Scope: what this audit covers and what it does not
The question this audit answers is whether a programmatic buy is measured, safe and structured well enough to show where the money went and what it bought, and which weakness should be fixed first. It covers a running buy in Display & Video 360 (DV360Google's enterprise DSP.) with Floodlight measurement from Campaign Manager 360 (CM360Ad server and measurement platform.): how the account tree mirrors the plan, whether conversions are counted correctly, where impressions come from and whether deals spend, whether safety and suitability match the brand, how audiences and frequency are set, how bids and pacing are chosen, whether creatives serve, and whether delivery is verified and reconciled with invoices. It suits a team bringing a buy in-house or a new agency taking one over; the guide to moving a programmatic account to a new agency covers the handover itself, from seats and Floodlight history to deals and live flights.
It does not re-teach the work. Running a flight from set-up to close belongs to the programmatic campaign management playbook, and a build about to go live is checked with the DV360 campaign launch checklist. Commercial terms, contract compliance and the wider practice of independent audits belong to the media auditing and verification guide; this scorecard uses only its reconciliation and verification steps.
The site analytics the buy is compared against are scored by the analytics implementation audit. Paid social buys are scored by their own scorecards, such as the Meta Ads account audit, on the same scale, and every scorecard sits in the audit library.
Before you score: access and evidence
Ask for read access to the advertiser in Display & Video 360 and to the linked Campaign Manager 360 account, including partner and advertiser settings. Without the Floodlight configuration and the verification reports the measurement and safety rows stay provisional. Fix an evidence window that covers at least one closed billing period, so delivery can be compared with invoices, and keep the same window for every rescore.
Collect:
- read access to the advertiser, with partner and advertiser settings, campaigns, insertion orders and line items;
- the Floodlight configuration with activity groups, activities, counting methods and values;
- the deal list with each deal's status, assignment and spend, and the inventory and exclusion lists in use;
- verification and ViewabilityShare of measured ad impressions that had a real chance to be seen. reports, the brand safety and suitability settings, and the brand's written tolerance;
- audience lists with their sources and any third-party data fees;
- delivery reports, insertion orders and invoices for a closed period.
Write down what each flight was bought to do before you open its results. The CPM calculator helps compare what the buy paid per thousand impressions across sources once fees are separated, and the reach and frequency calculator helps check that the plan's frequency assumptions are consistent with its budget.
Criteria and scoring
Rate each row missing (absent or broken), partial (present but incomplete, inconsistent or out of date) or solid (matches the "what good looks like" column). The scale is your judgement against that column, not a number. Rows marked gate decide the band on their own when they score missing.
Structure
| Criterion | What good looks like | Score guidance |
|---|---|---|
| Insertion orders follow the plan | Campaigns and insertion orders carry the plan's goals, budgets and dates, and line items hold targeting and bids | Missing: one insertion order holding the whole year / Partial: mostly mapped, some budgets set in the wrong place / Solid: the tree reads like the plan |
| Naming and reportability | Names let delivery be reported by plan line, audience and creative without manual mapping | Missing: default or personal names / Partial: a convention, applied unevenly / Solid: one convention throughout |
The DV360 account structure guide explains what each level decides and which settings pass downwards.
Measurement
| Criterion | What good looks like | Score guidance |
|---|---|---|
| Floodlight activities (gate) | Activities match business outcomes, with the right counting method and values, deployed through the tag manager and tested | Missing: sales counted on page loads, or no activities for the outcome / Partial: activities exist, counting or values wrong / Solid: tested, with methods that match each outcome |
| One ad server for the buy | Creatives are served and counted from one ad server, so delivery and verification can be compared across buys | Missing: each buy counted by its own seller / Partial: one ad server for some buys / Solid: one ad server throughout |
Score activities against the Floodlight tags guide, which explains counter and sales activities and how the order identifier stops a refreshed confirmation page becoming a second order. An inherited configuration that scores missing or partial is fixed with the tutorial on cleaning up an inherited Floodlight configuration. The case for one ad server is made in the ad serving basics guide.
Inventory and safety
| Criterion | What good looks like | Score guidance |
|---|---|---|
| Deals assigned and spending | Deals are accepted, assigned to line items and checked for spend, and the open auction is used on purpose | Missing: deals accepted and never assigned / Partial: assigned, spend never checked / Solid: each deal checked and diagnosed |
| Brand safety and suitability (gate) | Exclusions, suitability settings and verification match the brand's written tolerance and are reviewed | Missing: no written tolerance, or defaults nobody chose / Partial: settings in place, tolerance unwritten or news blocked wholesale / Solid: written, matched and reviewed |
| Made-for-advertising exclusions | Inventory built for ad arbitrage is excluded by lists or pre-bid filters, with the reach trade-off understood | Missing: no defence / Partial: an exclusion list never refreshed / Solid: lists or vendor segments, refreshed from the placement report |
Deals are scored against the DV360 inventory and deals guide, which lists the reasons an accepted deal fails to spend; the deal types guide explains how a private marketplace (PMPAn invitation-only programmatic auction a publisher runs for selected buyers through a deal ID.) differs from the open auction. Safety and made-for-advertisingA site built to sell ad impressions rather than serve readers: thin or scraped content, a high ratio of ads to content and traffic bought rather than earned. inventory are scored against the brand safety and viewability guide, which separates the Brand SafetyControls that keep ads away from content that would harm the advertiser's reputation. floor from the brand's own suitability judgement.
Audiences and frequency
| Criterion | What good looks like | Score guidance |
|---|---|---|
| First-party audiences | Audiences from the business's own data and Floodlight are built, sized and used before bought segments | Missing: bought segments only / Partial: first-party lists built, rarely used / Solid: first-party lists lead |
| Third-party data fees | Bought segments are used only where their fee is justified by results the business can see | Missing: fees nobody has read / Partial: fees known, never compared / Solid: each segment justified or removed |
| Frequency caps planned | Caps are set from the plan at the right level and reviewed against reach reports, not left at defaults | Missing: no caps, or defaults / Partial: caps set once, never read / Solid: planned, set at the right level, reviewed |
Audiences and fees are scored against the DV360 audiences guide, and caps against the cross-platform reach and frequency guide, which explains where each FrequencyAverage times a person saw an ad. cap applies and what it can see. This scorecard sets no cap of its own.
Bidding and pacing
| Criterion | What good looks like | Score guidance |
|---|---|---|
| Bid strategy fit | Fixed or automated bidding matches the goal, with constraints that let automated bidding learn | Missing: strategy chosen by habit / Partial: right strategy, constraints too tight to learn / Solid: deliberate on both |
| Pacing and under-delivery | Pacing mode and budgets suit the flight, and under-delivery is diagnosed rather than ignored | Missing: under-delivery found at the end of the flight / Partial: noticed, not diagnosed / Solid: read during the flight and traced to its cause |
The DV360 bidding and pacing guide covers both rows, including the troubleshooting figures that show which setting is throttling a line item.
Creative
| Criterion | What good looks like | Score guidance |
|---|---|---|
| Formats and rotation | Creatives exist in the formats the line items buy and rotate on purpose | Missing: line items buying formats with no creative / Partial: formats covered, rotation by default / Solid: covered and rotated on purpose |
| Rejected creatives handled | Disapproved or rejected creatives are found and fixed quickly, with the policy reason recorded | Missing: rejections found by accident / Partial: fixed, reasons unrecorded / Solid: checked routinely, reasons recorded |
The DV360 creatives guide covers formats, rotation and review by Google and by each exchange.
Verification and reporting
| Criterion | What good looks like | Score guidance |
|---|---|---|
| Viewability used where it fits | Viewability is measured, unmeasurable impressions are tracked and viewability filters are used where they fit the goal | Missing: not measured / Partial: a viewable rate read without the measured share / Solid: measured, unmeasurable tracked, filters by goal |
| Delivery reconciled | Delivery, insertion orders and invoices are reconciled each period, and verification data is compared with platform reports | Missing: invoices paid unchecked / Partial: occasional checks / Solid: every period, gaps explained |
| Reports feed decisions | Scheduled reports include reach and frequency and safety views, and feed decisions about inventory, audiences and creative | Missing: delivery totals only / Partial: the views exist, nobody acts on them / Solid: reports end in decisions |
The viewability and attention guide sets out the standard definition the guides take from the Media Rating Council and why the measured share matters. Reconciliation is scored against the media auditing and verification guide, which also covers invalid traffic (IVTAd impressions and clicks not caused by genuine human interest: bots, fraud and accidents.), and reports against the DV360 reporting guide.
How to score and prioritise
Score the two gates first. If Floodlight miscounts, automated bidding has been pursuing the wrong figure and every efficiency row is provisional: a line item that looks strong may be counting refreshed pages. If safety and suitability settings do not match the brand's written tolerance, or no tolerance is written, the buy may be unsafe whatever it reports. Fix both before anything else.
Then rank the remaining weaknesses in this order:
- Money that buys nothing: deals that do not spend, made-for-advertising inventory, unjustified data fees and frequency left at defaults.
- Bidding and pacing, because they decide what the remaining budget buys.
- Reconciliation and verification, which must be in place before any budget increase.
- Naming and creative rotation polish last, unless they block reporting.
Within each group, rank by the spend the weakness touches. A fix on the insertion order carrying most of the budget outranks the same fix on a small test.
Reading the result: what to do with each band
Red means a gate scored missing. Hold budget increases and new flights, repair and test the Floodlight activities, and agree a written safety and suitability tolerance with the people who own the brand before the settings are changed. Rescore after a full billing period on the repaired set-up.
Amber means the gates hold at least partially but several other rows are partial or missing. This is the usual result. Work a short ordered list, starting with the money that buys nothing.
Green means the gates are solid and most rows are too. Move effort into testing inventory sources and audiences against outcomes, and rescore each quarter or when the buy changes hands.
From findings to a prioritised fix list
Write each partial or missing row as a finding in plain words: what you saw, the evidence and the spend it touches. Turn the findings into one fix list with a line per fix: the finding, the fix, the owner, the guide that explains it and the rescore date. Order the list by the priorities above, not by how easy each fix is.
Group the lines by who does the work. Floodlight fixes sit with whoever runs the tag manager and the ad server, safety tolerance with the brand owner and the trader together, deals, audiences, frequency and bidding with the trader, creatives with the people who build and traffic them, and reconciliation with finance and the trader. Present the band, the gate findings and the first few fixes to the budget holder, and keep the full scorecard as the working copy for the rescore. The academy programmatic operating plan capstone practises setting an audit cadence that scores the buy and feeds the fix list.
Illustrative example
A furniture rental service brings its buy in-house
Wolmerax Furniture Rental, an invented furniture rental service spending about £25,000 a month on programmatic, moves the buy in-house from an agency and audits the last three months first. Floodlight activities score missing: the sale activity has no order identifier, so every refresh of the confirmation page records another rental, and the account reports 640 rentals against 410 in the booking system. Brand safety scores partial: defaults are on, but no tolerance is written down and news is blocked wholesale. Deals score missing, because two private deals were accepted and never assigned to a line item. Frequency caps score missing on prospecting. Delivery reconciliation scores missing: invoices were never compared with delivery reports. The band is red. The fix list puts a sales activity with the order identifier first, a written suitability tolerance second, deal assignment and a planned prospecting cap third, and a monthly reconciliation fourth, and holds new flights. After a full billing period, reported rentals sit within a few of the booking system, the gates score solid and the band moves to amber.
Common mistakes
- Reading efficiency before measurement. A line item with a low cost per rental may be counting refreshed pages. Score Floodlight first.
- Treating safety defaults as a tolerance. The floor is on for everyone; suitability is the brand's own call. Score the gate on whether the tolerance is written and matched.
- Blocking news to feel safe. Wholesale keyword exclusions push budget towards low-quality inventory. Score suitability on judgement, not on how much is blocked.
- Assuming an accepted deal is spending. Acceptance is not assignment, and assignment is not delivery. Check spend by deal.
- Rewarding the cheapest viewable impression. Optimising for it alone can lead the buy to made-for-advertising sites. Score viewability on fit with the goal.
- Paying invoices without reconciling. Delivery, the insertion order and the invoice should agree. Score the row on what was compared, not on what was paid.
