Marketing Analytics · ServiceMOFU
Attribution and incrementality that platforms can't mark themselves
The short answer
Every ad platform over-credits itself. Add up what Meta, Google and your affiliates each claim, and the total often exceeds your actual revenue. Attribution software can reassign that credit between channels, but it can't tell you what would have happened without the ad. Only experiments can.
PMG uses three layers: blended metrics that can't be double-counted, platform attribution for day-to-day allocation within channels, and incrementality tests to calibrate both.
Three layers of measurement
- Blended (MER, new-customer CAC) — the truth about the business, reviewed weekly
- Platform attribution — useful for allocating within a channel, never for judging it against others
- Incrementality — geo holdouts, conversion lift studies and scheduled pauses that reveal causal contribution
- Marketing mix modelling — at scale, a statistical model of each channel's contribution over time
How a geo holdout works
- Pick comparable markets with similar historical sales patterns.
- Pause or reduce a channel in the test markets for two to four weeks.
- Keep it running in control markets.
- Compare total revenue between them, adjusted for pre-test differences.
- Calculate incremental return — what the channel actually added.
- Recalibrate budgets and platform targets using the result.
Attribution models compared
- Last click — simple, biased toward retargeting and brand search
- Data-driven (platform) — better, but each platform only sees itself
- Multi-touch (third-party) — cross-channel, but limited by tracking gaps
- Incrementality tests — causal, but periodic and costly
- MMM — holistic, needs history and scale
How a geo holdout test works
- Choose comparable regions — similar size, seasonality and sales patterns.
- Split — run the channel in test regions, pause it in control regions.
- Run for enough weeks to detect a meaningful difference.
- Compare sales change in test vs control, adjusting for pre-test differences.
- Calculate incremental return — the sales the channel actually caused, per rupee.
Choosing the right measurement approach
Measurement by business type
| Business | Recommended approach |
|---|---|
| eCommerce | MER plus platform attribution plus periodic holdouts |
| Lead generation | CRM-based attribution with offline conversions |
| B2B | Multi-touch and self-reported attribution |
| Large advertisers | Media mix modelling plus experiments |
A worked example
A brand's attribution tools gave conflicting answers. We ran a geo-holdout test on Meta prospecting — pausing it in some regions — and measured the sales difference. The test showed the true incremental contribution, guiding the next quarter's budget.
Before attribution work
- Clean conversion data
- Spend data by channel
- Back-end revenue
- Willingness to run experiments
Why teams choose PMG for Attribution & Incrementality
- Measurement first. Unit economics and clean tracking come before any increase in spend.
- You own everything. Ad accounts, data and creative stay in your name; ad spend goes straight to the platforms.
- Senior hands on the account. The strategist you speak to is the person running it.
- Weekly transparency. Scorecard, change log and test board — every week.
- Short minimum terms. We keep clients with results, not contracts.
- India and international. One Mumbai team running campaigns across Indian cities and overseas markets, Mon–Sat, 7am–9pm IST.
Where we run Attribution & Incrementality
We manage Attribution & Incrementality for businesses across India — including Mumbai, Delhi NCR, Bengaluru, Pune, Hyderabad, Chennai, Kolkata and Ahmedabad — and for companies in the UAE, Saudi Arabia, the United States, the United Kingdom, Canada, Australia, Singapore and Germany. Campaigns are planned market by market, with language, currency, platform mix and privacy rules matched to each. See all Indian locations and international markets.
Free growth audit
Get a written audit for Attribution & Incrementality.
No sales call required. Tell us where to look; a strategist replies within working hours and sends the written audit within 72 hours.
- Tracking check
- Pixels, Conversions API, GA4 events and offline conversions — what is firing, what is double-counted, what is missing.
- Wasted spend
- Search terms, placements, audiences and overlap that cost money without producing qualified leads or sales.
- Conversion path
- Landing page speed, message match, form friction and follow-up time — where interested people drop off.
- 90-day priorities
- The three changes we would make first, with the metric each one should move.
Frequently asked questions
Is last-click attribution useless?
Not useless, but biased toward the bottom of the funnel. It over-credits retargeting and brand search and under-credits prospecting and video.
When is marketing mix modelling worth it?
Typically at larger multi-channel budgets with a couple of years of reliable data. Below that, holdout tests give clearer answers for less effort.
How often should we run incrementality tests?
For major channels at meaningful spend, once or twice a year — and before any large budget reallocation.
Which attribution model should we use?
There's no single correct model. Use platform attribution for in-channel decisions, blended metrics for overall health and incrementality tests to calibrate channel budgets.
What is incrementality?
The share of conversions that happened because of the marketing, which wouldn't have happened otherwise. See incrementality.
Can small businesses run incrementality tests?
Simple versions — pausing a channel for a period or in some regions and watching total sales — are possible for many businesses, with care about seasonality.
What is media mix modelling?
A statistical method that estimates each channel's contribution to sales using historical data.
How long does a geo-holdout test take?
Usually several weeks to cover a full buying cycle.
Is there a long contract?
No. Minimum terms are short, because we would rather keep clients with results than with contracts.
Will we own the ad accounts and data?
Yes. Everything runs in your own accounts with partner access for PMG, and you can see it all at any time.
What do you need from us to start?
Access to ad accounts and analytics, your margins and order values (or deal values), brand assets, and a decision-maker available for a weekly 30-minute call.
Works best alongside
More in this family
All servicesTerms used on this page
Talk to a strategist
Thirty minutes. Your numbers. A straight answer.
Book a strategy call with the people who'd actually run your account. We'll tell you what we'd do — or that you don't need us yet.