# Incrementality Testing for B2B: Measuring What Actually Works

# Incrementality Testing for B2B: Measuring What Actually Works

> **Quick answer:** **Incrementality testing measures what a marketing channel actually *adds* — the conversions that wouldn't have happened without it — rather than what it merely gets credited for.** It matters because attribution overstates impact: it assigns credit to touchpoints that were correlated with conversions, not necessarily causal. The cleanest methods are experiments — geo holdouts (run a channel in some regions, not others, and compare) and pause tests (turn a channel off and measure the effect). Incrementality is the honest answer to "does this actually work?" — and it frequently reveals that brand search and retargeting are less incremental than attribution suggests.

**Key takeaways**

- **Incrementality = what a channel truly adds,** not what it's credited for.
- **Attribution overstates impact** — it measures correlation, not causation.
- **Experiments are the gold standard** — geo holdouts and pause tests.
- **It often humbles retargeting and brand search** — much of their "impact" isn't incremental.
- **Test the expensive and the doubtful** — where the answer changes budget decisions.

Every attribution model tells you which touchpoints got credit; none tell you which touchpoints *caused* the conversion. Incrementality testing does — by running actual experiments. This guide covers what incrementality is, why attribution overstates, the testing methods, how to run a clean geo holdout, and what these tests commonly reveal.

## What is incrementality?

**Incrementality** is the share of conversions a channel *causes* — the ones that wouldn't have happened without it. If a channel gets credited with 100 conversions but 70 of those people would have converted anyway (through organic, brand search, or another channel), its incremental contribution is 30. Incrementality testing measures that true, causal contribution through experiments, rather than inferring it from correlational [attribution models](https://www.growthspreeofficial.com/blogs/multi-touch-attribution-b2b-saas). It answers the only question that ultimately matters for budget: if I turned this off, what would I actually lose?

## Why does attribution overstate impact?

Because attribution measures correlation, not causation. It credits touchpoints that *appeared* on the path to conversion, but appearing isn't the same as causing. The clearest example is [brand search](https://www.growthspreeofficial.com/blogs/brand-bidding-b2b): someone already decided to buy, searches your brand, clicks your ad, and converts — attribution credits the ad, but that person would have found you anyway. The same is true of much [retargeting](https://www.growthspreeofficial.com/blogs/linkedin-ads-abm-retargeting-companies-viewed-ads-didnt-convert): you show ads to people already far down the funnel and take credit for conversions that were coming regardless. Attribution systematically over-credits channels that sit close to conversions and under-credits the [demand creation](https://www.growthspreeofficial.com/blogs/marketing-budget-allocation) that actually started them. Incrementality corrects for this.

## Incrementality vs. attribution

| Dimension | Attribution | Incrementality |
|---|---|---|
| Measures | Correlation (who got credit) | Causation (what was added) |
| Method | Models on observed paths | Experiments (holdouts) |
| Bias | Over-credits last-touch channels | Isolates true lift |
| Effort | Low (always-on) | Higher (designed tests) |
| Answers | "Which touchpoints appeared?" | "What would we lose if we stopped?" |

They're complementary: attribution runs continuously for day-to-day optimization; incrementality tests periodically to check whether attribution is lying to you about specific channels.

## What are the main incrementality testing methods?

- **Geo holdout tests.** Run a channel in some geographic regions ("test") and not others ("control"), then compare outcomes. The difference is the incremental lift. The cleanest, most common method for B2B.
- **Pause (on/off) tests.** Turn a channel off for a defined period and measure the change in total conversions. Simple, but confounded by timing and seasonality if not controlled.
- **Matched-market tests.** A refined geo test that pairs similar markets to improve the comparison.
- **Ghost/PSA ads.** Show a control group unrelated ads instead of yours, then compare conversion rates — cleaner but harder to run, especially at B2B's lower volumes.

For most B2B teams, geo holdouts and pause tests are the practical options; the more rigorous methods often need volume B2B doesn't have.

## How do you run a geo holdout test?

1. **Pick the channel and question.** E.g., "is our brand search bidding incremental?"
2. **Split geographies into test and control** that are as similar as possible in size and behavior.
3. **Run the channel in test, pause it in control** for a defined window long enough to capture your [sales cycle](https://www.growthspreeofficial.com/blogs/reduce-saas-churn).
4. **Measure total conversions in both** — not just the channel's own reported conversions, but overall pipeline in each region.
5. **Compare.** If test regions produce more total conversions than control, the channel is incremental; if they're similar, it wasn't adding much.
6. **Account for noise.** B2B's lower volumes make results noisier, so run long enough for a meaningful read and interpret modest differences cautiously.

The key discipline: measure *total* outcomes in each region, not the channel's self-reported numbers — the whole point is to see what happens to the business, not to the channel's own dashboard.

> **Field note:** The most humbling — and valuable — incrementality tests are usually on the channels everyone assumes are winners. Retargeting and brand search look phenomenal in attribution reports precisely because they sit next to conversions, so they get the credit. Run a clean holdout and you often find a chunk of their "impact" would have happened anyway: the retargeted user was going to convert, the brand searcher already decided. That doesn't make these channels worthless — some incremental lift and defensive value is real — but it resizes them honestly, and it usually reveals that under-credited demand creation deserves more budget than last-click reporting suggested.

## What do incrementality tests commonly reveal?

Patterns that recur across B2B:

- **Brand search is partly non-incremental** — some of it defends against competitors, some just buys free organic clicks.
- **Retargeting is less incremental than it looks** — it often reaches people already converting.
- **Demand creation is more valuable than attribution credits** — its impact shows up as lift that last-click reporting misses.
- **Some channels are pleasant surprises** — occasionally a channel attribution under-credited proves genuinely additive.

The consistent theme: incrementality shifts credit from close-to-conversion channels toward demand creation, correcting attribution's built-in bias.

## When should you run incrementality tests?

Test where the answer would change a decision: expensive channels (is the spend justified?), channels you suspect are non-incremental (brand, retargeting), and before making a big budget shift. You don't need to test everything continuously — incrementality testing is periodic and targeted, complementing always-on [attribution reporting](https://www.growthspreeofficial.com/blogs/marketing-attribution-reporting). Test the big bets and the doubtful cases; trust attribution for the routine.

## Honest limitations

- **B2B volume makes tests noisy.** Lower conversion counts mean results are harder to read cleanly; you need longer windows and larger geographies, and modest differences may not be conclusive.
- **Confounds are real.** Seasonality, other campaigns, and market events can muddy holdout and pause tests; design carefully and interpret with judgment.
- **Long cycles slow the read.** A B2B channel's incremental impact may take months to appear, so tests must run long.
- **It's effort.** Proper incrementality testing takes design, patience, and discipline — it's not a dashboard toggle.
- **Results aren't permanent.** Incrementality changes as your market and mix change; re-test periodically rather than treating one result as forever true.

## Frequently Asked Questions

### Q1. What is incrementality testing?
Incrementality testing measures the conversions a channel actually causes — the ones that wouldn't have happened without it — through experiments like geo holdouts and pause tests, rather than inferring impact from correlational attribution. It answers "if we turned this off, what would we actually lose?"

### Q2. How is incrementality different from attribution?
Attribution measures correlation — which touchpoints appeared on the conversion path and got credit. Incrementality measures causation — what a channel actually added — using experiments. Attribution over-credits channels close to conversions; incrementality isolates true lift. They're complementary, not interchangeable.

### Q3. What is a geo holdout test?
A geo holdout test runs a channel in some geographic regions (test) and pauses it in others (control), then compares total conversions between them. The difference is the incremental lift. It's the cleanest, most practical incrementality method for most B2B teams.

### Q4. Why does attribution overstate channel impact?
Because it credits touchpoints that appeared on the conversion path, and appearing isn't causing. Brand search and retargeting sit close to conversions, so they get credit for people who would have converted anyway — over-crediting close-to-conversion channels and under-crediting the demand creation that started the journey.

### Q5. What do incrementality tests usually reveal?
That brand search is partly non-incremental, retargeting is less incremental than it looks, and demand creation is more valuable than attribution credits. The consistent pattern is that credit shifts from close-to-conversion channels toward demand creation, correcting attribution's built-in bias.

### Q6. When should you run incrementality tests?
Where the answer would change a decision: on expensive channels, on channels you suspect aren't incremental (like brand and retargeting), and before major budget shifts. It's periodic and targeted, complementing always-on attribution — test the big bets and doubtful cases, trust attribution for the routine.

### Q7. Is incrementality testing hard for B2B?
It's harder than for high-volume ecommerce because B2B's lower conversion counts make results noisier and long sales cycles slow the read. Geo holdouts and pause tests are the practical methods; run them over longer windows and larger geographies, and interpret modest differences cautiously.

**Sources & further reading**

- Google Ads and Meta experiment tools (geo experiments, brand lift) for structured incrementality testing (confirm current features).
- Measure total conversions in test vs. control regions using your own CRM data, not channel self-reported numbers.

*This guide is educational; incrementality testing requires careful design and B2B volumes make results noisy, so run tests long enough and interpret with judgment against your own data.*

---

*Related guides: [Multi-Touch Attribution for B2B SaaS](https://www.growthspreeofficial.com/blogs/multi-touch-attribution-b2b-saas) · [Should B2B Bid on Its Own Brand?](https://www.growthspreeofficial.com/blogs/brand-bidding-b2b) · [Marketing Budget Allocation](https://www.growthspreeofficial.com/blogs/marketing-budget-allocation) · [Retargeting for B2B SaaS](https://www.growthspreeofficial.com/blogs/google-ads-budget-split-b2b-saas-brand-nonbrand-retargeting-demand-gen-2026) · [Marketing Attribution Reporting](https://www.growthspreeofficial.com/blogs/marketing-attribution-reporting).*