Marketing return: CAC, payback, incrementality, and why last-click credit misleads
How to tell whether marketing spend paid for itself: blended and paid CAC, marginal CAC, break-even return, and incremental tests.
Firm processes and online tests change from year to year and differ by office. Use this to prepare, and confirm the exact current steps on the firm's own careers page.
Key takeaways
- Marketing pays when it causes sales that would not have happened anyway, and when the profit on those extra sales is more than the spend.
- Common mistakes: Comparing ROAS with 1 instead of with 1 divided by the gross margin.
- Blended CAC: all marketing spend divided by all new customers, including those who came on their own.
- Paid CAC: paid spend divided by the customers that paid marketing actually brought in.
- Marginal CAC: the cost of the next customer. The first customers are the cheapest to reach; each extra unit of spend usually brings fewer.
Key idea
Marketing pays when it causes sales that would not have happened anyway, and when the profit on those extra sales is more than the spend. Most reports give credit to the last ad clicked before a purchase, which flatters ads shown to people who were already going to buy.
Four numbers beyond the basics
- Blended CAC: all marketing spend divided by all new customers, including those who came on their own. It looks lower than the truth for paid channels.
- Paid CAC: paid spend divided by the customers that paid marketing actually brought in. This is the number to compare with lifetime value.
- Marginal CAC: the cost of the next customer. The first customers are the cheapest to reach; each extra unit of spend usually brings fewer.
- Break-even return on ad spend (ROAS): revenue per unit of spend needed to cover the spend, which is 1 divided by the gross margin. At a 40 percent margin you need 2.5 of revenue for every 1 spent, just to break even.
Incrementality: would the sale have happened anyway?
Incremental sales are the extra sales caused by the marketing. The clean way to measure them is a test: show the ads to one randomly chosen group, hold them back from a similar group, and compare purchases. Google Ads describes its Conversion Lift tool this way: people who see the ads form the test group, people who do not form the control group, and the difference in conversions is the lift. Large studies back the warning. In experiments run with eBay and published in Econometrica in 2015, ads on searches for the brand's own name had no measurable short-term benefit, and most spending on other search ads went on frequent buyers who would have bought anyway. People who type a brand's name have usually already decided to visit. A 2019 Marketing Science study of large randomised advertising experiments at Facebook found that common methods that compare people who saw ads with people who did not often failed to recover the effects the experiments measured.
Attribution is a rule for sharing credit for a sale among the ads a buyer saw. Last-click gives all the credit to the final ad clicked. Google Ads now offers only last-click and a "data-driven" model that spreads credit using past data, and says the first-click, linear, time decay and position-based models are no longer supported. Any such rule shares credit among ads; only a test with a held-back group shows what the ads caused.
Worked case
A furniture retailer in the UAE: which ads really pay?
The prompt
A fictional online furniture retailer in the UAE has a 40 percent gross margin. Its last-click report says ads on searches for its own brand name returned AED 1.2 million of sales on AED 100,000 of spend, and social media ads returned AED 400,000 on AED 200,000. The marketing team wants to cut social. A holdout test then shows only 10 percent of the brand search sales were extra, while social ads caused AED 600,000 of extra sales (many buyers saw a social ad, then later searched and clicked a brand ad). What should the retailer do?
The structure
- Profit from an ad = incremental sales x gross margin minus spend
- Break-even ROAS = 1 / gross margin
- Key: Last-click ROAS versus incremental ROAS for each channel
- Profit from each channel
Working it through
1. Break-even ROAS
At a 40 percent margin.
Break-even ROAS:1 ÷ 0.4 = 2.52. Brand search, last-click ROAS
AED 1.2 million of credited sales on AED 100,000.
Brand search last-click ROAS:1,200,000 ÷ 100,000 = 123. Brand search, incremental sales
Only 10 percent of the credited sales were extra.
Brand search incremental sales (AED):1,200,000 × 0.1 = 120,0004. Brand search, profit
Margin on the extra sales minus the spend.
Brand search profit (AED):120,000 × 0.4 - 100,000 = -52,0005. Social, last-click ROAS
AED 400,000 of credited sales on AED 200,000.
Social last-click ROAS:400,000 ÷ 200,000 = 26. Social, incremental ROAS
AED 600,000 of extra sales on AED 200,000.
Social incremental ROAS:600,000 ÷ 200,000 = 37. Social, profit
Margin on the extra sales minus the spend.
Social profit (AED):600,000 × 0.4 - 200,000 = 40,000
The recommendation
The retailer should keep social ads and cut most brand search spend, the opposite of what the last-click report suggests. First, brand search looks like the star with a ROAS of 12, but only AED 120,000 of its sales were extra, so it loses about AED 52,000. Second, social looks weak at a ROAS of 2, below the 2.5 break-even, but it caused AED 600,000 of sales, an incremental ROAS of 3, and earns about AED 40,000. The risk is that a rival bids on the retailer's brand name once it stops, taking some searches. As a next step, cut brand search in a few test weeks or regions and watch whether total sales move.
Risks: One test is one period; repeat it before moving large budgets; Social ads may also build demand that a short test does not capture.
A fitness app in the UK spent GBP 1,000,000 on paid marketing and won 10,000 new customers. A test shows 6,000 of them would have joined anyway. What is the paid CAC, in GBP, counting only customers the marketing brought in?
A delivery app in Thailand wins 1,000 customers on THB 100,000 of ads a month. Raising spend to THB 150,000 wins 1,300 customers. What does each of the extra 300 customers cost, in THB? (One decimal place.)
A software customer costs USD 1,800 to win and brings USD 150 of contribution a month. How many months until the acquisition cost is paid back?
Comparing ROAS with 1 instead of with 1 divided by the gross margin. Using blended CAC to justify paid spend. Judging the next unit of spend on the average CAC. Trusting a channel because the report credits it with sales, without asking whether those buyers would have come anyway. Treating any attribution model, however clever, as proof of cause.
Gross margin is 25 percent. What return on ad spend is needed just to break even?
Ads on searches for your own brand name show a very high return in the last-click report. Why might that mislead?
Sources for this lesson (5)
- Recognized public explanations of case-interview concepts and terms
- Econometrica (2015), "Consumer Heterogeneity and Paid Search Effectiveness: A Large-Scale Field Experiment" (experiments run with eBay)
- Marketing Science (2019), "A Comparison of Approaches to Advertising Measurement: Evidence from Big Field Experiments at Facebook"
- Google Ads Help, About Conversion Lift
- Google Ads Help, About attribution models
My notes on this lesson
0 of 5,000 characters. Saves automatically.
Try the 5 remaining checks and drills above to complete this lesson (0 of 5 done).
Spotted something wrong or out of date? Report a mistake. We check every report and correct the page.