months and called because a tenant left. Both facts matter and only one is a channel.
Roughly when did you first come across us? The single most under-asked question in intake. It reveals delays your reporting window never sees and it is what tells you your mail is working on a lag.
If they say online: what made you search? The highest-return follow-up in the whole conversation, because it recovers the first touch that self-reporting systematically loses.
Write the answers into the lead record as free text. The temptation is to build a tidy dropdown, and a dropdown discards exactly the detail that makes the record worth having. Categorize monthly when you read them in a batch, which is also when the patterns become visible.
The Trap of Measuring Only What Is Measurable
Worth stating plainly, because the bias is strong and it costs investors real money.
Digital channels report themselves. Mail, signs, referrals and word of mouth do not. An investor who weights decisions by data quality will systematically move budget toward whatever is easiest to measure, regardless of what actually produces.
This is how investors end up abandoning direct mail, which is under-reported by nature, in favor of channels with dashboards. The dashboard is not evidence of performance, it is evidence of instrumentation.
The correction is to weight by outcome rather than by measurability, and to use holdout tests specifically on the channels that resist tracking. If a channel cannot report itself, that is an argument for testing it deliberately rather than for discounting it. What to do with the answer is in comparing marketing channels, and where attribution sits in the wider set is in marketing metrics for real estate investors. The rule to carry out of all of it: a channel that cannot report itself is not a channel that is not working. It is a channel you have to test deliberately, and most investors never do.