A campaign reports almost 600 conversions at roughly €0.63 each. Advertising wants to scale; sales should first ask how many represent people worth following up. An impressive event count is not yet an answer about business demand.
The tension appears in Arnau Vendrell’s iSocialWeb review, published on 29 June 2026. It explicitly distinguishes platform conversions from qualified leads. This is not a GoogleAsk project. We use the measurement gap to develop a diagnostic method, not to adopt the author’s customer results as our own.
Two systems can mean different things by success
The public review describes connecting acquisition with nurturing and qualification, and reports close rate changing from 3.17% to 5.73%. Those are the author’s reported project outcomes, not a guaranteed uplift or proof that one setting caused the difference.
The operational question is simpler: if the platform observes a form event while sales evaluates product fit, geography and purchasing conditions, they are judging different outcomes. One report can look excellent while the other contains little quotable demand.
First investigation: follow an event to a real record
Before discussing a bidding change, select a period with complete follow-up records. Compare conversion actions, stored website submissions and sales records. If they lack a shared lead identifier, repair that gap first.
For each sample, ask whether the event was a click or a confirmed receipt; whether it duplicated another request; whether sales received it; whether it fitted the service; and whether anyone followed up. This separates technical, targeting and operational problems instead of passing responsibility between teams.
| Finding | Initial location | First action |
|---|---|---|
| Platform event, no stored request | Event definition or submission chain | Trigger after confirmed receipt |
| Stored request, no sales receipt | Notification and assignment | Inspect handoff and ownership |
| Mostly out-of-scope demand | Traffic and page promise | Review queries, locations and service limits |
| Suitable request, no timely reply | Sales workflow | Define response and follow-up ownership |
Second investigation: separate reasons for rejection
Distinguish tests or spam, incomplete information, out-of-scope requests and qualified leads awaiting contact. Missing quantity may require a follow-up; an unsupported product is a different problem. Labelling both simply “invalid” obscures what advertising should change.
Sales needs to agree written, reviewable criteria and a treatment for uncertainty. If different reviewers classify the same request inconsistently, automating the return of their judgements only accelerates inconsistent data.
The repair order
Website records first. A successful submission requires server receipt and storage, not a click. Test validation errors, repeated clicks and thank-you refreshes. Assign a stable lead_id to a real request.
Sales records second. Preserve state, occurrence time, reviewer and reason. A controlled spreadsheet can be a starting point without a CRM, provided someone maintains it continuously rather than reconstructing a month from memory.
Business-outcome imports last. Return verified stages through the supported integration. Google’s enhanced conversions for leads documentation provides the starting point. Interfaces changed during 2026; check the current Data Manager route and the account’s capabilities instead of reusing an old implementation uncritically.
Why closed sales are not automatically the only target
Long sales cycles and low counts can delay feedback. Establish whether qualified leads are consistently defined, reviewed promptly and imported reliably before deciding which stage to optimise.
When observing a new action, inspect campaign goals as well. A submission, qualification and sale may belong to one person, not three new customers. Google’s implementation checklist helps organise the configuration review; upload acceptance alone is incomplete verification.
How to accept the repaired workflow
Walk a clearly marked test lead through submission, storage, assignment, review and import. Confirm that failed submissions and repeated state changes do not inflate results. Then sample anonymised real records and compare source, occurrence time and import outcome.
Acceptance, matching, attribution and reporting are distinct checkpoints. Preserve diagnostics when a stage fails and reconcile identifiers before retrying. Re-uploading a batch merely because a report has not refreshed can create another problem.
Read outcomes without returning to vanity metrics
Compare submission cost, qualified-lead cost, quotation rate and mature sales outcomes. Give cohorts comparable follow-up time. Yesterday’s enquiry cannot fairly compete with a month-old lead on close rate.
We do not have the original project’s complete underlying records and cannot independently audit its uplift. The transferable lesson is the measurement boundary and repair order: connect advertising, the website and sales through the same lead before deciding what the budget bought.
If the storefront itself keeps visitors waiting, start with the image-priority case. For account and landing-page coordination, use the search advertising scope to assign work rather than attributing every issue to bidding.
Questions after the case
Should cheap clicks be excluded when lead quality is poor?
Price alone does not establish quality. Review intent, fit and follow-up before deciding which traffic to restrict.
What if historic sales states were never recorded?
Analyse what exists and state the gap. Establish reliable records now rather than manufacturing past classifications.
