Recruitment advertising attribution: reconcile ad reports with ATS applications
Recruitment ad reports and ATS totals can differ because they count different events, time windows or records.
Recruitment ad reports and ATS totals can differ because they count different events, time windows or records. Reconcile the definitions and the application-to-ATS handoff before treating the difference as missing candidates or attributing hires to a particular campaign.
The useful starting point is a shared ledger that explains what each total represents. A number in an advertising report and a number in a recruiting system are not comparable until the team has agreed the stage, period and counting rule. Work through those definitions before using either total in a campaign evaluation.
Key facts
- Ad clicks, application starts, confirmed submissions and accepted ATS records are separate events.
- Accepted records still need the declared duplicate rule before they become a count of unique applications.
- An unresolved record needs an owner and investigation status, not an assumed cause.
- Recruiter qualification and hires require their own recorded outcomes and a clear cohort age.
- Attribution describes an observed source under a rule; reconciliation alone does not establish incremental hiring.
Which definitions should be aligned first?
Write down the event behind each report column. An ad click describes an interaction with an advertisement. An application start describes entry into the application process. A submission confirmation records completion of the submission step, while an accepted ATS record requires evidence from the receiving recruiting system. Keep each stage visible even when a report uses the broad label “applications”.
Next, define a unique application under the team’s stated duplicate rule. Do not silently switch from counting application records to counting people. A recruiter-qualified application also needs a recorded assessment against agreed requirements. A hire is a later outcome and must not be inferred from a submission, receipt or qualification decision.
Ask whether the advertising report includes modelled or view-through attribution, and record the configured attribution window. The question is a check, not an assumption that every platform report uses those methods. Keep the report’s definition alongside the receiving-system definition so a reviewer can see what is being compared.
Finally, align the reporting period and timestamp convention. Record the time zone and which event date determines inclusion. A submission-time view and a receipt-time view answer different questions unless the comparison accounts for that distinction. Retain the definitions with the ledger so the next reporting cycle can use the same rules.
What should a reconciliation ledger contain?
Use the following ledger as a copyable working structure. It is an editorial proposal for organising evidence, not a benchmark or a claim that every system exposes the same fields.
| Ledger item | Meaning | Check |
|---|---|---|
| Reported event | The stage counted in the ad report | Definition, attribution window and time zone |
| Submission | Confirmed candidate submission | Available application and job reference |
| ATS receipt | Record accepted in the recruiting system | Timestamp, job and mapped source |
| Duplicate | Record removed under the declared rule | Keep the rule and count visible |
| Unresolved | Expected record not yet reconciled | Owner and investigation status |
| Outcome | Recruiter review or later hire | Cohort age and recorded assessment |
Match records using the appropriate identifiers that are available within the authorised systems, such as application ID and job ID. Record the timestamp convention and source fields beside those references. A source label without a job reference or event definition should not be treated as a complete explanation of the journey.
Keep the matching method explicit. If an identifier is unavailable, record that limitation rather than implying that identity resolution is complete. The ledger should distinguish a confirmed match from a record still being investigated. Both belong in the account of what happened, but they carry different levels of certainty.
Work within the approved systems and access arrangements. Do not move personal candidate data to an unapproved tool to make the comparison easier. If reconciliation would require new tracking, treat that as a separate, scoped implementation decision with the appropriate owner. A reporting worksheet does not itself authorise collection or transfer of additional data.
How should discrepancies be investigated?
Investigate the difference in a declared order, keeping observations separate from possible explanations. This sequence is a practical review process; it does not establish which cause applies to a particular campaign.
- Check the period and time zone. Confirm that both reports use the intended date boundaries and identify the event timestamp used for each total.
- Check duplicates. Apply the declared rule consistently and show the removed count rather than simply replacing the original total.
- Check test records. Establish whether synthetic submissions are included and how the reporting definition treats them.
- Check failed delivery. Ask the authorised technical owner for handoff evidence before describing an unmatched record as a delivery failure.
- Check wrong-job mapping. Compare the intended vacancy reference with the receiving record rather than assuming a missing entry belongs elsewhere.
- Check the attribution definition. Review the source rule and configured window, including any modelled or view-through treatment identified earlier.
Keep an unresolved bucket throughout the investigation. Give each unresolved item a named owner and a recorded status. If the evidence does not yet explain the discrepancy, that is the result to report. Forcing every difference into a familiar explanation hides what the team still needs to verify.
Retain the starting totals as the investigation progresses. A correction should explain the movement between categories, so another reviewer can reconstruct the calculation and distinguish a newly confirmed match from a changed counting rule.
What does a reconciled example look like?
Illustrative ledger: 50 confirmed submission events in the defined period; 48 have matching accepted ATS records and 2 remain unresolved. Among the 48 accepted records, 2 are duplicate submissions under the stated rule, leaving 46 unique applications. Therefore 50 = 46 unique + 2 duplicates + 2 unresolved. Do not claim there were 50 unique candidates, 46 qualified people or two confirmed integration failures. No causal or vendor performance inference follows.
The first reconciliation is between confirmed events and receiving-system evidence: 48 accepted records plus 2 unresolved events accounts for all 50 confirmed submission events. The second applies the duplicate rule within those accepted records: 48 minus 2 duplicate submissions leaves 46 unique applications.
Those steps answer different questions. Receipt evidence establishes which records were accepted. The duplicate rule establishes which accepted records count towards the unique-application total. Neither step provides a recruiter assessment, so the ledger cannot turn the 46 unique applications into a qualification result.
Trace the two unresolved events with the authorised technical owner using the available references. Keep their status unresolved until the evidence supports a more specific classification. They are not automatically rejected candidates, and the arithmetic does not establish that an integration failed. This fictional example illustrates the counting method only; it is not an Adway campaign result or a vendor comparison.
Can reconciliation prove incremental hiring?
No. Matching an observed source to a record describes attribution under the chosen definition. It establishes the records a report is discussing, but it does not answer whether the advertising caused additional hires.
Incremental impact requires a suitable comparison design and mature outcomes. Keep that evaluation separate from the ledger, and state what the available evidence can support. An accepted application record is useful handoff evidence; a recorded hire is a downstream outcome. Neither becomes a causal finding merely because it carries a campaign source.
After establishing the ledger, use the guide to measuring ROI from AI recruiting software for the separate task of valuing outcomes. Carry the agreed definitions, duplicate rule and unresolved count into that discussion. This prevents an apparently precise value calculation from concealing uncertainty about which applications it includes.
How should the proposed Adway integration be checked?
Adway publishes an ATS and CRM integration directory. Buyers should verify the actual connection, field mapping and responsibilities for their configuration. A directory entry does not prove that the buyer’s particular mapping works, that a connection is bidirectional or that setup requires no configuration.
Request a job-specific synthetic submission and a receiving record containing the agreed fields. Compare the intended job, available application reference, timestamps and mapped source with the evidence returned by the recruiting system. Agree who owns the submission step, the receiving-system check and any unresolved investigation before accepting the handoff.
Use the demonstration to examine the proposed journey, not to infer universal integration support from a single example. Record what was actually checked and any remaining limitation. Keep synthetic test evidence separate from candidate results, and preserve the distinction between a successful receipt and a recruiter-qualified application.
Book a demo to walk through one approved role, its application journey and the handoff to your recruiting system.
Sources
- Adway integrations directory: first-party documentation for the directory statement. Inclusion does not establish tested mapping, universal support or guaranteed governance compliance.
- How to measure ROI from AI recruiting software: related guidance for valuing outcomes after defining the application ledger.
- The ledger, investigation sequence and numerical example are illustrative editorial guidance. They are not observed customer results or evidence of causal impact.
Start with one role. Verify the whole journey.
Frequently asked questions
Why can ad reports and ATS application totals differ?
They may count different stages, attribution windows, time zones or duplicate records. Investigate those definitions and the receiving-system evidence before concluding that applications were lost.
Is a submitted form the same as an accepted ATS record?
No. Submission and receiving-system acceptance are distinct events. Verify the handoff with an appropriate record or receipt.
Does an attributed hire prove the ad caused the hire?
No. Attribution assigns an observed source under a stated rule. Incremental impact is a different question requiring an appropriate evaluation design.
What should happen to unmatched records?
Keep them in an unresolved category with a named owner and investigation status. Do not silently discard them or infer a technical failure without evidence.
Related reading
-
Recruitment operations
Job category taxonomy for recruitment advertising reports
-
Which Recruitment Tech Vendors Are Best at Programmatic Job Advertising and Recruitment Marketing? A 2026 Buyer's Guide
-
Using applicant quality feedback to improve recruitment advertising
-
One recruitment advertising platform or specialist tools: how to choose the operating model
Want to see this in practice?