One action, different views
A website conversion is an action observed on your website or application experience, such as an application start or a confirmed submission. An ad conversion is an action an advertising platform attributes to one of its ad interactions under that platform's reporting rules. The two can describe the same underlying event, yet their totals may differ. They answer related but distinct questions: what happened in the digital experience, and what an ad platform credits to its campaigns.
Imagine an auto loan applicant who first clicks a Google Search ad, later sees a Meta ad, and finally returns directly to finish the vendor-hosted form. The loan origination system has one submitted application ID. Analytics may record one submission if the handoff is instrumented, while Google and Meta may each credit that outcome under their own eligibility and attribution settings. Adding both ad reports would overstate the number of unique loan applications.
Why the totals diverge
Differences can come from conversion definitions, timing, attribution windows, click versus view credit, identity matching, consent, and time zones. A website event may count every confirmed submission, while an ad report includes only submissions it can associate with eligible ad activity. Conversely, a poorly configured ad conversion may count an Apply button click that never became a submitted application. A repeated event can inflate either system if it fires twice.
A vendor handoff creates a particularly common gap. The credit union site may register an Apply click, while the application platform confirms a smaller number of submissions. If source data does not carry into intake, the approved loan may have an application ID but no campaign attached. Treat the Apply click as an intent metric. Do not relabel it a completed application or infer approval attribution from it.
Assign a job to each number
Use website and intake data to understand the full journey and reconcile actual application activity. Use ad platform conversions to guide campaign bidding and optimization within that platform, provided the conversion action represents the intended milestone. An ad platform's number is useful for its own decisions, but it should not automatically become the institution's unduplicated total across channels.
For a credit union scorecard, distinguish total confirmed applications in the intake system from the subset with a known campaign source. Match approvals to submissions by application ID in an authorized reporting process. Label application conversion rate as submissions divided by the selected visit or click base; label approved conversion rate as approved applications divided by submitted applications. Keep product, date window, and decision lag visible so auto loan and checking figures are not compared as though they describe the same journey.
Build a reconciliation habit
Reconcile by product each month: vendor-confirmed submissions, analytics submission events, known-source applications, and conversions credited in each ad platform. Investigate a sudden gap after a vendor release, a new application URL, or a campaign tagging change. Do not expect the counts to be identical; the point is to know which system is the record of submitted applications and why an ad platform credits only a portion of them.
The strongest setup lets each system do its job while connecting campaign source, application ID, and decision data where permitted. That creates a more useful answer than forcing every dashboard to display the same number.
What to ask when numbers conflict
Before changing budgets, ask whether the reports cover the same product, event, date range, and time zone. Check whether one counts every submission and another counts only submissions attributed to ads. Then review attribution settings and whether view-based credit is included. If a platform has several conversion actions, confirm which are included in its headline total and in bidding.
Keep a short metric dictionary beside the dashboard. It should identify each source, event trigger, attribution rule, and known limitation. That shared reference turns a recurring argument about mismatched totals into a productive discussion about what each measure can and cannot answer.