The advertising dashboard reports conversions. The sales team records customers. A business owner asks how many purchases happened because of the advertising. Those observations can lead to different answers. A useful discussion begins by naming the question each report is meant to answer.
What attribution explains
Attribution allocates recorded outcomes to sources using a chosen rule. It helps a team analyse journeys and channel activity. An outcome credited to a source does not, by itself, establish that the purchase would not have happened without the advertisement.
Incrementality concerns additional impact: what changed because of an intervention. Estimating it requires a defensible comparison. A simple before-and-after revenue comparison can mix advertising impact with seasonality, price changes or changes in sales activity.
What the new measurement tools address
In its 10 September 2026 measurement announcement, Google described developments in advertiser data foundations and causal measurement tools. The useful distinction is between reporting more observed conversions and assessing contribution.
The Meridian documentation describes an open marketing mix modelling framework. Its contribution estimates depend on model assumptions and prepared data. Access to a library does not make every spreadsheet sufficient for a reliable analysis.
Agree on the result first
What counts as a lead: a form submission, a suitable inquiry or an agreed project? How does the team treat another message from the same person? Who records payment and a refund? Advertising, analytics and sales need compatible answers.
Collect costs and outcomes for a comparable period. Inspect duplicates, missing events and the delay between an inquiry and its review. If one channel is assessed using every message while another is assessed using qualified prospects, the resulting lead costs describe different things. The guide to Telegram CPL and CAC explains how to keep cost definitions consistent.
The recording process should also preserve the original context. A request that looks unsuitable in a short spreadsheet label may make sense when a reviewer sees the conversation. Keep a way to resolve those disagreements instead of silently changing the definition between reports.
Plan the comparison before changing the campaign
Write the hypothesis, outcome, observation period and comparison method before launch. A controlled experiment needs comparable groups and rules that remain stable during the test. Sample size and uncertainty require their own assessment.
Here is an arithmetic illustration only: a comparable test group records 120 outcomes and a control group records 100. The difference is 20. Without information about variation, group composition and other interventions, the difference is not yet proof of advertising impact.
If the team cannot support that kind of experiment yet, an honest operational report remains useful. It can show where requests are being lost and how quickly the team responds. Its value does not depend on presenting it as a causal study.
Our editorial view
For a small business, a consistent table of qualified inquiries, owners and follow-up outcomes is often a sensible first improvement. Choose a contribution assessment method after agreeing on the source data. A complex model will not resolve two departments using different definitions of a customer.
Use the lead qualification checklist to align those definitions. Then consider the tools discussed in our review of marketing changes against the question your team actually needs to answer.
Sources
- Google: data and measurement updates, 10 September 2026.
- Google Meridian: purpose and modelling assumptions.
Sources checked on 23 October 2026. The numerical example is illustrative and does not describe a KUZKO experiment.