← All playbooks·Digital attribution

What you need first

• Multi-touch path report for the conversion window, pulled from the analytics platform (GA4 Explorations, Adobe Attribution IQ, or the MTA vendor), with touchpoint order and timestamp per converting user.
• Last-click conversion totals by channel for the identical date range, pulled from the same platform's standard conversion report.
• Agreed attribution weighting model and split (for example 40/40/20 U-shaped, or a time-decay half-life), signed off with the analyst or client rather than picked ad hoc for this run.
• Channel taxonomy mapping so touchpoints tagged with different UTM values roll up to one channel (paid-search-brand and paid-search-generic both roll to Paid Search).
• Materiality threshold for flagging credit shifts, taken from the variance-flag standard the team already uses.

The procedure

1. Pull the multi-touch path report for the campaign and conversion window, producing the raw list of touchpoint sequences and timestamps per converting user.
2. Group the raw paths into pattern clusters by channel sequence, producing path groups with a conversion count attached to each sequence (for example Search>Display>Social>Email: 250).
3. Assign position weights from the agreed model to first, middle, and last touchpoints in each pattern, producing a weight table per position.
4. Apply the weight table to each pattern group's conversion count, splitting the middle-position share evenly across however many middle touchpoints that pattern has, producing credited conversions per channel per pattern.
5. Sum credited conversions across all pattern groups by channel, producing the final MTA credit table.
6. Pull the last-click credit table for the same window and calculate the delta by channel, producing the over-credit and under-credit list.
7. Flag any channel whose credit shift crosses the materiality threshold, producing the short list the analyst has to explain before handoff.

Worked through with numbers

250 conversions in the review period share the exact path Paid Search > Display > Paid Social > Email (last click). Under last-click, Email gets all 250. Applying a 40/40/20 U-shaped split: first touch (Paid Search) gets 40% of 250 = 100; last touch (Email) gets 40% of 250 = 100; the remaining 20% (50 conversions) splits evenly across the two middle touchpoints, 25 each to Display and Paid Social. Total check: 100 + 100 + 25 + 25 = 250, matching the path's raw conversion count. Read it as Email's real job is closing, not creating demand: its credited conversions drop from 250 to 100, a 60% cut, while Paid Search picks up 100 conversions it got zero credit for under last-click, which is the number that should move next quarter's search budget.

Where it goes wrong

• Split the middle-position weight across however many middle touchpoints a path actually has, not a flat 20% per middle touch, or paths with three middle touches will over-total past 100%.
• Collapse repeat touchpoints from the same channel within one path into a single position before weighting, or a channel that appears three times in the middle of a path inflates its own share against channels that appear once.
• Pull the last-click comparison from the identical date range and conversion window as the MTA run, since a mismatched window is the usual reason the delta table looks wrong, not the weighting model.
• Weight view-through touches separately from click touches, or a display impression scores the same as a paid search click and overstates display's share of the path.

How to know it is right

Terms used

• Multi-Touch Attribution (MTA)
• Last-Click Attribution
• Assisted Conversions

← All playbooks Calculated in ValueMetrics →