How a store uses multi-touch attribution to move budget

Multi-touch attribution sounds like a reporting feature. In practice it is a budgeting tool. When every touch and every verified order sit in one record, you can ask which channels start sales, which finish them, and which are claiming credit for work someone else did. This post walks through how a store actually uses that to move money between Google, Microsoft, Meta and TikTok, one decision at a time.

Start with the disagreement

The useful signal in multi-touch attribution is not any one model. It is the gap between models.

Open the Attribution page and put first click next to last click for the same 30 days. If a channel gets roughly the same credit under both, it works alone: it introduces the buyer and closes the sale. If a channel gets far more credit under first click than last click, it opens journeys that other channels finish. If it gets far more under last click, it is closing sales that something else started.

Say your store did $400,000 in verified online revenue last month across four paid channels. An illustrative comparison might look like this:

Ad spend is counted once per channel and does not change when you switch models. That is deliberate. The revenue moves, the cost does not, so the return on ad spend column changes meaning as you change the lens.

Read the table. TikTok opens a lot of sales and closes few. Meta closes more than it opens. Google and Microsoft are fairly balanced, with a tilt toward closing. Those are four different jobs, and you would not fund them the same way.

Decision one: protect the openers

TikTok's last-click number looks weak. On the platform's own dashboard it may look weaker still, because TikTok's tag misses the buyer who comes back through Google three days later. A store looking at last click alone would cut it.

The first-click view says the opposite. TikTok is where a large share of new journeys begin. Cut it and the Google and Meta numbers fall a few weeks later, because there are fewer journeys for them to close.

The practical move is to judge a prospecting channel on first click and on the new-customer share of its orders, not on last click. Keep the budget where the journeys start. Then confirm it with the influenced view, which gives every touching source full credit and shows how many Google and Meta orders had a TikTok touch somewhere earlier.

Decision two: question the closers

Meta's last-click number is strong. A lot of that is retargeting: ads shown to people who already visited. The position-based model gives Meta less than last click does, because the first touch belonged to someone else.

The question is whether Meta is closing sales that would have closed anyway. Two checks help. First, look at time between touches. If the Meta click lands hours before the order on a journey that already reached checkout, the ad probably nudged a buyer who was already there. Second, compare the customer mix. If most of Meta's last-click orders are from repeat customers, it is harvesting, not hunting.

Neither check proves the retargeting is worthless. Reminder ads have a job. But it justifies moving a portion of Meta budget from retargeting toward prospecting, and watching whether the first-click column moves.

Decision three: fund the quiet channel

Microsoft Shopping is small and easy to ignore. In the table above it earns a higher return than Google under every model. Lower volume, less competition, same catalog.

Because the pixel captures Microsoft's click ID on arrival and matches it to the verified order, the number is real rather than an estimate from a tag that fires part of the time. The move is to raise Microsoft's budget in steps and watch whether the return holds as volume grows. It often does for a while, then flattens. The attribution pages will show you where.

Decision four: choose the window from your own data

Every platform ships a default lookback window. Yours should come from your buyers. The Pixel shows the time from first touch to order, and lets you set a 7- to 90-day lookback and compare models under each.

If your product is a consumable, most orders close within a day and a short window with a time-decay model is honest. If your product is researched for weeks, a short window pushes late orders into "direct" and starves the campaigns that started them. Say a third of your orders close more than a week after the first click. Under a seven-day window, that third belongs to nobody. Under a 30-day position-based model, the campaigns that opened them get their share.

Set the window from your distribution, not from the platform's default.

Make the change, then measure the change

Attribution tells you where credit belongs today. It does not tell you what happens after you move money. So each budget move is a small experiment:

1. Move a bounded amount. Ten to twenty percent of a channel's budget is enough to see a signal without breaking anything.
2. Note the date. Compare the period after against the period before, under the same model and window.
3. Watch the whole table, not one row. If TikTok spend goes down, watch Google's first-click share two weeks later.
4. Reverse quickly if the openers dry up.

The comparison view shows the previous period beside the current one with the change per cell, so the before-and-after reads in one glance.

What the platforms see while you do this

One more piece makes the budget moves stick. Each platform's automated bidding learns from the conversions it can see. If you shift money toward a channel whose tag misses half your orders, its bidding will underperform no matter what your attribution says.

That is why Orkie sends verified orders back to Google, Microsoft, Meta and TikTok as enhanced or offline conversions, matched to the click that started the journey and gated by the shopper's advertising consent. The platform learns from your real sales, its number starts to agree with yours, and the budget decision you made on the attribution page is the same decision its bidding is acting on.

Where this leaves you

The dashboards will keep disagreeing with each other. That is fine. Your job is not to reconcile them. It is to keep one record you trust, read it under a few models, move money in small steps, and check the result. Multi-touch attribution is the tool that makes each of those steps a decision instead of a guess.

Want to see your four channels under one model? Talk to us.

---

Want this for your business? Talk to us How the pixel works