Why this matters
Last-click attribution makes one partner look like a hero and hides the rest. A buyer might read a review on an SEO affiliate, click a coupon affiliate a week later, and finally convert via a paid social publisher — but only the social publisher shows up in your report. Every other partner looks like dead weight and gets paused.
Trcker runs three multi-touch models every night against the past 30 days of converted journeys so you can see the full picture without changing how you pay.
What Trcker computes
For every approved conversion, Trcker stitches together every click made by the same visitor in the prior 30 days, dedupes repeats by the same partner, and runs three models against the path:
Linear — Every partner in the path gets an equal share of the credit. Works great for short paths, ignores recency and position.
Position-based (40/40/20) — Same as Google Analytics' default: the first touch gets 40%, the last touch gets 40%, and any middle touches split the remaining 20%. If there are only two touches in the path, each gets 50%. If there's only one, it gets everything. Good for programs where the awareness moment and the closing moment matter most.
Time-decay (7-day half-life) — Every touch's weight decays as 2-age/7, so a touch from 7 days ago is worth half of a touch today, and a touch from 14 days ago is worth a quarter. Best for shorter purchase cycles where recency matters more than order.
Each model produces attributed conversions and attributed revenue per partner, per offer, per brand. Last-click attribution is still the truth for payouts — this is a reporting overlay, not a payment rule.
Payouts stay on last-click
Nothing about your payouts changes. Partner commissions are still computed from last-click attribution as captured at the moment of conversion. Multi-touch attribution is purely for understanding the _contribution_ each partner makes to your program — it answers "who's really helping?" without changing "who gets paid?"
How often it runs
Every day at 08:30 UTC. Snapshots are upserted per-brand per-partner per-offer per-model, so re-running the cron is safe and fast. The window is a rolling 30 days; older conversions roll off the report automatically.
Where to see it
Dashboard: open your brand and go to Reports > Attribution. You'll see every partner with last-click as the baseline column plus all three multi-touch models, a delta chip on each showing +/- vs. last-click, and a "paths touched" count.
The green/red chip is the actionable signal — green means the partner contributes more under multi-touch than their last-click number suggests (undercounted by your current reports). Red means they contribute less (overcredited).
What you'll see
A partner who closes 100 conversions a month under last-click might show:
| Model | Attributed conversions | |---|---| | Last-click (baseline) | 100 | | Linear | 78 | | Position-based | 88 | | Time-decay | 95 |
That 78 in the linear column is the signal. It means this partner is often involved in conversions other partners close — not a solo hero. Conversely, if a partner's linear attribution is _higher_ than their last-click number, they're an awareness driver whose contribution your last-click reports undercount.
Future: Shapley & Markov
Shapley-value and Markov-chain attribution are planned as a follow-up Python job. Both require training a path-to-conversion probability model on converted _and_ non-converted journeys, which is a heavier pipeline than the closed-form models above. The three models Trcker ships with today are ~80% of what most operators actually use.
Related
- Incrementality holdouts — measures the _causal_ lift from a partner (stronger than attribution)
- API — query attribution snapshots programmatically