Analytics and Measurement
Attribution and incrementality
One part of Analytics and Measurement, described on its own: work out what actually caused the sale, rather than which touchpoint was last.
What does an attribution and incrementality actually involve?
Work out what actually caused the sale, rather than which touchpoint was last. In practice that means an attribution approach chosen deliberately and applied consistently, rather than whichever the tool defaults to, and then the harder half: holdout testing where the spend justifies it, which is the only method that proves causation.
Why this work exists
The work has no natural owner, which is why it survives three reorganisations untouched.
It belongs to Analytics and Measurement and is described here because it is the piece clients recognise and ask for.
The nearest piece of work to it is deciding which numbers go in front of the board and which stay internal. Where that is already handled, this moves faster than the estimate; where it is not, the two get done together or neither of them holds.
What good looks like
Working well means the decisions downstream get made faster, because the argument about what is true has already been had.
The gap between that and where most accounts start is why this is scoped as attribution and incrementality rather than as a line inside something larger. Working out the cost of a bad data decision before it compounds is the usual second step once it is in place.
When this is the right work
It is worth doing when:
- Every channel claims the same conversion
- Budget arguments are decided by whoever has the most convincing dashboard
- Nobody has ever tested what happens when a channel is switched off
If none of those is true, something else is probably the constraint, and the Analytics and Measurement overview maps the alternatives. The usual neighbours are finding the segment where the average is hiding the story and getting a baseline recorded before the work starts.
What the work covers
Concretely, this covers:
- An attribution approach chosen deliberately and applied consistently, rather than whichever the tool defaults to. This is the piece a client can verify themselves, which matters more than it sounds.
- Holdout testing where the spend justifies it, which is the only method that proves causation, which clients tend to want evidence for, so it is documented as it goes.
- Marketing mix modelling where the business is large enough to support it, with the evidence kept so the decision can be defended later.
- Honest treatment of the channels that leave no trace at all, including agent-driven visits, revisited rather than finished, because the inputs keep moving.
How it runs
The shape of the work, stage by stage.
Before anything is changed
Nothing moves until the current position is written down and agreed. That takes days, not weeks, and it is the only defence against arguing about whether something improved.
The first pass
The obvious problems get fixed first, in one batch, so the picture clears before the harder judgement calls have to be made against a noisy background.
Where the judgement goes
What is left needs a decision rather than a fix. These get taken with whoever owns the consequence, not resolved quietly in a ticket.
What each part runs into
Each part meets a different obstacle depending on the market it is done in:
- An attribution approach chosen deliberately and applied consistently, rather than whichever the tool defaults to, where a Gurugram engagement contends with an incumbent whose position comes from presence rather than effort
- Holdout testing where the spend justifies it, which is the only method that proves causation – working in Hyderabad this means accounting for a buying pattern that resolves faster than the reporting cycle
- Marketing mix modelling where the business is large enough to support it, and in Indore the constraint on it is a sector mix that pulls the brief in several directions
- Honest treatment of the channels that leave no trace at all, including agent-driven visits, harder in London because of a seasonal peak that arrives before the content is indexed
What goes wrong first
Defending last click because it is simple. It systematically overcredits the bottom of the funnel and starves the work that creates demand.
The damage is less in the wasted work than in the conclusion it leads to.
How it is measured
Incremental revenue from holdout tests, and the gap between modelled and measured. Reported monthly against a baseline taken before anything changed.
The point of the number is to settle an argument, so it is chosen for being hard to dispute rather than for being flattering.
What you end up with
In concrete terms the output is an agreed attribution approach, a holdout test plan, and a reporting view everybody accepts.
Given with the raw working attached, so the conclusion can be re-derived.
How it differs by sector
How the sector changes it:
- Home and trade services, where emergency intent, service-area geography, and a phone that has to be answered. which moves this work earlier in the plan.
- In B2B services the binding constraint is that a buying committee, a long cycle, and attribution that has to survive both.
- SaaS and technology: the thing that changes the scope here is that product-led companies where the buyer self-educates and the trial is the sales call.
The industry pages carry the rest of that detail. Home and trade services and B2B services are the two where this piece of work most often decides the outcome.
Common questions
Which attribution model should we use?
The one everybody agrees to manage against, applied consistently. The choice matters less than the consistency and the honesty about its limits.
How do we measure channels with no referrer?
Holdouts where possible, branded search as a leading indicator, and mix modelling at scale. Not last click, which simply does not see them.
Is the output a report or a change?
Both, in that order. An agreed attribution approach, a holdout test plan, and a reporting view everybody accepts. The report exists so the change is accountable, not instead of it. AI Platform Optimization for Healthcare and clinics covers the setup, Analytics and Measurement for Healthcare and clinics the maintenance nobody budgets for.
How do we know this is the right piece of work?
The signals above are the test. If none of them describes your situation, something else is the constraint, and saying so is more useful than selling this. Two pages carry the detail this one summarises: Marketplace Marketing in Bengaluru, and then Paid Social and Meta Ads in Bengaluru.
What happens if it does not work?
We report that it did not, with the reason. Incremental revenue from holdout tests, and the gap between modelled and measured agreed before the work starts makes that an honest conversation rather than an argument. This sits between Paid social measurement and Video and short-form advertising, and borrows from both.
Ask about attribution and incrementality
One line on the goal and one on the obstacle. That is all the first exchange needs.
Ask what we would do first, and why
Bring the numbers you have. We will tell you which ones are measuring the wrong thing before we suggest anything new.