CMO playbook & advanced tactics: mastering CAC, LTV & ROAS at scale

The request arrives in the operating plan review: another €40 million for paid acquisitionpaid acquisitionVisitors arriving via paid ads or sponsored placements, where you pay a platform to display your message rather than earning visits organically.View full definition →, backed by a blended 3.2x ROASROASReturn on Ad Spend (ROAS) measures the revenue generated for every unit of currency spent on advertising, calculated as revenue divided by ad cost.View full definition → and a deck of channel dashboards. The CFO asks what the last €40 million bought that the company would not have got anyway, and the room goes quiet. That silence is the arbitration this lesson is about. At $10 million of annual spend, an optimistic CACCACCustomer Acquisition Cost (CAC) is the total sales and marketing spend divided by the number of new customers gained in a period. It measures how efficiently you grow.View full definition → number is an error you correct next quarter. At $500 million, it is a capital allocation decision made on the wrong evidence, and unwinding it takes two or three years, a headcount reduction and usually a new CEO.

Core concept: when a good ROAS is still value-destroying

Take the three metrics and the cohort maths as given from the earlier rungs. What no dashboard tells you is that a strong return can be wrong in three independent ways, each owned by a different person.

The first is incrementalityincrementalityThe share of results (sales, conversions, revenue) that only happened because of a marketing action, not what would have occurred anyway.View full definition →. The campaign is credited with revenue that would have arrived without it. The second is the numerator: ROAS is usually computed on gross revenue, while the business only banks contribution. At Zalando, roughly half of everything shipped comes back, so a fashion ROAS that ignores returns, reverse logistics and payment fees describes a transaction that never happened. The third is timing. A payback window that is fine when capital is free is not fine when it is not: US policy rates went from near zero in 2021 to above 5% by mid-2023, and every subscription business carrying a 24-month payback discovered that its growth model had been an interest rate bet.

Stack them. A 4x reported ROAS at a 45% contribution margin is a 1.8x contribution return. Strip out 25% of conversions as non-incremental and you are at 1.35x, before you have paid a single person in the marketing team. The number never looked bad at any point in that chain, and nobody in the meeting was lying.

Key sub-concepts

Sub-concept 1: who owns the payback target

In most companies, nobody does, which is why it moves every quarter. The payback ceiling is a financing decision: it depends on cost of capital, cash on hand and how much dilution the board will accept. That makes it the CFO's number. What sits under the ceiling is the CMO's: which channels, which segmentssegmentsDividing a market into distinct groups of customers who share similar needs, characteristics or behaviours, so each group can be served with a tailored approach.View full definition →, which countries clear it. Write it down that way and the argument stops being about whether marketing is optimistic and starts being about mix.

The failure mode is the reverse assignment. When marketing sets the payback target, it drifts outward with each budget cycle, because a longer window always makes today's spend look affordable. When finance sets the channel mix, you get a company that cuts brand and upper funnel first, then wonders in eighteen months why performance CAC keeps climbing.

Sub-concept 2: what an incrementality-blind CAC costs

Platform-reported conversions overlap. Ask Meta, Google, your affiliate network and your email tool to each claim their share of last month's orders and the total routinely exceeds 100%. If 30% of your attributed acquisitions would have happened anyway, your true CAC is not 30% higher, it is roughly 43% higher, because the same spend now sits on a smaller denominator. Apply that to a business spending $200 million a year and the misstatement is worth more than most marketing departments cost.

The evidence base here is old and inconvenient. eBay's controlled experiment on paid search, published in 2015, found returns on branded search close to zero: the traffic came back through organic links when the ads were switched off. Branded search, retargetingretargetingShowing ads to users who have previously visited your site or interacted with your brand, to bring them back and drive conversion.View full definition → and lower-funnelfunnelThe customer journey from awareness to purchase, typically Awareness, Interest, Consideration, Decision, Action, with prospects narrowing at each stage.View full definition → app install campaigns are where non-incremental spend concentrates, and they are also the channels with the prettiest reported ROAS.

Sub-concept 3: marginal CAC, not average CAC

Budget requests are argued with averages and spent at the margin. If the current $10 million buys customers at $120 average and the next $4 million buys them at $210, the honest number for the incremental decision is $210. A CMO who cannot produce a spend-versus-CAC curve for the top two channels is asking the board to approve a purchase at an unknown price. Saturation shows up first as rising frequency and falling new-to-file rate, weeks before it shows up in CAC.

Sub-concept 4: governing the LTVLTVLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → horizon

LTV is a forecast, and forecasts expand to justify the spend they are attached to. The governance answer is a written horizon that only finance can change: realised contribution over a fixed window, with anything beyond it excluded from acquisition decisions. Second-order effect worth naming to the board: capping the horizon will make some genuinely good long-payback channels look unaffordable. That is the price of not repeating the Peloton arithmetic below, and it is usually worth paying.

Real-world cases

Case 1: uber

In 2017 Uber's performance marketing team, led by Kevin Frisch, ran what began as a fraud investigation into mobile app install advertising and ended as an incrementality test. They switched off roughly $120 million of annual digital ad spend. App installs stayed essentially flat. Uber later sued its mobile ad agency Fetch over fraudulent attribution. The organisational lesson is sharper than the fraud story: the spend had been approved every quarter by people looking at a CAC number that attributed installs to whoever claimed them last. Nine figures of budget survived for years because no one was accountable for proving the counterfactual, and the test that killed it cost almost nothing to run.

Case 2: peloton

At its 2020 to 2021 peak, Peloton was acquiring connected fitness subscribers at well over $1,000 each, funded by sales and marketing spend running near a billion dollars a year. The LTV model behind that was defensible on the data available: monthly churn under 1%, high engagement, subscription revenue at strong margin. What the model missed is that its inputs were correlated. Post-pandemic demand normalisation hit hardware sales, churn and referral volume at the same time, so the LTV side collapsed in the same quarter the CAC side deteriorated. In February 2022 Peloton cut around 2,800 jobs and John Foley handed the CEO role to Barry McCarthy. There is a second trap here for any hardware or first-order-subsidised business: part of the real acquisition cost sat in cost of goods, not in the marketing line, so the marketing dashboard never showed the full price of a subscriber.

Case 3: zalando

In 2022 Zalando stopped buying growth. Facing softening European demand, it shifted the operating goal from GMV expansion to profitable growth, pulled back marketing and pushed harder on existing customers. The trade was explicit and it was not free: active customer numbers stopped growing and then edged down over the following year, while adjusted EBIT recovered into the hundreds of millions of euros. That is the arbitration in its cleanest form. A CMO who presents this as "efficiency" without naming the cost in cohort supply is setting up the next crisis, because the customers you did not acquire in 2022 are the retention base you do not have in 2025.

CMO action items

• Get the payback ceiling written into the operating plan and countersigned by the CFO, with the cost-of-capital assumption stated next to it. When rates or the funding position move, the ceiling moves by an agreed rule, not by argument.
• Put a standing incrementality budget in place: geo holdouts or scheduled dark periods on the two channels with the highest reported ROAS, at least twice a year. Report the incremental CAC alongside the platform CAC in the same table, permanently.
• Require a marginal CAC estimate on every budget increase above a threshold you set (many companies use $250K per month). Approve against the marginal number, not the blended one.
• Name one owner for the LTV horizon and one for the definition of what counts inside CAC, including discounts, referral credits and hardware subsidies. Two people, named in writing.

Common mistakes that kill results

Mistake 1: compensating the media team on platform-reported ROAS

Whatever you pay bonuses on gets optimised, including the attribution. Teams measured on in-platform return will drift toward retargeting and branded search because those channels report beautifully. Measure the paid media function on contribution after returns at a fixed cohort age.

Mistake 2: letting CAC hide outside the marketing p&l

Rider and driver incentives, free trial hardware, first-order discounts and referral credits often sit in contra-revenue or in cost of goods. Marketing then reports a CAC that is technically accurate and materially wrong, and the board approves scaling on it.

Mistake 3: treating a broken ratio as a marketing problem

When the ratio inverts, the reflex is to cut media. Sometimes the fix is pricing, retention or the product's second-purchase rate, and cutting acquisition just shrinks the cohorts that would have funded the recovery. Decide which lever you are pulling before the quarter's number forces the decision for you.