Cross-channel attribution splits credit for a conversion among the ads a customer touched on TV, search, social and the web. It shows which ads came before a sale but can’t prove which ones caused it. To learn what caused it, run an incrementality test, such as a holdout or geo experiment, which compares results with the ads and without them.
What cross-channel attribution means
Cross-channel attribution is the practice of deciding how much credit each ad channel gets when a customer saw several of them before converting. With one channel the answer is simple. Once you run two or three, customers move between them, and someone has to decide how to split the sale.
Here’s an example. A homeowner sees a roofing company’s ad on a streaming TV app, and the name sticks. A few days later they scroll past the same company’s ad on Instagram. Two weeks after that, a storm hits. They search “roofer near me,” recognize the name in a Google ad, and click through to request a quote. Three channels touched that lead. Attribution is the set of rules that decides whether Google gets all the credit, whether TV gets some, or whether the credit is shared.
Those rules matter because your budget follows them. If your reports say search did all the work, you’ll move money into search. If TV planted the name that made the search click happen, that move can backfire. Our guide to planning a multichannel approach covers how channels fit together. This page is about how to read the credit.
Why platform reports add up to more sales than you made
Every ad platform reports on its own contribution. Google’s dashboard measures Google, and Meta’s measures Facebook and Instagram. Each counts the conversions its ads touched, using its own conversion definition and attribution window. If a customer saw your Instagram ad and later clicked your Google ad, both platforms may claim the same sale.
You can check this in about ten minutes:
- Pick one month.
- Write down the conversions or revenue each platform reports for that month.
- Add them up.
- Compare the total with your actual sales records.
If the platform total is well above your real total, the reports overlap. That doesn’t mean any platform is lying. It means none of them can see the others, so you can’t add their numbers together or treat any single one as the full picture. For context on how a campaign reaches people on search, social and video, see our channels overview.
Common attribution models and what each one assumes
Every attribution model is a rule about who deserves credit. Each rule makes an assumption, and each assumption tilts results toward certain channels.
| Model | How it assigns credit | Good for | Known bias |
|---|---|---|---|
| First-touch | All credit to the first ad the customer interacted with | Seeing which channels introduce new people to the business | Ignores everything that persuaded the customer afterward |
| Last-touch (last-click) | All credit to the final click or view before the conversion | Quick to set up and easy to read | Over-credits search and retargeting; can credit ads for sales that would have happened anyway |
| Multi-touch / data-driven | Splits credit across touchpoints, using either a fixed rule or a model trained on converting and non-converting paths | A fuller view of trackable digital journeys | Needs user-level tracking, so it can’t follow linear TV or other media it can’t tie to a person |
Last-touch is still widely used because it’s quick to set up and easy to read, even though it overvalues the final touchpoint and ignores everything that came before it, as the Communications of the ACM guide to attribution explains. Google has moved away from it as a default: in Google Ads, Google’s attribution model documentation says data-driven attribution is now the default for most conversion actions. In Google Analytics 4, Google’s attribution reports guide lists three models in its Attribution reports: data-driven attribution and two last-click models.
Data-driven attribution looks more complete, but it can only share credit among the touchpoints it can track. Media that can’t be tied to a person, such as linear TV, gets little or no credit.
Why TV and connected TV get under-credited
Nobody clicks a TV ad. A streaming spot often works days or weeks later, when a viewer searches your name or types your web address. Search or direct traffic then collects the credit. The conversion also usually happens on a phone or laptop, a different device from the TV, which makes matching the two harder. Under last-touch, TV can create the demand and still appear to do nothing.
Some TV-to-web matching happens at the household level, but it covers only part of the journey. Our guide to measuring CTV campaigns in detail goes further into those methods.
Under-crediting is only half the risk. Attribution can also flatter a channel. Branded search and retargeting often reach people already on their way to buying, so they can claim sales they didn’t cause. Attribution can’t tell those two cases apart, which is why the next question matters.
Attribution vs. incrementality
Attribution shows who converted after seeing an ad. Incrementality shows who converted because of the ad.
An incrementality test works like a controlled experiment. You show ads to one group (a set of people or a set of geographic markets) and hold them back from a similar group. The difference in sales between the treatment and holdout groups is the lift the ads caused. In a geo test, you turn off spend in some markets and keep it running in others. Streaming TV suits this because most streaming platforms support geographic targeting exclusions.
| Method | Question it answers | Strength | Tradeoff |
|---|---|---|---|
| Attribution | Which ads came before conversions? | Fast and frequent, good for day-to-day changes | Not causal; depends on what can be tracked |
| Media mix modeling | How have spend and sales moved together over time? | Broad view for budget planning | Correlation-based; needs a long data history |
| Incrementality testing | How many sales did these ads cause? | Causal, and reaches non-clickable media like TV | Slower; needs a large enough budget and a long enough run to produce a clear result |
Tests have costs. The held-back market sees no ads while the test runs, and a test typically takes several weeks, long enough to cover the lag between seeing an ad and buying. In one geo holdout on streaming TV by Measured’s incrementality research, roughly half of the conversions the ad platform had claimed were actually incremental. A test’s result can be higher or lower than what your dashboards show.
How a small business can measure TV, search and social together
You don’t need an analytics team to measure honestly. Here’s a workable process:
- Install tracking on your website and count only the actions that matter to the business, whether that’s a call, a booking, or a purchase.
- Give each conversion a value. A rough average job value lets you compare revenue against spend instead of only counting clicks.
- Read last-touch and assisted conversions side by side. Last-touch shows who closed the sale; assisted conversions show which channels appeared earlier. A channel with few last-touch sales and many assists is probably doing early work.
- Watch branded search and direct traffic around any change to your TV campaign. A rise in people searching your name is a signal, not proof.
- Test the channel you doubt most. Run a geo or holdout test long enough to cover your normal sales cycle. Our guide to choosing which channels to test can help you decide where to start.
- Judge results against your own sales records, not platform totals.
Be realistic about the limits. A small budget or short test may not produce a clear result, and no measurement setup guarantees a return. Keeping campaigns organized makes all of this easier; a campaign workflow across channels shows one way to do it.
What Adwave reports and where its limits are
Adwave creates ads from your website and runs a single campaign across eligible channels, funded from a prepaid wallet. According to Adwave’s pricing page, the daily budget runs from $30 to $10,000. Not every channel runs at every budget: TikTok starts at $91/day and requires early-access selection. The Adwave FAQ says what can run depends on your budget and audience size, along with your creative, campaign settings, and content eligibility. Our guide to how one daily budget spreads across channels explains the mix, and the Adwave features page lists what’s included.
For measurement, you install your business’s tracking snippet, configure the actions you want to count, and give conversions a value so Adwave can report revenue and return on spend. On the How Adwave works page, the final step is tracking campaign activity and website conversions. The Adwave FAQ also notes that last-touch credit answers a different question than assisted conversions, and a Lift experiment answers a different question still, recommending the latter when you need to measure incremental impact.
The limits are the ones described above. Forecasts are estimates, and Adwave does not guarantee sales or a return on spend. Results depend on your offer and audience, as well as your budget, creative, and website, so check actual campaign results against your own sales records before deciding what to change.
