Why verification is separate from effectiveness
A delivery report is a seller's claim about what ran. Verification is the process of checking that claim against independent evidence. Effectiveness is the question of whether what ran changed anything. The three are often merged into a single campaign report, and the merge hides problems.
If delivery is overstated, every downstream metric is distorted. A cost per thousand calculated on inflated impressions looks better than it was; a reach estimate built on impressions that were never viewable looks larger than it was; an attribution model fed with fraudulent clicks credits the wrong channel. Effectiveness work cannot correct for errors it does not know about.
Separating the questions also clarifies ownership. Verification is the buyer's and finance team's responsibility, sometimes supported by an independent auditor: it is about getting what was paid for. Effectiveness belongs to analytics. The post-campaign analysis guide takes verified delivery as its input, and the media measurement hub shows how the three questions fit together. A plan whose delivery cannot be verified is a plan whose ReachUnique people who saw an ad. and FrequencyAverage times a person saw an ad. claims rest on trust alone.
Audience measurement by medium
Each medium has its own system for estimating who saw or heard the advertising, and each produces a different kind of evidence. The planner and buyer need to know what stands behind every figure.
| Medium | How audience is measured | Proof that the ad ran | Main limits |
|---|---|---|---|
| Linear television | Industry panels of households with metering equipment, run or commissioned by joint industry bodies | Broadcaster transmission logs and spot-level audience data | Panel size limits precision for small audiences; viewing outside the home harder to capture |
| Connected television and online video | Ad server and platform impression counts, sometimes combined with panel data | Ad server logs, third-party verification | Platform-reported figures; deduplication across platforms is hard |
| Radio and audio | Listening surveys or diaries for broadcast; download and impression counts for digital audio | Station logs; ad server data for digital | Survey recall; podcast downloads are not listens |
| Out-of-home | Audience models combining traffic counts, visibility and location data | Posting reports, photographs, digital screen play logs | Estimated rather than measured exposure |
| Audited circulation and readership surveys | Tear sheets or digital copies of the published ad | Readership estimates lag and rely on survey | |
| Display and social | Ad server counts, platform reports, third-party verification | Ad server logs, verification reports | Viewability, invalid traffic and walled-garden access |
| Sponsorship | Varies: broadcast audience, attendance, social metrics | Contractual deliverables, photographs, logs | Rarely a common currency |
The practical rule is to know, for every line, which source counts delivery, whether that source is independent of the seller, and whether its method is accredited. Lines that rely only on the seller's own count deserve more scrutiny than lines measured by an accredited industry system or an independent ad server. The guide to buying out-of-home advertising shows how to write posting reportsEvidence that offline ads ran as booked, such as posting photos, play and airing logs, transmission logs and tear sheets, checked before invoices are paid., photographs and play logs into the booking so each site can be checked before the invoice is approved.
Third-party verification for digital
Digital media can be checked at the level of the individual impression, and an industry of verification vendors exists to do it. Verification tags or integrations run alongside the ad and report on several things the seller's count alone cannot prove.
Viewability. Whether the ad had a chance to be seen, measured against the industry standard for display and video. The viewability and attention guide covers the definitions and what they do and do not say about attention.
Invalid traffic. Impressions and clicks generated by bots, data centres, hidden placements or other non-human or fraudulent sources. Verification vendors classify and filter invalid trafficAd impressions and clicks not caused by genuine human interest: bots, fraud and accidents., and the contract should say whether the advertiser pays for it.
Brand safety and suitability. Whether the ad appeared next to content the advertiser has excluded. The brand safety and viewability guide covers how exclusions are set and reported, and when a complaint about a placement arrives the tutorial on responding to a brand safety incident uses the verification record as the evidence.
Geography and audience. Whether impressions were served in the markets bought and, where data allows, to the audience described.
Independent ad serving is the companion to verification: an ad server the advertiser controls counts impressions and clicks independently of each seller, so delivery reports can be compared against a neutral source. The Campaign Manager reporting guide covers how that reporting is set up and read. Walled-garden platforms limit what third parties can measure inside them, and the advertiser should know which checks each platform permits before the buy, not after.
Reconciling delivery, invoices and the insertion order
Reconciliation is the routine, line-by-line check that what was bought, what ran and what is being charged agree. It is unglamorous and it is where most recoverable money is found.
The process compares three documents for every line. The insertion order states what was bought: volume, placement, dates, rate and terms. The delivery evidence states what ran, ideally from an independent source: ad server counts, verification reports, broadcaster logs, posting reports. The invoice states what is being charged. Differences fall into a few familiar groups: under-delivery against a guarantee, delivery outside the agreed placements or dates, invalid or non-viewable impressions the contract excludes, rates billed above the agreed rate, and fees that were not agreed.
Each difference is resolved under the insertion order's terms: a make-good, a credit, or a corrected invoice. Where the contract names a counting source, that source wins disputes; where it is silent, disputes become negotiations. The cost per thousand (CPMCost per 1,000 impressions.) recalculated on verified delivery is the honest figure for the post-campaign analysis. For the budget holder's side of the same check, the tutorial on purchase orders and invoice approval covers querying a media invoice before approving it, alongside agency, production and event invoices.
A bank reconciles a quarter of digital video
Corbelreach Building Society, an invented lender, reconciles a quarter of online video bought across four publishers for £210,000. The insertion orders guaranteed 14 million impressions at a £15 cost per thousand. Publisher reports claim 14.6 million. The independent ad server counts 13.9 million, and the verification vendor flags 3% of those as invalid traffic, leaving 13.5 million valid impressions. Under the contract, which names the ad server as the counting source and excludes invalid traffic, the society is billed for 13.5 million impressions, a saving of £16,500 against the publishers' invoices, and agrees make-goods for the remaining shortfall in the following month.
Independent media audits and what they check
A media audit is a periodic, independent review of the advertiser's media investment, usually carried out by a specialist firm with access to the agency's and sellers' records under the contract's audit clause. Where verification checks individual impressions, an audit checks the whole system.
Audits typically cover several areas. Price performance: whether the rates achieved were competitive against the auditor's benchmarks for similar buys in the same market and period. Delivery: whether the buys delivered against plan across the year, and whether shortfalls were recovered. Financial compliance: whether invoices reconcile to media owner invoices, whether rebatesA volume-based discount or payment from media owners to an agency or large advertiser, which the agency agreement should assign and report. and volume incentives were handled as the contract requires, and whether the agency acted as agent or principal where the contract specifies. Process: whether plans were signed off, buys matched plans and changes were documented. Contract compliance: whether the agency met its obligations on transparency, data access and reporting.
The value of an audit depends on the contract that permits it. An agency agreement without clear audit rights, access to media owner invoices and definitions of what counts as a rebate leaves the auditor with little to examine. Advertisers that rely on agencies for large offline buys tend to benefit most, because those buys carry the least impression-level evidence.
Choosing when to audit is part of the decision. Some advertisers commission an audit every year as a matter of governance; others audit after a change of agency, a large increase in spend or a pattern of unexplained reconciliation gaps. For a programmatic buy, the programmatic advertising audit scorecard gives an in-house team a scored view of its measurement, safety, deals and reconciliation to start from. Whichever trigger applies, the findings should be discussed with the agency before they are final, because many apparent discrepancies turn out to have contractual explanations, and the ones that do not become the basis for recovering money or changing terms at the next contract review.
Accreditation and standards
Measurement is only as credible as the method behind it, and accreditation is how the industry checks methods. In the United States, the Media Rating Council audits audience measurement and ad verification services against its published standards and accredits the specific products and metrics that pass. The MRC has worked with the IAB on industry guidelines for digital measurement, including viewability and invalid traffic, which verification vendors and platforms reference. Other markets have their own joint industry bodies that commission or oversee television, radio, print and out-of-home audience measurement. For an advertiser buying airtime for the first time, the guide to planning a first television campaign turns this into questions to settle before booking: which source the deal is traded on and what spot-level evidence will be supplied.
Three practical points follow. First, accreditation applies to specific measurements, not to a vendor as a whole, so check which metrics are covered. Second, a figure produced by an accredited method is still an estimate with error, particularly for small audiences or fine-grained breakdowns. Third, contracts should name accredited sources as the counting basis wherever they exist, so disputes are settled by method rather than by argument.
Common mistakes
- Paying on the seller's own count. Without an independent counting source in the contract, there is nothing to dispute with.
- Treating verification as optional. Invalid traffic and non-viewable impressions inflate every metric built on delivery.
- Skipping reconciliation. Invoices approved without checking against the insertion order and delivery evidence leave errors uncorrected.
- Assuming accreditation covers everything. It applies to specific metrics; check what is accredited.
- Signing agency contracts without audit rights. An audit is only as good as the access the contract grants.
- Confusing verified delivery with effect. A perfectly verified campaign may still have changed nothing; effectiveness needs its own evidence.
- Ignoring offline proof of delivery. Transmission logs, posting reports and tear sheets are the offline equivalent of an ad server count and should be requested as routinely.
