A WORKED EXAMPLE: THE DEATH SPIRAL IN ACTION
A DTC brand sees Paid Social at 0.7× last-touch ROAS. The dashboard says cut it. The team pauses Paid Social entirely; they keep Email and Retargeting at full spend, expecting "efficient" channels to absorb the volume.
Thirty days later, total revenue is down 22%. Email and Retargeting ROAS both crater — not because they got worse, but because there are fewer people for them to retarget and email. The paid social pause starved the funnel of new customers; the "closer" channels had nothing left to close.
They turn Paid Social back on. With tiered attribution — first-touch for ToFU reporting, linear for mid-funnel, last-touch for tactical optimization — Paid Social shows up as 25% of total contribution across the funnel, not 0.7× ROAS on a closing channel it was never supposed to be.
## Common Objections
### "But last-touch is what my CFO understands"
Your CFO understands incorrect data. The question is whether you want decisions based on simple-but-wrong or accurate-but-complex.
Show them the holdout test results. CFOs understand "we cut this channel and revenue dropped."
### "Our sales cycle is short—does this still apply?"
If your average journey has 3+ touchpoints, yes. Even in e-commerce with 7-day cycles, multi-touch journeys are common:
```
E-commerce Journey (7 days):
Day 1: Facebook ad (browse)
Day 3: Google Shopping (compare)
Day 5: Retargeting ad (reminder)
Day 7: Direct visit (purchase)
Last-touch: 100% to Direct
Reality: Facebook started it, Google Shopping was consideration
```
Short cycles don't mean single-touch journeys.
### "We don't have enough conversions for complex attribution"
Tiered attribution uses the same models you already know—just applied to different questions. First-touch and last-touch require no additional data. Linear is a simple formula.
If you can do last-touch, you can do tiered. It's a framing change, not a data requirement change.
## Summary
Last-touch attribution fails for funnel forecasting because:
1. **It ignores 80-90% of customer journeys** — Only the last touchpoint gets credit
2. **It over-credits closers** — Email, branded search, retargeting look like heroes
3. **It under-credits introducers** — Paid social, content, display look like failures
4. **It leads to the death spiral** — Cutting awareness starves the funnel
**The solution:** Use tiered attribution—different models for different funnel stages.
| Stage | Model | Question |
|-------|-------|----------|
| ToFU | First-touch | What introduces prospects? |
| MoFU | Linear/Participation | What nurtures them? |
| BoFU | Last-touch | What closes them? |
Last-touch isn't wrong everywhere—it's wrong for the wrong question. For understanding closers, it's appropriate. For forecasting full-funnel revenue, it's misleading.
## Further Reading
**On Building Tiered Attribution:**
- [How to Use Different Attribution Models by Funnel Stage](/articles/funnel-stage-attribution) — The implementation guide
- [How to Build a Bottom-Up Revenue Forecast with MTA](/articles/bottom-up-revenue-forecast) — Complete forecasting workflow
**On Last-Touch Limitations:**
- [Last-Touch Attribution: When to Use It](/articles/last-touch-attribution) — When last-touch IS appropriate
- [How to Choose the Right Attribution Model](/articles/how-to-choose-attribution-model) — Model selection framework
## Key takeaways
- Last-touch ignores 80-90% of customer journey touchpoints
- Forecasts built on last-touch over-credit closers (email, retargeting)
- Cutting 'underperforming' awareness channels tanks the entire funnel
- Use tiered attribution: different models for different funnel stages
## FAQ
**Why do most companies still use last-touch for forecasting?**
Legacy systems, platform defaults (GA4, ad platforms), and simplicity. Last-touch is easy to implement and explain. The problem is that 'easy to understand' doesn't mean 'accurate for planning.' Most teams don't realize the distortion until they've cut awareness channels and watched their funnel dry up.
**What attribution model should I use for revenue forecasting?**
Use a tiered approach: First-touch for top-of-funnel (awareness), Linear or Participation for mid-funnel (consideration), and Last-touch only for bottom-funnel (conversion). This captures contribution at each funnel stage and produces more accurate forecasts than any single model.
**How much does last-touch distort channel credit?**
Dramatically. In typical multi-touch journeys, channels like paid social get 40%+ of first-touch credit but under 10% of last-touch credit—a 4x distortion. Email and branded search show the opposite: they get 3-5x more credit under last-touch than their true funnel contribution warrants.
**Can I fix last-touch forecasting with lookback windows?**
No. Adjusting lookback windows changes which touchpoint is 'last,' but doesn't change the fundamental problem: only one touchpoint gets credit. A 90-day window with last-touch is still ignoring 90% of the journey.
**What happens if I forecast with last-touch and then cut awareness channels?**
Classic attribution death spiral: you cut 'underperforming' awareness channels → fewer people enter funnel → email/retargeting audience shrinks → closing channel ROAS drops → you increase closing spend → diminishing returns → CAC rises, revenue plateaus. Recovery takes 6-12 months.