CMO playbook & advanced tactics: mastering GRP, TRP & offline metrics

Your agency brings two numbers to the annual planning meeting. The television line arrives with a panel-audited delivery report, a post-campaign reconciliation, and make-goods if the buy underdelivers. The digital video line arrives with a reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → figure generated, counted and reported by the company selling the inventory. Both get pasted into the same reach chart. That act of pasting, treating an audited currency and a self-reported claim as the same unit, is the most expensive unexamined decision in a large advertiser's budget. Put 40% of a $500 million plan behind numbers overstated by a fifth and you have mis-priced $40 million of reach, with nothing in the monthly dashboard to flag it.

What you are actually arbitrating

Take the vocabulary as given: the impacts, GRP, TRP, reach and frequency definitions the foundations lesson builds, and the reach-frequency arithmetic and impressionimpressionThe total number of times an ad or piece of content is displayed, regardless of clicks. Each display counts as one impression, even to the same person.View full definition → conversion the methodology lesson sets out. The CMO question sits a layer above the maths. Two currencies claim to measure the same human attention. One rests on a probability panel with published methodology and third-party accreditation. The other rests on logged-in device data with a methodology its owner can revise between quarters, and no external referee. Your job is not to declare a winner. It is to set an exchange rate between them, write it into contracts, and be able to defend the rate when the CFO asks how you arrived at it.

Sub-concept 1: the audit premium, and the week the audit failed

What accreditation buys is narrow and valuable: a documented sample, a known error margin, an outside body checking the arithmetic, and a commercial right to compensation when delivery misses. BARB in the UK runs on the order of 5,000 metered homes; Nielsen's US national television panel runs into the tens of thousands. Neither is large by digital standards, and that is the point. A small measured sample with published confidence intervals is auditable in a way a census of self-reported impressions is not.

The failure mode is that the audit itself can lapse. In September 2021 the Media Rating Council suspended accreditation of Nielsen's national television ratings after Nielsen was found to have undercounted viewing during the pandemic, when in-home panel maintenance stopped. Accreditation was restored in 2023. In between, US networks ran their own certification processes and advertisers began transacting against alternative currencies from VideoAmp, Comscore and iSpot. The lesson for a CMO is contractual, not technical: your protection comes from the audit, not from the vendor's name. Write clauses that reference accreditation status and specify what happens to pricing if it is withdrawn mid-year.

Sub-concept 2: pricing the discount on unaudited reach

The discount you apply to platform-reported reach is a risk price, not a moral verdict. The historical record gives you the range. In September 2016 Facebook disclosed that it had overstated average video view time for roughly two years by excluding views under three seconds from the average, an error large enough that corrected figures came in dramatically lower. In 2017 its own planning tool showed potential reach among US 18 to 34 year olds above the census population for that age band. Industry work by the ANA with White Ops put global bot-fraud losses in the multiple billions of dollars a year. None of these were fraud by the buyer's definition. All of them moved the number in the same direction.

So apply a haircut and make it explicit. An illustration: your plan claims 65% four-plus reach, 45 points of it audited panel delivery, 20 points contributed by unaudited digital video. Discount the unaudited contribution by a third and your four-plus reach is nearer 58%. In most brand response models that is the difference between clearing an effective frequency threshold and buying an expensive near-miss. The budget has not changed. The decision has.

Sub-concept 3: share of voiceshare of voiceYour brand's share of total advertising or conversation volume in your category, measured against competitors over a defined period.View full definition →, the one argument that survives the board

Share of voice is your GRP delivery as a percentage of total category GRPs. The Binet and Field analysis of the IPA Databank, covering more than 700 case studies, established that brands holding share of voice above their market sharemarket shareThe percentage of total industry sales your company captures in a given period. It measures competitive position relative to rivals in a defined market.View full definition → (excess share of voice) grow, with roughly 0.5 points of annual share growth per 10 points of ESOV. That relationship is quotable in a board meeting precisely because it is computed in an audited currency that measures your competitors on the same basis. No platform will report a rival's reach to you.

The edge case matters. ESOV is a category average, and it breaks in categories with heavy price promotion, where share moves with the promotional calendar faster than with salience. It also flatters you if you compute it on television GRPs alone in a category where a challenger has gone digital-heavy: your ESOV looks positive while your actual voice share is shrinking.

Sub-concept 4: mix modelling as tie-breaker, and where it quietly lies

Marketing mix modelling is the tool that prices the two currencies against a single outcome: revenue. Nielsen and Analytic Partners both sell mix modelling services, so read their case studies as marketing. The structural weakness is not the regression, it is the data. When television and digital video are flighted together week after week, the model cannot separate them, and the coefficient it hands you is an artefact of collinearity. Channels bought continuously with no variation look efficient because they correlate with demand they did not create, which is how always-on search takes credit for television's work. Meta open-sourced its own mix-modelling package, Robyn, and Google has released one too. Useful code, written by the parties being measured.

The correction is cheap by comparison: run at least one geographic holdout or blackout test a year per major channel and force the model's decomposition to reconcile with the experiment. If it cannot, the model is a narrative, not evidence.

Real-world cases

Case 1: Unilever, spending on the order of €7 billion a year on brand and marketing, used that scale as the lever. In February 2018 at the IAB's leadership meeting, then CMO Keith Weed said Unilever would not invest in platforms that failed on transparency, measurement and content responsibility, and made third-party verification a condition of spend. Around the same programme Unilever cut its number of ads by about 30% and halved its agency roster, reporting savings in the hundreds of millions of euros. The instructive part is sequencing: the demand for verified measurement came before the reallocation, not after. An advertiser that reallocates first has no exchange rate to negotiate with.

Case 2: CocaCocaCustomer Acquisition Cost: total sales and marketing spend divided by the number of new customers acquired over the same period.View full definition →-Cola pulled roughly a billion dollars of marketing spend out of 2020 when consumption in restaurants, cinemas and stadiums collapsed. CEO James Quincey said afterwards that the pause had exposed spending that was not working, and the company rebuilt investment above pre-pandemic levels while consolidating global creative with WPP at the end of 2021 and running fewer, larger platforms such as Real Magic. That decision only holds up because there was an audited baseline to return to. A brand that cuts its measured offline weight and then cannot say what it used to deliver has lost the ability to argue for the money back.

CMO action items

• Publish an internal exchange rate. One number, reviewed twice a year: the discount applied to unverified reach claims before they enter any combined reach chart. Give it to your agency in writing so plans arrive pre-adjusted.
• Put right-to-audit and delivery-shortfall remedies in every insertion order, on both sides of the plan, and add language covering loss of accreditation mid-flight.
• Fund one incrementalityincrementalityThe share of results (sales, conversions, revenue) that only happened because of a marketing action, not what would have occurred anyway.View full definition → test per major channel per year, and refuse to approve a mix model whose decomposition has never been checked against an experiment. A credible third-party model costs low to mid six figures; the test costs a fraction of that and decides whether the model is worth anything.
• Ask who deduplicated any cross-media reach figure you are shown. If the answer is a company that sells part of the inventory being deduplicated, treat it as a sales document.

Common mistakes that kill results

Mistake 1: mistaking precision for accuracy. Platform reach arrives to the individual, panel reach arrives with an error margin, and the second looks weaker in a slide. A number with published error bars is worth more than a number with none, because you can price the uncertainty. Executives who reward the tidier figure train their teams to buy the less accountable one.

Mistake 2: letting the two budgets run on separate KPIs and separate owners. Without shared reach deduplication you over-serve the same households across television and connected devices, driving frequency up and incremental reach flat. The failure shows up as a rising cost per incremental point that no single channel report contains, which is why industry cross-media projects such as ISBA's Origin in the UK exist at all.

Mistake 3: cutting the audited line first under board pressure, because it is the only line with a visible price per point and therefore the easiest to defend cutting. Salience decays over months rather than weeks, so the first two quarters look like free savings. The IPA evidence base is consistent that recovering share of voice costs more than holding it, and the recovery argument is far harder to make once you have surrendered the currency that proved the loss.