Revenue Attribution
| Model | What it credits | Proof strength | When to use |
|---|---|---|---|
| Last touch | Final interaction | Weak (correlation) | Simple journeys |
| Multi-touch (W-shape) | First/mid/last split | Moderate | Long journeys |
| Causal (holdout/geo) | Incremental revenue | Strong | Budget decisions |
💡TL;DR
Revenue attribution is moving from click-credit to causal proof. Use multi-touch for direction, then validate big bets with holdouts or geo experiments. SMBs avoid over-investing in noisy channels and defend budgets with evidence.
Definition
Methods for assigning revenue credit to marketing and sales interactions. Modern practice blends multi-touch models with causal tests to validate lift.
🏢What This Means for SMB Teams
When dollars are scarce, proof matters. Attribution with experiments shows which channels truly add revenue.
📋Practical Example
An 80-person SMB lender ($28M ARR) relied on last-touch credit, so paid search consumed 42% of spend while partner referrals looked weak. They implemented W-shape multi-touch plus quarterly geo holdouts. Before: CAC $640, payback 7.2 months. After 120 days, holdouts showed referrals drove +22% incremental approvals and search +6%. Budget shifted 12% from search to partner enablement. CAC dropped to $510 and payback improved to 5.6 months while monthly originations grew from 1,150 to 1,340.
🔧Implementation Steps
- 1
Map all touchpoints and standardize UTMs/IDs so every form, call, and referral is captured.
- 2
Select a directional MTA model (e.g., W-shape) and run it weekly to show contribution by channel.
- 3
Run quarterly causal tests (holdout/geo split) on the top two spend channels to validate incremental lift.
- 4
Publish an attribution readout with CAC, payback, and confidence level; freeze budget shifts without evidence.
- 5
Refresh model weights every quarter using the latest conversion and test data.
❓Frequently Asked Questions
Do we still need experiments if we run multi-touch attribution?
Yes. MTA is directional and correlational. Use it to spot candidates, then run holdouts or geo splits to prove incrementality before moving budgets. Without experiments, noisy channels can still look good.
How much data do we need for reliable attribution?
For directional reads, 200–300 conversions per month across channels is workable. For causal tests, design for minimum detectable effect (e.g., 10-15% lift) and run until you hit the sample size; smaller volumes may need longer test windows.
⚡How Optifai Uses This
ROI Ledger stores both multi-touch reads and causal results for each play.
📚References
- •
- •
Related Terms
Revenue Attribution Model
A causal-first framework that attributes incremental revenue to specific plays and signals using holdouts or geo splits, not just multi-touch click weights.
Multi-Touch Attribution
An attribution approach that distributes credit across multiple touchpoints in the buyer journey rather than only first or last touch.
ROI Ledger
A ledger system that tracks every AI action with a UUID and attributes actual revenue contribution using holdout testing.
Experience Revenue Action in Practice
Now that you know the terms, see them in action. Experience signal detection, automated actions, and ROI proof with Optifai.
Learn More