Incrementality Testing for Indian D2C Brands: A Practical 2026 Guide

Incrementality testing estimates the conversions or revenue caused by advertising by comparing an exposed group with a credible control group. Indian D2C brands should use it alongside attribution and MER, especially before major budget shifts. A useful test needs comparable groups, clean conversion tracking, enough sample size, a fixed decision rule and uncertainty analysis.

Platform dashboards are useful for daily optimisation, but attributed revenue is not the same as revenue caused by advertising. A customer who clicked a retargeting ad and purchased may have bought anyway. Incrementality gives the growth team a way to estimate that counterfactual.

What does incrementality testing actually measure?

A lift study separates an eligible audience into a treatment group that can receive the advertising and a control group that cannot. The difference in outcomes estimates the campaign's causal contribution.

Google defines incremental conversions as treatment conversions minus control conversions. When the groups are different sizes, compare normalised conversion rates rather than raw totals. Google's current Conversion Lift measurement guide also distinguishes incremental conversions from the standard attributed conversions in campaign reporting.

Three decision metrics are especially useful:

| Metric | Formula | Decision it supports | | --- | --- | --- | | Incremental conversions | Test conversions - expected control conversions | How many additional outcomes did the campaign cause? | | Incremental CPA | Relevant test spend / incremental conversions | What did each caused conversion cost? | | Incremental ROAS | Incremental conversion value / relevant test spend | Did the caused revenue justify the spend? |

Use the HML Incrementality Lift Calculator to normalise unequal group sizes and estimate observed lift, incremental CPA and iROAS. The calculator does not calculate statistical significance, so it is a planning and interpretation aid rather than a substitute for experimental analysis.

Why can platform ROAS and incremental ROAS disagree?

Attribution asks which touchpoint receives credit. Incrementality asks whether the conversion would have happened without the advertising.

A campaign can report a strong ROAS while producing modest incremental value when it mostly reaches existing customers, brand searchers or shoppers already close to purchasing. The reverse can also occur: an upper-funnel campaign may look weak in last-click reporting but create later demand that a holdout study detects.

This is why a useful measurement stack combines:

• platform attribution for campaign operations;
• MER for blended business efficiency;
• incrementality for causal validation; and
• contribution margin and cash flow for the final business decision.

The ROAS benchmark guide explains why even an accurate attributed ROAS must be interpreted against unit economics. The first-party data guide covers the measurement inputs needed before a lift study is trustworthy.

Which incrementality test should an Indian D2C brand use?

User-level platform holdout

The advertising platform randomly assigns eligible users to treatment and control groups. This can provide strong randomisation, but availability and campaign eligibility vary. Google's current setup documentation says user-based Conversion Lift is not available to every account and some campaign types require contact with a Google representative. Check eligibility before building a quarterly plan around it.

Geo holdout or matched-market test

Comparable cities, states or market clusters are assigned to treatment and control conditions. This can work when user-level suppression is unavailable or when the business wants a cross-channel view. The design must account for regional differences, spillover, distribution coverage, promotions and seasonality.

Campaign experiment

An A/B experiment compares a changed campaign setup with the original. This is useful for questions such as whether Performance Max adds value beyond an existing mix. Google documents Performance Max uplift experiments, but the exact experiment available depends on account eligibility.

Time-based pause test

Turning activity off before and on is easy to understand but vulnerable to seasonality, competitor activity, promotions and demand shifts. Use it only when a better control is not feasible, and avoid treating a simple before-versus-after difference as causal proof.

A worked example: interpreting lift without overclaiming

This hypothetical example uses round numbers and is not an HML client result.

An Indian D2C brand runs a randomised holdout with equal-sized groups:

• test audience: 50,000 people;
• test conversions: 650;
• control audience: 50,000 people;
• control conversions: 500;
• relevant test spend: INR 1,50,000; and
• average order value: INR 2,500.

The test conversion rate is 1.3%, while the control conversion rate is 1.0%. The observed results are:

• 150 incremental conversions;
• 30% relative lift;
• INR 1,000 incremental CPA; and
• 2.5x incremental ROAS.

These figures describe the observed difference. Before reallocating budget, the analyst still needs to review uncertainty, randomisation, contamination, conversion lag and whether the spend definition matches the decision.

Google's lift-study setup guidance recommends checking study power and says a longer study or larger holdout may be needed when the expected signal is weak. It also recommends allowing the duration to cover the normal conversion lag.

How should the team plan a defensible test?

1. Write the decision first. Specify what budget or campaign decision will change if the test is positive, negative or inconclusive.
2. Choose one primary outcome. Purchases, qualified leads or contribution value should match the commercial goal. Do not choose the metric after seeing the result.
3. Confirm tracking readiness. Deduplicate conversions, validate purchase values and keep observation windows consistent. Google's setup guide recommends compatible conversion actions and stronger first-party measurement inputs.
4. Estimate feasibility. Use historical baseline rate, minimum effect worth detecting, group size and expected conversion lag to plan sample and duration.
5. Protect the control. Avoid overlapping campaigns, remarketing or regional promotions that expose the control to the tested intervention.
6. Predefine the decision rule. Document the required certainty, acceptable iCPA or iROAS, and how an inconclusive result will be handled.
7. Read the result with business economics. Compare incremental value with margin, fulfilment cost, repeat purchase behaviour and the opportunity cost of the holdout.

Google says its lift studies go beyond clicks and impressions by comparing exposed and control groups. It also announced lower spend thresholds and methodology improvements for incrementality testing in 2025. Availability is still account- and campaign-dependent, so teams should verify the current account interface rather than assume access.

What mistakes make a lift test unreliable?

• Comparing raw conversions when the groups are different sizes.
• Running the test across a major sale or festival without equivalent control conditions.
• Allowing the control group to see overlapping ads from another campaign.
• Stopping as soon as the result looks favourable.
• Ignoring delayed conversions for products with longer consideration cycles.
• Calling an observed lift statistically conclusive without uncertainty analysis.
• Using attributed revenue in the numerator when calculating incremental ROAS.
• Generalising one channel or campaign result to the entire marketing mix.

An inconclusive study is not a failed study. It may show that the expected effect is smaller than the design can detect, that tracking needs improvement or that a longer test is required.

How should incrementality change budget decisions?

Do not replace platform reporting with a single lift result. Calibrate it.

If a campaign has a 5x platform ROAS but a 2x iROAS, use the causal estimate for strategic allocation and the platform metrics for tactical optimisation. Repeat material tests around major budget cycles rather than assuming one result is permanent. Creative, competition, product mix and customer behaviour can all change the incremental effect.

For an example of coordinated upper- and lower-funnel execution, see HML's full-funnel consumer electronics case study. It provides operating context, not a substitute for a controlled lift result.

What should you do before the next budget review?

Run your proposed or completed holdout through the Incrementality Lift Calculator, document the assumptions and list the sources of contamination. If the decision is commercially material and the test design is unclear, request a performance measurement review before moving the budget. HML's performance marketing service connects campaign operations with MER, margin and incremental outcomes.