When the marketing budget comes down by a fifth, the usual method is quick and feels rigorous. Sort channels by cost per lead. Cut from the bottom.

That method answers a question nobody asked. It tells you which channels were present when leads arrived. It cannot tell you which channels caused them, and those are different things with different consequences.

The distinction has a name. The first is attribution. The second is incrementality. Your lead source report primarily tells you where credit was assigned. A budget decision asks a different question: what would happen if you changed or removed the channel?

Present is not the same as responsible

Attribution tells you who was in the room. Incrementality tells you what would have happened if they had not been.

A channel can appear on thousands of lead records and contribute almost nothing, because those renters would have found you anyway. Another can appear on few records and still contribute earlier in the decision process, without receiving much last-touch credit. Our article on why your lead source report measures the last thing that happened covers why the credit ends up where it does.

Here is the part that matters for a budget decision. Observational reporting on its own generally cannot settle which is which. Google's own data science team puts it directly: it is generally not possible to determine the incremental impact of advertising by merely observing clicks, visits or sales across time.

Cutting on a cost-per-lead ranking is a decision made on exactly the kind of evidence they describe as insufficient on its own. Attribution alone cannot answer "what if we cut this?" An experiment or another credible causal-inference approach is needed to estimate that.

Why a bad cut looks free at first

The trap has a specific shape, and it is worth understanding before you act, because it explains why so many cuts seem to work.

Renters do not decide on the day they inquire. There is a consideration period before the inquiry, and some channels may influence prospects during it, before they ever submit a guest card.

Cut one of those channels and the effect is not immediate. Renters who were already exposed before the cut continue to arrive for a while. Your lead volume holds. The cost saving is real and the damage appears to be zero.

The effect arrives later, when renters who would have been reached during the period you switched it off reach the point of inquiring, and fewer of them know you exist. By then the drop is removed in time from the decision that caused it, and it looks like a softening market.

The length of that consideration period depends on the property, market, renter and channel, so estimate it from your own data rather than borrow a timeframe from another business. It is also why a cut should never be judged on its first month of results.

Before cutting anything, sort by role

For a budget review, it can help to think about channels by the role they may play. The cost-per-lead ranking treats them identically. They are not.

This classification changes the question you ask of each number.

A closer with a high cost per lead may still be worth the spend if it produces signed leases efficiently. That is why cost per lead should be considered alongside downstream leasing outcomes.

An assister with a high cost per lead is behaving exactly as expected. The report is structurally unable to show you what it does, which is not the same as it doing nothing.

Dead weight is often the most overlooked saving, because nobody looks at a small line item. The useful test here is not attribution performance but direct evidence: duplication, staleness, or no identifiable owner. If you syndicate the same unit through overlapping feeds, some of that spend is duplicated. We cover how that happens in why syndication multiplies your listing errors.

The test: switch it off somewhere, keep it on somewhere else

For anything you cannot classify with confidence, the most direct way to find out whether a channel matters is to remove it and watch what happens, with a comparison group that indicates what would have happened otherwise.

This is the logic behind a method Google has published for its own advertising. Vaver and Koehler's work on measuring ad effectiveness using geo experiments describes randomly assigning non-overlapping geographic regions to a treatment or control condition, then comparing outcomes. They describe the approach as conceptually simple, systematic in design, and easy to interpret.

A portfolio may give a larger operator enough properties or markets to construct a comparison that a single-property operator simply does not have. But having multiple properties does not automatically make the test valid. Properties can differ in demand, pricing, availability, asset type and competitive conditions. The comparison needs to be designed carefully, and random assignment is preferable where it is practical.

The protocol, adapted for leasing:

1. Choose the channel and the question. One channel at a time. "Does pausing social advertising reduce signed leases?" is testable. "Is our marketing working?" is not.

2. Create comparable groups. Use the properties and markets available to you, with random assignment where practical. Where possible, keep groups in different markets so that advertising in one does not reach renters looking in the other.

3. Establish a baseline first. Run everything as normal and record signed leases per property. In Google's published geo-experiment methodology, the pretest period is typically four to eight weeks. You are checking that the groups move together before anything changes. If they do not, the comparison will not tell you much.

4. Pause the channel in one group only. Keep everything else identical. Same pricing, same concessions, same staffing, same other channels. Any other change contaminates the result.

5. Run the test long enough to reach leases, not just leads. In the same methodology, the test period is typically three to five weeks, with additional time potentially needed when effects are delayed. For leasing, the appropriate duration depends on how long prospects normally take to move through your funnel, and on the volume of leases available to measure.

6. Compare signed leases, not leads. Measure how each group's leases changed relative to its own baseline. If the paused group falls behind the control group in a well-designed experiment, that difference can provide evidence of the channel's incremental contribution. If both moved together and the difference is not distinguishable from normal variation, the test may indicate that the channel's incremental effect is small relative to what this experiment could detect.

A well-designed control group helps account for changes that affect both groups, including some seasonal or market-wide movements. The remaining difference can then be used to estimate the incremental effect of the channel, subject to the assumptions and quality of the experiment. That matters in leasing more than most places, because demand shifts through the year, as we covered in what your days-on-market number is actually measuring.

The honest limit: you may not have enough leases

This deserves stating plainly, because a method is only useful if you know when it will not work.

Measuring the effect of advertising is hard, and it is hard for a statistical reason rather than a technical one. Lewis and Rao's widely cited research in the Quarterly Journal of Economics examined why the returns to advertising are so difficult to measure even in large, well-designed experiments: the effect of any single channel is often small relative to the natural variation in outcomes.

For leasing, the implication is direct. At a property with relatively few monthly leases, even a small change in the number signed can represent a large percentage movement. That makes it difficult to distinguish a channel's effect from normal variation.

So the test works best when:

• You have enough properties or markets to detect a meaningful difference. More observations make the estimate more precise, but there is no universal property count that guarantees a useful test.

• The channel is large enough for the decision to matter. If the potential savings are very small, the cost and complexity of a formal test may outweigh the value of the answer.

• You read a null result carefully. A test that shows the paused group down noticeably is a strong signal. A test showing no clear difference tells you the effect is smaller than this test could detect, which is not the same as the effect being zero.

If your portfolio is too small to construct a meaningful comparison, the test will not give you a reliable answer, and it is better to know that before spending a quarter on it.

If you cannot run a clean test

Not every portfolio can run a clean incrementality experiment, and not every decision can wait for one. When a decision has to be made without it, start with the evidence you actually have.

Review channels with little recent activity, duplicated spend, unclear ownership or no identifiable leasing outcome. Then distinguish between channels where the available data shows weak downstream performance and channels where the data is simply insufficient.

That distinction matters more than any ranking. A lack of evidence that a channel works is not the same as evidence that the channel does not work.

Explaining the cut to owners

A marketing cut changes the numbers owners see, and the timing described above means the change may not show up until well after the decision. That is worth raising before it happens rather than explaining afterwards.

If you frame it at the time as a test with a defined measurement period, a later dip becomes an expected data point rather than a surprise. We cover what belongs on that page in what your owner report doesn't say about leasing.

What you need before any of this works

Every step above depends on one thing: counting signed leases per property per week, reliably, and knowing which channels were involved.

That sounds basic. In practice it is often the obstacle. Leads live in one place, applications in another, signed leases in a third, and joining them by property and week becomes a manual exercise that gets done once and abandoned. Our article on the four conversion rates behind your headline number covers why the stages between inquiry and lease tend to sit in different systems.

It also depends on counting leads consistently in the first place. If one group's properties dedupe inquiries across channels and the other's do not, the comparison is broken before it starts, which is the problem covered in what counts as a lead.

A cost-per-lead ranking will always be available, and it will always look authoritative, because it is a sorted list of real numbers.

It is still the wrong instrument for this decision. It describes where credit landed, which is a record of the past, while a budget cut is a prediction about the future. Those need different evidence, and the evidence that answers the second question is best established by changing something deliberately and watching what follows.

A portfolio may be able to do that where a single property cannot, provided it can see signed leases by property and by week. RIOO's leasing workflow brings inquiries, applications and leasing activity into a structured process, while its dashboards and reports provide visibility into property and portfolio performance. That creates a stronger foundation for examining leasing outcomes alongside the operational data behind them.

Before the next budget conversation, check whether you could produce last quarter's signed leases per property, per week, by originating channel. If that takes an afternoon rather than a query, the test has a prerequisite you need to solve first.

Frequently asked questions

Q1. How do I decide which multifamily marketing channels to cut?
Classify channels by role first: closers, assisters and dead weight. Review dead weight using direct evidence such as duplication, staleness or unclear ownership. For anything else, consider testing before cutting by pausing a channel in one group of properties while keeping it in a comparable group, and compare signed leases.

Q2. What is the difference between attribution and incrementality?
Attribution tells you which channels were present when a lead or lease occurred. Incrementality tells you what would have happened without a channel. Budget decisions depend on incrementality, which observational reporting alone generally cannot establish.

Q3. Why do cuts sometimes look successful at first?
Some channels may influence prospects during a consideration period before they inquire. After a cut, renters already exposed continue to arrive for a while, so lead volume holds. The effect can appear later, when fewer newly exposed renters reach the point of inquiring.

Q4. How long should a marketing test run?
In Google's published geo-experiment methodology, pretest periods are typically four to eight weeks and test periods three to five weeks, with extra time potentially needed for delayed effects. For leasing, the right duration depends on your inquiry-to-lease cycle and the lease volume available to measure.

Q5. Can a small portfolio run this kind of test?
Often not reliably. The effect of any single channel can be small relative to the natural month-to-month variation in leases, so testing works best with enough properties or markets to detect a meaningful difference, and on channels large enough for the decision to matter.