The quarterly review begins with a familiar problem. The media platform reports conversions. The ATS reports applicants. Recruiters report viable candidates. Finance reports spend. The totals do not match, and the meeting turns into a debate about whose dashboard is correct.

Reconciliation is not a reporting cleanup task

When systems count different events, the disagreement changes budget decisions, vendor evaluations and internal trust. It can also hide real performance. A campaign may be working while the integration undercounts it. A dashboard may look strong because it counts a lead as an application. Both situations require more than choosing the preferred number.

Create an event dictionary

Write one sentence for every event used in performance reporting. Define what has to happen, where it is recorded and which system is authoritative.

  • Click: a user activates a tracked link.
  • Apply start: the application workflow begins.
  • Application: all required fields are submitted under the employer’s definition.
  • Candidate record: a person exists in the ATS, regardless of creation method.
  • Qualified candidate: the person meets the agreed minimum standard.
  • Interview: a completed interview, not merely a scheduled event.
  • Hire: the organization’s chosen employment milestone.

Reconcile in sequence

  1. Compare clicks and sessions.
  2. Compare apply starts and completions.
  3. Compare completions and ATS-created records.
  4. Compare records and source assignments.
  5. Compare source assignments and recruiter dispositions.
  6. Compare dispositions and downstream outcomes.

The first point of material divergence is where the investigation should begin. Do not wait until the final hire count to discover that the application definition differed at the first stage.

A shared dashboard is not a shared truth unless the events underneath it are shared.

Use tolerances, not false precision

Perfect one-to-one counts may be impossible because of privacy settings, attribution windows, duplicate handling and timing. Define an acceptable variance for each stage and investigate when the difference exceeds it. The goal is a defensible explanation, not a performance theater of exactness.

Bring the explanation into the meeting

The review should show the agreed number, the variance, the likely cause, the business effect and the correction owner. That turns a dashboard dispute into an operating decision.

Keep the dictionary alive

Update definitions when a vendor changes a product, an ATS workflow changes or the organization adopts a new apply method. A metric can keep the same label while its underlying event changes. That is how trend lines become misleading without anyone intentionally manipulating them.