First, recompute the trend and split it into parts
Before diagnosing anything, make sure the series is comparable. Recalculate acquisition cost for every month in the period using one written definition: which costs are in the numerator and what counts as an acquired customer. If the reported trend was assembled from monthly decks, it may already mix definitions. Then break the cost into three ratios for each month and each main channel: spend per visit or click, the share of visits that became a lead or an order, and the share of leads that became paying customers. Multiplied together, these reproduce the acquisition cost, and whichever ratio moved is where the diagnosis starts.
Read the split by channel as well as blended. Blended cost can rise because one channel got dearer, because budget moved into an expensive channel or because a cheap source such as organic search shrank, and only the channel view tells them apart. Kaushik's measurement model insists that each key measure has a fixed definition tied to the business objective; a cost series that changes its definition halfway through cannot be diagnosed at all.
If the figure being tracked is a platform's cost per acquisition rather than customer acquisition cost, the CPA versus CAC comparison shows how the two differ and how to reconcile them.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Cost per customer in the largest channels rose as their budgets grew | Marginal cost of recent budget increases against the channel average | Channels past the point of diminishing returns | Cap each channel where marginal cost meets target; move the surplus |
| Each channel's cost is stable but the blended figure rose | Share of spend by channel now against a year ago | Budget mix shifted toward costlier channels | Reallocate on marginal returns, keeping deliberate investments labelled |
| Spend per visit flat; fewer visits become leads or orders | Conversion rate by step, device and source over the period | Site or funnel conversion falling | Find and fix the step that lost the most visitors |
| The figure jumped in one month with no change in the market | Definition history: which costs were added, and when | Definition changed | Restate history on one definition and label the figure |
| Cost per click or per thousand impressions rising in the main platforms | Like-for-like price trends in each platform's own reports | Auction prices rising | Diagnose on the platform's troubleshooting page; adjust bids and mix |
| Leads per pound unchanged; fewer leads become customers | Lead-to-customer rate by cohort and source | Close rate falling after the hand-over | Review qualification, follow-up and stage definitions with sales |
Fix, verify and prevent each cause
Channels past the point where extra spend pays
Every channel buys its cheapest customers first. As budget grows, each additional pound reaches people less ready to buy, so the average cost per customer creeps up while the cost of the last tranche rises much faster. This is diminishing returnsEach extra unit of spend or effort buys less extra result than the one before it., and it explains most slow increases in channels that have been scaled steadily.
Fix: estimate the marginal cost of recent increases in each large channel, using the budget-step method in the marginal returns and forecasting guide, and hold each channel at the level where that marginal cost meets the target.
Verify: after holding or trimming, the channel's average cost per customer stabilises and total customers fall less than the spend cut would suggest.
Prevent: record a marginal-cost ceiling for every major channel and re-estimate it whenever budget moves materially. Where the channel to trim is Google Ads, the guide to cutting a Google Ads budget finds the marginal spend by campaign, network, location and time, and chooses between a lower budget and a tighter target.
A mix drifting toward dearer channels
Acquisition cost is a weighted average across channels, so moving budget toward expensive channels raises it even if no channel's own cost changes. Sometimes the move was deliberate, such as entering a new market; sometimes it was a series of small reallocations nobody added up.
Fix: compare the spend mix now with the mix in the base period, and re-run the allocation on marginal returns as the budget allocation guide describes, separating deliberate investments from drift.
Verify: blended cost moves back toward its earlier level, or the remaining increase is fully explained by labelled investments.
Prevent: report blended cost with a mix-adjusted version beside it, so a change in weighting is visible as a change in weighting.
Fewer visitors becoming leads or orders
If the price of traffic is unchanged but a smaller share of it converts, every customer needs more traffic and the cost rises without any change in the media. Site changes, slower pages, a new form or a price rise are the usual triggers.
Fix: find the step that lost the most people using the method in the funnel analysis tutorial, which builds the funnel exploration Google describes in its guide to getting started with Explorations, segment it by device and source, and fix that step first. Read the conversion rate (CVRShare of visits or clicks that convert.) by step rather than end to end.
Verify: the step's completion rate recovers, and visits per customer fall back.
Prevent: add step-level conversion to the monthly cost review, and check the main paths after every site release.
A definition that changed
Acquisition cost has more than one defensible definition, and the figure can jump simply because the business started including agency fees, tools, salaries or incentives that were previously left out, or started counting only paying customers rather than sign-ups.
Fix: use the definitional choices in the LTV and CAC guide to write one definition, restate the history on it, and label every published figure with the definition used. Customer acquisition cost (CACCost to acquire one new customer.) reported to finance and to channel teams can differ, but each must say what it includes.
Verify: the restated series shows whether a real increase remains once the definition is constant.
Prevent: keep the definition in a shared metrics dictionary and treat any change as a dated event in the report.
Dearer auctions on the main platforms
Paid channels price attention in auctions, and more competition, seasonal peaks or changes in bidding can raise what each click or impression costs. That pushes acquisition cost up even when conversion holds.
Fix: confirm the price rise like for like inside each platform, then diagnose it there: the Google Ads CPC increasing diagnosis and the LinkedIn costs rising diagnosis separate real price increases from shifts in what you are buying. If prices have risen for good, move budget to where marginal cost is lower.
Verify: cost per click or impression stabilises on the segments you kept, or the budget moved to channels where customers cost less.
Prevent: watch platform prices beside cost per customer every month, so a price change is seen before it reaches the blended figure.
Fewer leads becoming customers after the hand-over
In businesses with a sales team, the cost per lead can hold steady while acquisition cost rises because fewer leads close. The cause may be looser qualification, slower follow-up or a change in the sales process.
Fix: read lead-to-customer rates by cohort and source using the stage model in the lead stages guide, and if the fall sits between marketing qualification and sales qualification, work through the diagnosis of MQLs not converting to SQLs.
Verify: the close rate for recent cohorts returns toward its earlier level, by source.
Prevent: report acquisition cost on won customers, not on leads, so a weaker close rate shows up in the cost the business actually pays.
A prevention routine for acquisition cost
The routine is built so that a rise shows up as a specific component moving, not as a vague upward drift. Monthly, publish acquisition cost on its written definition, blended and by channel, with the three ratios beside it: price of traffic, conversion to lead or order, and lead-to-customer rate. In the same review, compare the spend mix with the base period and read platform price trends for the largest channels. Quarterly, re-estimate marginal cost for each channel whose budget moved, using a planned budget step where possible, and refresh the ceilings. Quarterly too, check the definition against what finance uses and restate history if it changed. Twice a year, compare acquisition cost with contribution per customer and lifetime value by cohort, so the question "is this rise acceptable?" has an answer. When a rise appears, look at the newest change first: a budget increase, a new channel, a site release, a pricing change or a change in what the figure includes explains most increases.
A meal-kit subscription unpicks its rising cost
A meal-kit subscription saw blended acquisition cost rise from £62 to £81 over nine months. Restating the series on one definition removed £6 of the increase: from month five, the report had begun adding agency fees. Of the remaining rise, the channel split showed paid social steady at about £58 per customer, while paid search had grown its budget by 70% and its marginal cost on the last increase was about £120 against a target of £75. At the same time a new checkout step had cut visit-to-subscription conversion from 4.1% to 3.4%. The team trimmed search back to the level where marginal cost met target, removed the extra checkout step, and labelled the fee-inclusive figure for finance. Two months later blended cost on the original definition was £66.
Common mistakes
Reading only the blended figure. Comparing months built on different definitions. Treating every increase as an auction problem. Scaling a channel on its average cost while its marginal cost has passed target. Ignoring the close rate because the cost per lead looks fine. Changing the mix, the site and the definition in the same quarter, so nothing can be attributed.
