Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Programmatic Advertising › Intermediate

Capstone: Auditing and Optimising a Programmatic Campaign

The closing project of the intermediate level. You audit an invented cruise line's live programmatic campaign across deals, brand safety, viewability, line item structure, bidding, pacing, Floodlight, verification and frequency, separate measurement faults from performance problems, and write a prioritised plan for the rest of the flight with an incrementality test to settle what the programmatic budget adds.
Capstone projectIntermediate levelModule 3: Measurement and optimisation7 minUpdated

Brief

Meridian Isle Cruises is an invented small cruise line selling river and coastal voyages to customers across Great Britain. Its agency built a DV360 campaign for the spring booking season, alongside a direct video sponsorship with a newspaper's travel section, all served through one Campaign Manager 360 advertiser. The flight is past its midpoint. The agency's reports say bookings are strong; the finance team says the booking system does not agree, and the newspaper's invoice is higher than expected.

The marketing manager wants to know what is actually happening, what should change for the rest of the flight, and whether programmatic earns its budget at all. The programmatic hub is the map for the audit: every stage of the buying chain it describes is a place this campaign can go wrong.

Inputs and constraints

Illustrative example — not a real result

Meridian Isle Cruises: plan and delivery after five of eight weeks

Flight budget: £192,000. Planned to date: £120,000. Finance's maximum cost per booking: £400.

Line item or buyInventoryAudienceBidding and pacingPlanned to dateSpent to dateImpressionsMeasurableViewable on measurable
Prospecting displayOpen auctionIn-market cruises plus third-party "affluent retirees" segment (£1.00 CPM data fee)Beat target CPA £250; shares an insertion order with the PMP, unlimited line item budgets, even flight pacing£45,000£38,0009,500,00070%48%
Travel publishersPrivate marketplaceNoneFixed £9 CPM; same insertion order£15,000£6,000670,00094%72%
YouTube prospectingYouTube via DV360Affinity travelMaximise conversions, no ceiling; own insertion order, as soon as possible pacing; flight budget £40,000£25,000£30,0002,000,00097%86%
RemarketingOpen auctionBrochure downloaders, 19,000 matchableFixed £6 CPM; no caps£10,000£7,2001,200,00085%66%
Newspaper videoDirect, CM360 tagsNoneInsertion order at £20 CPM£25,000Invoiced £28,0001,250,000 counted by CM360; 1,400,000 claimed96%81%

No campaign-level frequency cap. DV360 reach report for the campaign: 2,100,000 reached, average frequency 6.1.

Illustrative example — not a real result

Meridian Isle Cruises: measurement, verification and constraints

  • Floodlight "Booking confirmed": sales activity, transactions counting, ord set to a random number, cost carrying the deposit rather than the voyage price, view-through window 30 days.
  • CM360 Floodlight report: 1,020 bookings in five weeks. Booking system: 610 bookings from all channels.
  • DV360 attributed bookings, click-through / view-through: prospecting display 6 / 140; travel publishers 4 / 22; YouTube 9 / 95; remarketing 25 / 310.
  • Verification vendor: 18% of prospecting display impressions served on sites on its made-for-advertising list, with an 85% viewable rate on those sites; no exclusion list is applied. Invalid traffic on the newspaper buy: 2%.
  • Newspaper contract: CM360 is the counting source and invalid traffic is not billable.
  • The booking system records bookings by postcode area; voyages sell nationally.
  • Constraints: total budget unchanged; one agency trader with limited hours; no Floodlight change may break the booking flow; the plan must be explainable to finance.

Task

Produce one document with the sections below, in this order. Use the site's tools, templates and checklists as named, state expected effects as a direction with a way to check them, and invent no uplifts.

  1. Fix the measurement first. Using the Floodlight tags guide and the tracking verification checklist, list every fault in the booking activity, its effect on reported bookings and on the automated strategies, and the fix. Say why no bidding verdict can be trusted until it is done.
  2. Pull the right reports. Name the DV360 report types you would run and at what level, following the DV360 reporting guide: delivery and cost with data fees separated, the reach report, Active View by inventory source and the verification report.
  3. Audit the deal mix and inventory quality. Judge each buy on cost, viewable delivery and placement quality. Cover the made-for-advertising problem with the brand safety and viewability guide, and read measurable and viewable rates together as the viewability and attention guide describes.
  4. Audit structure, bidding and pacing. Explain why the private marketplace is under-delivering, why YouTube is ahead of plan and what happens next, and whether the target strategy on prospecting can work on the evidence. Use the CPM calculator to check what each remaining budget buys.
  5. Audit frequency. Calculate remarketing exposure per identifier and set caps at campaign and line item level, using the cross-platform reach and frequency guide and the reach and frequency calculator.
  6. Reconcile the newspaper invoice against the contract and say what Meridian Isle should pay.
  7. Write the prioritised plan for the remaining weeks: each finding with evidence, affected spend, fix, owner and how it will be verified, measurement first. Record the revised split for the rest of the flight in the media plan template, and run relaunched line items through the campaign launch QA checklist.
  8. Design the incrementality test. Using the incrementality testing guide and the experiment hypothesis template, specify a geo test of programmatic prospecting: regions, outcome, duration, minimum detectable effect set in advance and how the result will change next season's budget.

Evaluation criteria

CriterionWhat a strong answer showsCommon gap
Measurement firstRandom ord, deposit value and the long view-through window found and fixed before any bidding verdictJudging the target CPA on inflated, view-through bookings
ReportingNamed report types at the right level, data fees separated from media, reach never summed across rowsQuoting a blended cost that hides the data fee
Inventory qualityMade-for-advertising sites excluded; viewable rate read with measurable rate; deal value judged on viewable costPraising the high viewable rate on the worst sites
Structure and pacingPMP starvation traced to the shared insertion order; YouTube's early exhaustion predicted and fixedRaising the PMP bid, which does not touch the cause
BiddingProspecting moved to a fixed bid or a volume strategy with a ceiling until clean conversions accumulateLowering the target on a strategy fed by bad data
FrequencyRemarketing exposure calculated, caps at campaign and line item level, a plan to enlarge the listLine item caps only
VerificationInvoice settled on CM360's count less invalid traffic, as the contract statesPaying the publisher's count or splitting the difference
IncrementalityA geo design with matched regions, a pre-set minimum detectable effect and a decision ruleTreating attributed bookings as proof programmatic works
CommunicationA short summary finance can follow, with directions of effect rather than promised upliftsInvented percentage improvements

To check your arithmetic, the key figures are below. Work them out yourself first.

Illustrative example — not a real result

Reference figures for self-checking

  • Floodlight overcount: 1,020 recorded against 610 bookings from all channels, so the activity overstates bookings by at least 410 before attribution is considered.
  • Prospecting display data fee: 9,500,000 × £1.00 ÷ 1,000 = £9,500 on top of £38,000 media. Viewable share of served impressions: 70% × 48% ≈ 34%. Click-through cost per booking on media plus data: £47,500 ÷ 6 ≈ £7,917.
  • Travel publishers: £6,000 of £15,000 planned, 40% delivery, while the open-auction sibling took the shared allowance.
  • YouTube: £30,000 of a £40,000 flight budget spent in five weeks, leaving £10,000 for three weeks.
  • Remarketing: 1,200,000 ÷ 19,000 ≈ 63 impressions per identifier in five weeks, about 13 a week.
  • Newspaper: billable 1,250,000 × 98% = 1,225,000 impressions; 1,225,000 × £20 ÷ 1,000 = £24,500 against an invoice of £28,000, a difference of £3,500.

A strong plan will refuse to judge programmatic on attributed bookings at all until the Floodlight fix has run for a clean period, and will let the incrementality test, not the DV360 interface, answer the marketing manager's last question.

Submission checklist

0/12

When you have finished, compare your plan with the DV360 line item planning exercise, the Floodlight specification exercise and the frequency rebalancing exercise, which practise its parts in isolation, then return to the intermediate programmatic level to mark the level complete. The Incrementality test is the part to carry into next season's plan.

Pages this lesson applies

Terms used on this page

Also related

Related