← All posts

Blog

AttributionROASEcommerce

Multi-touch attribution ecommerce guide: models, trade-offs and what to expect from your data

22 November 20267 min readSyncBeacon Team

Multi-touch attribution ecommerce teams actually use is less complicated than the vendor diagrams suggest. A shopper rarely buys after one ad. They see something on Pinterest, click a Meta ad a few days later, search your brand on Google, and then order. Attribution is the rule you use to divide credit for that order among those touches.

This guide explains the common rules, where each one misleads, and what you can and cannot expect from an attribution tool. In the last sections it describes what SyncBeacon's Beacon Attribution Engine does, including what it leaves out.

Key takeaway: No attribution model tells you the truth about what caused a sale. Each one is a different, defensible way to split credit. Use more than one model side by side, check how much your answer changes, and treat the output as a decision aid for budget moves, not as a measurement of cause and effect.

What attribution is, and why platforms disagree

Every ad platform reports conversions on its own terms. Meta counts orders it can link to its ads, Google does the same for Google, and so on. Add the platform totals together and they usually exceed the orders in your store, because each platform claims credit for a shopper who touched several of them. That is the basic problem attribution tries to fix.

A multi-touch approach starts from your own orders instead. It builds a record of the touches before each order (campaign links, click IDs) and splits each order's revenue across them. The total never exceeds your real revenue, which is the useful property.

This depends on having the data first. If purchases are missing from the ad platforms because of ad blockers or browser limits, any attribution built on top of them inherits the gaps. That is why attribution and server-side tracking belong together: orders from your store are the reliable end of the journey.


The common models

Take one journey: a shopper clicks a Pinterest ad (touch A), a Meta ad (touch B), and a Google brand search ad (touch C), then buys.

ModelHow it splits creditWhere it misleads
First touchAll credit to A, the touch that started the journeyIgnores everything that closed the sale
Last touchAll credit to C, the final touch before the orderOver-rewards brand search and retargeting; ignores discovery
LinearEqual credit to A, B and CTreats a passing click like a decisive one
Time decayMore credit to touches closer to the order, so C gets most and A leastUndervalues discovery on long journeys
Position-basedMost credit to the first and last touches, less to the middleThe weights are a convention, not a measurement

Ad platforms also offer data-driven models that estimate each touch's contribution from their own data. Those can be useful inside one platform, but they only see that platform's touches.


First-click vs last-click, in practice

First-click vs last-click is the comparison most teams start with, and the gap between the two is informative on its own.

Shops with short, impulse-style purchases may find the models agree closely. Higher-priced or considered purchases usually spread out more. You will not know which you are until you look at your own journeys.


What you need for multi-touch attribution to work

  1. Touch data. Landing-page parameters such as UTMs and ad click IDs have to be captured when the shopper arrives and remembered until they order.
  2. Order data you trust. Orders read from your store, not only from browser tags.
  3. Ad spend by channel. Revenue credit alone is not ROAS. You need spend from the same ad accounts.
  4. Consent handling. If a visitor declines marketing tracking, their touches cannot be followed. Journeys will be incomplete for those shoppers, and that is the correct behaviour.

Attribution-adjusted ROAS

Platform ROAS divides the revenue a platform claims by what you spent there. Attribution-adjusted ROAS divides the revenue credited to that channel by your attribution model by the same spend. The numerator changes; the spend does not.

Because the credited revenue depends on the model, so does the ROAS. That is not a flaw to hide. Seeing a channel's ROAS move between models tells you how sensitive your budget decision is. If a channel clears your break-even under every model, the decision is easy. If it passes under one and fails under another, you need more than one report to decide.

One naming note: SyncBeacon's dashboard uses the label Hybrid True ROAS for the blended figure. In plain words it is attribution-adjusted ROAS. It is a modelled number, and nothing in it is validated against a holdout test or guaranteed.


What the Beacon Attribution Engine offers

The Beacon Attribution Engine is SyncBeacon's attribution feature. The attribution page has the product overview; here is the working summary.

What it does

What it does not do

Plan and data limits

Attribution is included on Growth and Enterprise, not on Pilot or Essentials. See pricing for the current plans. Two data limits matter. First, click IDs are only remembered for a period set by your plan, and they are not stored when marketing consent is denied. A 90-day window does not help if your plan remembers click IDs for less than that, so check both numbers before you pick a window. Second, a journey only includes the touches the browser script could see. A visit where the script never loaded or was blocked will not appear.


How to use attribution without fooling yourself

  1. Start with last touch and first touch side by side. Note which channels move most.
  2. Then look at linear or time decay for the same period. Channels that hold their position across models are the safe calls.
  3. Change budgets in small steps and watch store revenue, not only the attribution report. The report explains; your revenue confirms.
  4. Re-run the comparison after each campaign change. Do not reuse last quarter's conclusion.

Remember too that every number here is only as complete as your tracking. Closing the gap between store orders and what platforms see is covered in the Google Ads conversion gap post.


Get started

See where your orders really come from

Connect your store and ad accounts, then compare attribution models on your own journeys.

Get started

Attribution models allocate credit; they do not prove causation. Any ROAS figure from a model is an estimate and not a guarantee of results.