How many additional sales does advertising generate? Planning a contribution study

An advertising report shows sales, but some buyers might have arrived without an ad: through a search for the company name, a recommendation or a repeat enquiry. The question “how many sales were attributed to advertising?” differs from “how many additional sales did it generate?”

Measuring the incremental contribution helps you decide whether to keep a channel or increase spending. Define the comparison, outcome and limits of interpretation before the study begins.

Start with a specific hypothesis: for example, whether a selected campaign generates additional paid orders in a particular country. Avoid combining every market, service and advertising system in one study.

Record:

• which campaigns are included;
• the audience and market being studied;
• which action counts as an outcome;
• how much uplift would be useful to the business;
• the decision that will follow the assessment.

If the goal is new customers, track repeat purchases separately. If the goal is sales, opening a form does not replace payment. For a long B2B sales cycle, you can study qualified enquiries, but the conclusion will concern enquiries rather than sales that have yet to happen.

Attribution assigns outcomes to interactions according to selected rules. One order may appear in several systems' reports, so their figures cannot be added together as independent sales. See the article on differences between Google Ads, Meta Ads and GA4 data.

Measuring incremental impact requires control conditions: what happened to a comparable group without the advertising being studied. Google describes this approach in its Conversion Lift documentation.

Campaigns that address existing demand are particularly useful candidates for a check. One example is advertising on your own company name.

A user experiment. With a properly randomised split, part of the audience is eligible to receive the advertising being studied, while the control group is excluded. Assess data sufficiency and the ability to maintain this separation before launch.

Built-in studies are not available to every business. Google, for example, specifies account, campaign, budget and conversion requirements. Check them for your particular launch using the Conversion Lift requirements.

A geographical study. Regions are assigned to test and control groups after reviewing demand and sales trends. Simply comparing a capital city with a small town is unreliable: differences may predate the advertising. Account for baseline trends, movement of buyers between regions and other channels.

A comparison over time. Switching advertising on and off can provide observations, but seasonality, holidays and purchase delays complicate interpretation. Without an appropriate study design, treat this comparison as an indication rather than proven causal impact.

Choose a primary metric and collect it consistently in both groups. For a store, this could be unique paid orders adjusted for cancellations and returns; for a service, qualified enquiries assessed using agreed criteria.

Do not count only clicks on the test advertising. Its effect may appear in a later direct visit or another channel. The measurement method needs to capture the selected outcome in both groups.

Record changes to prices, availability, the website, enquiry handling and other campaigns during the study. Define its duration, stopping rules and the waiting period for late sales in advance. Ending early at the first attractive result makes the assessment unreliable.

Suppose a properly designed experiment produces 120 and 100 paid orders from two comparable groups of equal size. The estimated incremental outcome for the test group is 20 orders.

If spending on the advertising being studied for that group was €1,000, the estimated cost per incremental order is 1,000 ÷ 20 = €50.

This is hypothetical arithmetic, not proof of impact from two numbers. Assess uncertainty and the quality of the comparison. Raw differences in order counts cannot be used when group sizes differ. Also, compare €50 with an order's contribution after variable costs rather than its revenue.

1. Decision: what will change in the budget or advertising structure.
2. Hypothesis: the campaign and incremental outcome being studied.
3. Control: who does not receive the advertising being studied and how comparability is maintained.
4. Data: sources, definitions, duplicates, cancellations and late sales.
5. Timing: observation period and time allowed for deals to close.
6. Limitations: other channels, seasonality, audience crossover and business changes.
7. Conclusion: estimated impact, its uncertainty and the next action.

With limited data, the result may remain inconclusive. This does not prove there is no impact or justify declaring advertising profitable. Record which question was assessed and what the next decision requires.

Study feasibility and measurement preparation can be discussed through our advertising setup and management offer. Prepare the campaigns, sales history, selected market and the decision you want to make.