Real-world application of paid digital marketing

One row in a HelloFresh growth spreadsheet governs everything else in the paid programme: how many boxes a newly acquired customer has to receive before the money spent to acquire them comes back. Not a 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 → target, not a channel split. Boxes. HelloFresh sells a subscription with no shop window and no supermarket shelf, so close to all of its demand is bought. At its peak scale that meant marketing spend of the order of a billion euros a year against roughly €7.6bn of 2022 revenue and around seven million active customers. This lesson stays inside that one programme, across Google, Meta and LinkedIn, through the years the arithmetic worked and the years it stopped.

The number the programme is managed to

Take 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 → as the foundations lesson defines it. What matters here is what sits on the other side of the equation. HelloFresh's average order value has run in the region of €50 to €60, with contribution margin around a quarter of revenue, so a delivered box throws off something like €13 to €15 towards paying back acquisition. A €60 CAC needs four or five boxes. A €90 CAC needs six or seven.

Now put the retention curve beside it. Meal-kit churn is front-loaded: a large share of new customers never 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 → a fifth box and many stop after the discounted first one. The CAC ceiling is therefore not set in the marketing plan. It is set by how many boxes the median cohort actually takes. If that number falls from five to three, the ceiling drops by roughly 40% and every channel bidding above it turns loss-making overnight, with nothing having changed in the auction itself.

There is a second acquisition cost that most teams file in the wrong place. HelloFresh buys first orders with steep discounts, frequently more than half off with free delivery. Economically that discount is acquisition spend; in the accounts it appears as lower revenue and thinner gross margingross marginGross margin is the share of revenue left after subtracting the direct cost of producing goods or services, expressed as a percentage of revenue.View full definition →, not as media cost. A €40 first-box discount on top of €50 of media is a €90 acquisition, and only the €50 reaches the media report. Any payback model built on the media number alone is flattering itself.

Sub-concept 1: Google, where the demand already exists

Search splits into two businesses that look like one line in the account. Generic queries ("meal kit delivery", "recipe box for two") reach people choosing a category. Branded queries ("hellofresh discount code") reach people who have already decided, usually because Meta, a friend or a podcast host put the name in their head.

Branded search will always report the lowest CAC in the account, and most of it is harvest rather than creation: the click would often have arrived free through the organic result immediately below the ad. The counter-argument is real too, because competitors bid on the HelloFresh name and an unbid brand term hands the click to whoever is willing to pay for it. Google sells this inventory, so its own guidance is not a neutral read on the question. The honest resolution is the holdout design the frameworks lesson sets out, run market by market, not an argument in a planning meeting.

The structural constraint matters more than either. Category search volume is capped by how many people type the query, and HelloFresh had already captured most of it in its large markets. Growth beyond that cap has to come from social, where CAC is worse by construction. So blended CAC rises as you scale even when no single channel has deteriorated. Teams that miss this go hunting for a broken campaign that does not exist.

Sub-concept 2: Meta, and what creative actually buys

Meta is where HelloFresh creates demand rather than collects it, and it is where creative production becomes the real cost centre. Food advertising fatigues fast: the same overhead shot of a chicken traybake reaches the same feed for the fourth time and click-through collapses while frequency climbs. The output required is dozens of variants a month per market, in local languages, with local recipes, refreshed on a calendar rather than when performance has already sagged.

Then the measurement floor moved. Apple's App Tracking Transparency rollout from April 2021 cut the signal Meta receives from iOS users; Meta itself told investors in early 2022 to expect a revenue impact of around $10 billion that year. Meta sells the inventory and also supplies the conversion report, and as observed conversions thinned, more of that report became modelled. For a subscription advertiser the second-order effect is nastier than the reporting noise: a platform that overstates conversions understates CAC, the payback model says there is headroom, and the team scales into cohorts that were never clearing the ceiling. The correction arrives one or two quarters later in the retention data, by which time the money is spent.

Sub-concept 3: Where the credit lands

Last-click gives the conversion to the final touch before checkout, which in a subscription funnelfunnelThe customer journey from awareness to purchase, typically Awareness, Interest, Consideration, Decision, Action, with prospects narrowing at each stage.View full definition → is almost always branded search or a 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 → placement. Both then look like the best-performing lines in the plan, and both get more budget. Neither created the customer. Meanwhile the prospecting video that did the work reports a CAC three times higher and gets cut.

The distortion compounds with discounting: an influencer code or a comparison site captures the last click on a customer who was already converting, and pays for the privilege out of margin. HelloFresh's practical defence is the same one available to any subscription advertiser, which is to test at the market level rather than argue at the campaign level, using the method the frameworks lesson lays out.

Sub-concept 4: The LinkedIn question, and the answer

Every consumer subscription business eventually asks whether it should be on LinkedIn. Run the numbers before the debate. LinkedIn prices impressionsimpressionsThe 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 → well above Meta because the targeting is by employer, job title and seniority (LinkedIn sells this inventory, so treat its case studies accordingly). A box paying back €13 to €15 cannot absorb that CPMCPMCost Per Mille: the cost to deliver 1,000 ad impressions. A pricing and benchmarking metric for awareness campaigns where reach matters more than clicks.View full definition → at any plausible conversion rate against a general working audience. Floor prices and minimum audience sizes make it worse in practice: a €5,000 test buys too little volume to produce a readable result, so the channel gets judged on noise.

LinkedIn earns its place where a single conversion is worth hundreds or thousands of euros rather than fifteen. For a company operating fulfilment sites across dozens of markets, that means hiring and employer brand, and it would mean any corporate or office-catering line where one signed account buys hundreds of boxes. The discipline is to name which P&L the spend is paying back into before the first euro moves, not to run consumer creative on a B2B network and call the CPL disappointing.

Three episodes from one programme

Episode 1: the 2020 surge. Revenue roughly doubled as lockdowns pushed people to home cooking. Demand arrived without being bought, CPCs in the category stayed reasonable, and reported CAC fell. The trap is reading a demand shock as a marketing achievement and setting the next year's budget on it.

Episode 2: 2021, signal loss and expensive scale. Post-ATT measurement got looser precisely as the company was spending hardest, and acquisition costs across the category climbed as every subscription business chased the same feed inventory.

Episode 3: the correction. From 2023 HelloFresh told investors it was prioritising profitability over customer growth, leaning on retention and on its ready-meal line Factor, acquired in 2020, rather than on ever-larger acquisition budgets. The shares had lost the great majority of their 2021 peak value by then. Paid media did not cause that on its own, but a payback model that had been reading optimistic inputs for two years made it worse.

CMO action items

• Rebuild your payback model on delivered orders or renewals, not on months, and rerun it every quarter against the current cohort curve rather than last year's. The CAC ceiling moves when retention moves.
• Add discounting to your acquisition cost line. If the first order is half price, that discount belongs next to media spend in every channel comparison you show the board.
• Before any B2B channel test, write down the conversion value that would make it clear its floor price, and the budget needed to reach a readable sample. If either number is uncomfortable, you have your answer without spending.

Common mistakes that kill results

Mistake 1: reading a demand shock as media performance. HelloFresh's 2020 CAC looked superb because the world had changed, not because the campaigns had. Budgets set on that basis meet a normal market twelve months later.

Mistake 2: trusting platform-reported conversions during a measurement break. When more of the report is modelled, reported CAC drifts below true CAC and the payback model quietly authorises spend the retention curve will not support.

Mistake 3: treating branded search savings as free money. Cutting brand bids because the organic result sits underneath works until a competitor buys the term. Test it in one market for a fortnight and read total orders, not the search account's own CAC.