What is linear attribution?

Linear attribution splits an order's credit equally across every paid touchpoint in the attribution window. Three touches, a third of the margin each; five touches, a fifth each. It takes no position on whether starting or finishing a journey matters more, which is exactly why it makes a sensible default.

Formula

linear margin (touch) = order margin ÷ number of paid touches

Worked example

An order with €90 of contribution margin follows a four-touch journey: two search clicks (different sessions), one social, one retargeting. Each touch is credited €22.50, so search books €45 and the other two €22.50 each. Under last-touch the same order reads retargeting €90, search €0 — same data, same month, different story.

How Saldo Metrics computes it

The attribution engine builds each order's journey from pixel events — UTM-tagged page views by the same customer in the 30 days before the purchase event, one touch per session, in time order — and under linear writes one fact_attribution row per touch carrying 1/n of the order's EUR revenue and contribution margin. The fractions of one order always sum back to exactly its full margin. No-touch orders go to organic at full weight; uncosted orders are excluded outright.

Linear is also the app's default working model: unless an org configures another in its marketing settings, v_campaign_true_roas and v_channel_true_roas read the linear rows, so every true-ROAS figure you see is a linear-attributed one by default.

Why it matters

With multi-touch journeys, any single-touch model is a claim about which touch mattered; linear is the model that refuses to make that claim. Its per-channel totals are the least sensitive to journey shape, which makes it the steadiest basis for recurring decisions — and the natural baseline the Δ last-vs-linear column measures other models against.

Common mistakes

• Treating equal split as measured truth. Linear is a neutral convention, not evidence — the Markov model, once an org has enough conversions, is the only data-driven read available here.
• Comparing linear figures against platform dashboards. Platforms approximate last-touch; linear will read lower for closer channels and higher for prospecting ones, by construction.
• Overlooking that session deduplication shapes the split. Five page views in one session are one touch, not five; a channel generating many short sessions gets more weight than one generating one long session.

Where you see this in the app

Marketing → Attribution, as the linear column of the model comparison and the default model of the per-model breakdown — and implicitly in every true-ROAS figure while the org's model is the default.