How to Measure Marketing Incrementality Without a Data Team

Platform-reported ROAS tells you what each ad platform wants credit for, not what your marketing actually caused. You do not need a data team or a fancy attribution tool to find out what is real. You need a few structured tests and a spreadsheet.

Why platform-reported ROAS misleads

Every ad platform measures success using its own attribution window and its own claimed conversions. Meta, Google, and TikTok will each take credit for the same sale if a customer touched all three before buying. Add them up and you can easily see reported revenue that exceeds your actual total revenue for the period.

Platforms also count conversions that would have happened anyway. If someone already searched your brand name, already had your product in their cart from an organic visit, or was already a repeat customer on a subscription, a retargeting ad shown to them and then credited as “driving” that sale did nothing incremental. The platform still reports it as a win, because the platform’s job is to justify its own spend, not to tell you the truth about causality.

This is not a reason to distrust every number. It is a reason to stop treating ROAS as ground truth and start treating it as one noisy signal among several.

MER as your north star

Marketing Efficiency Ratio (MER) is total revenue divided by total marketing spend across all channels, over the same period. It sidesteps attribution entirely because it does not care which platform gets credit for which sale.

MER = Total Revenue / Total Marketing Spend

The advantage is simplicity and honesty. You cannot double-count with MER the way platforms double-count with ROAS, because it is calculated at the business level, not the channel level. The disadvantage is that MER is a lagging, blended number. It will not tell you which channel or campaign is responsible for a change. It tells you whether your marketing system as a whole is getting more or less efficient over time.

Use MER as the metric you watch weekly to sense-check whether things are trending the right direction. Use the tests below when you need to know why, or which specific spend is actually working.

Geo holdout tests

A geo holdout is the closest thing to a controlled experiment most DTC brands can run without technical infrastructure.

How it works: Split your addressable market into two groups of comparable regions, matched on size and past sales performance as closely as you can. Turn off (or hold flat) a specific channel or campaign in the holdout regions while running it as normal everywhere else. Compare revenue growth between the two groups over the test window.

What you need:

Enough regions to split meaningfully. This works better for national or multi-region brands than for a single-city local business.
A baseline period beforehand to confirm the groups were tracking similarly before the test started. If holdout regions were already growing faster or slower than test regions, your result will be confounded.
A test window long enough to smooth out day-to-day noise. A few days is not enough. Plan for several weeks where possible.

What it tells you: The true incremental lift (or absence of lift) from the channel or campaign you turned off, because you are comparing otherwise-similar populations with and without the spend.

This is the highest-confidence method on this list, and also the most operationally demanding. Reserve it for testing your biggest spend line, not every small campaign.

Simple on/off tests

If you cannot split geographies cleanly, a time-based on/off test is the next best option.

How it works: Pause a specific channel or campaign entirely for a defined period (commonly 1-2 weeks). Watch what happens to total revenue and MER, not just to the channel you paused. Compare against a similar prior period, ideally accounting for typical week-over-week or seasonal patterns.

What to watch for:

If total revenue drops meaningfully when the channel is off, and recovers when it is back on, that channel is likely doing real incremental work.
If total revenue barely moves, the spend was probably cannibalizing sales that would have happened anyway through other channels or organic traffic.
Isolate one variable at a time. Turning off two channels simultaneously, or running the test during a sale event, will contaminate your read.

The honest caveat: on/off tests are noisier than geo holdouts because you are comparing across time, not across a matched control group, and time introduces seasonality, competitor activity, and random variance. Run the test for long enough that a single unusual day does not skew the result, and be skeptical of a “lift” under 10-15%, since that is within the range normal week-to-week noise can produce.

Post-purchase surveys

The simplest incrementality signal available to any store, regardless of size: ask the customer.

Add a one-question survey to your post-purchase page or confirmation email: “How did you first hear about us?” with a short list of options (search, social ad, friend, influencer, email, other). This is self-reported and imperfect, people misremember or default to the most recent touchpoint, but it is directionally useful and completely free to run.

Track this over time rather than treating any single week as gospel. If a channel’s survey-attributed share is consistently far below its platform-reported ROAS share, that is a signal worth investigating with a more rigorous test. If organic and word-of-mouth show up as a large share, that tells you real brand equity is building, something platform ROAS will never show you credit for.

Running a basic test with a spreadsheet

You do not need specialized software for any of this. Here is the minimum workable setup.

Build a weekly tracking sheet with columns for total revenue, total marketing spend, MER, spend and platform-reported ROAS by channel, and order count.
Log test periods explicitly. Add a column noting whether each week was a “normal,” “holdout,” or “on/off test” week, and what changed. Without this, you will not be able to reconstruct what caused a shift six months later.
Chart MER over time as your primary line. Overlay major spend changes as annotations so you can visually connect cause and effect.
For each test, calculate the delta. Compare average daily revenue (or MER) during the test window against a comparable baseline period, adjusted for known seasonality (holidays, paydays, past promotions).
Set a threshold for what counts as signal. Given the noise inherent in small-sample tests, treat anything under roughly 10% change as inconclusive rather than a real result. This keeps you from overreacting to normal variance.

This does not require statistical software. A basic spreadsheet with SUM, AVERAGE, and a line chart covers the analysis most DTC brands need to make real decisions.

Putting it together

No single method here is perfect on its own. Platform ROAS is directionally useful for optimizing within a channel but untrustworthy for comparing across channels or judging true incrementality. Geo holdouts give the strongest read but need scale and patience. On/off tests are more accessible but noisier. Surveys are free and easy but self-reported. MER is your steady, honest backstop that catches problems the other methods might miss between formal tests.

Run geo holdouts or on/off tests periodically on your largest spend lines, not constantly on everything. Watch MER weekly as your health check. Keep the post-purchase survey running in the background at all times, since it costs nothing and compounds in usefulness the longer you track it.

Where to start

Set up a weekly MER tracking sheet this week if you do not already have one. It is the fastest way to stop being misled by channel-level ROAS.
Add a one-question post-purchase survey. It costs nothing and starts generating a trend line immediately.
Pick your single largest spend line and run one on/off or geo holdout test next quarter, with the test period logged clearly so you can reference it later.
If an agency runs your paid media, make the test part of the engagement rather than something you run around them. Partners that keep the ad accounts in your name, Branva for example, make on/off and holdout tests straightforward because you can see every campaign yourself.
Treat any lift under about 10% as noise, not signal, until you have run the test more than once.