Measure the journey you can see. Do not turn a retailer referral into a sale you have not observed.
A product campaign sends more visitors to your website. Retailer clicks rise. The team wants to increase spend—but the purchase happens in the retailer’s system, and you cannot see whether those visits became sales.
That is an evidence gap, not a reason to abandon measurement. The task is to separate what the data shows from what the business still needs to know.
Name the action before interpreting the number
Suppose a visitor reads a product page, selects a retailer and leaves your site. Each action answers a different question. None, by itself, records a completed purchase.
| Evidence you have | What it can tell you | What it cannot establish |
|---|---|---|
| Product-page activity | Which information visitors explore. | That they intend to buy. |
| Recorded retailer referral | Which purchase destination they choose. | That the product was available or a sale occurred. |
| Partner transaction records | Which sales the partner reports within the data’s scope. | That marketing caused those sales. |
Illustrative evidence map, not a measured funnel. Transaction coverage and matching may be incomplete.
For the budget decision, rising referrals are useful evidence of movement toward a retailer. They leave the sales question unresolved. Naming that boundary helps the business choose the next investigation instead of filling the gap with an assumed sales value.
Check whether the journey changed or the recording did
Before comparing products or campaigns, test the events. Does the same button mean the same thing across sites? Is the product and destination recorded clearly? Can one interaction create duplicate events?
Now return to the campaign. If one retailer’s referrals fall after a site update, inspect the link and recording before deciding interest has weakened. If clicks are recorded but land on the wrong product, the customer route needs repair.
| What you observe | What to inspect | Action the evidence may support |
|---|---|---|
| Referrals drop after a site change | The button, destination and event recording. | Repair the route or tracking before changing the campaign. |
| Visitors reach an unsuitable destination | Product match, availability and the page the buyer sees. | Correct the link or work with the retailer on the destination. |
| Referrals rise; sales remain unknown | Whether usable partner transaction data is available. | Close the sales-data gap before making a sales claim. |
Investigation prompts, not confirmed explanations. Check representative journeys and relevant records.
These checks can produce practical improvements even when sales data is unavailable. They do not establish how much revenue the campaign generated.
Ask for sales data with a clear purpose
If the retailer can provide transaction records, establish what is included before combining them with your reports. Check products, dates, geography, returns and reporting delays. Agree what information can be shared and used.
A partial sales feed can be useful, but its limits need to remain visible. Sales from one retailer cannot automatically stand for every destination. A total increase may also reflect price, distribution or stock availability.
For this campaign, partner records could help determine whether reported sales rose alongside referrals. That is stronger commercial context. It still does not isolate what would have happened without the campaign.
Match the strength of the evidence to the investment decision
McKinsey’s full-funnel marketing guidance discusses controlled incrementality tests as a way to check performance beyond attribution. For our campaign, the practical implication is to compare observable sales under a defined marketing change with a credible comparison.
Agree the outcome and design before starting. The comparison needs to account for the situation, available data and uncertainty. A test of referral clicks cannot answer a question about sales.
If a credible sales test is not feasible, keep that question open. Use journey evidence to improve the handoff, use partner records within their coverage, and avoid presenting either as proof of incremental revenue.
The next measurement investment should resolve a consequential uncertainty. Collecting another event is only useful if it helps the business make a better decision.