Why your CAC keeps rising

Blended CAC was $48 this time last year. Now it's $67. Nothing obviously broke, and nobody can say exactly why. Some version of that sentence comes up whenever DTC operators compare notes, and the standard explanation is always the same: ads got more expensive.

The standard explanation isn't wrong. It's small. Benchmark reports this year put average CAC increases in the low-to-mid teens year over year across most DTC categories. If your CAC is up 40%, the auction gave you maybe a third of that. The rest came from inside your own funnel, and it entered quietly.

This essay is that diagnosis – the same one a CAC audit starts with. You can run it on your own brand this week with numbers you already have.

CAC is a chain, not a price

You don't buy customers. You buy impressions. A chain of conversions turns those impressions into orders, and CAC is what falls out the other end:

CAC = spend ÷ orders. Orders = impressions × click-through × conversion.

Three multipliers, and spend divided by their product. This matters because multipliers compound. A small decay in each one produces a large move in the result, and no single decay is big enough to trigger an alarm on its own.

We break the path from spend to order into five gates: Ad, Page, Offer, Checkout, Measure. Traffic leaks at every one of them. When CAC drifts up over two or three quarters, the cause is almost never one gate failing loudly. It is three or four gates each getting 10% worse, silently, at the same time.

Here's what that looks like with plausible numbers. Same brand, same monthly budget, six months apart:

CPM up 8%. That's the auction, the part everyone blames, and the part you can't control. CTR down 14%: creative fatigue. Conversion down 14%: the page and the offer aging. Each of those is a slow bleed a weekly dashboard absorbs without comment. Multiplied together, they are a 46% CAC increase.

The auction contributed eight points. Your funnel contributed the rest.

That's the good news, even if it doesn't feel like it. You can't negotiate with Meta's auction. You can absolutely fix your own gates. So let's walk through them in order, with the numbers to pull at each one.

Gate one: fatigue announces itself before CAC moves

Creative fatigue is usually the first leak to check, and the easiest to see coming, because it shows up in three metrics before it shows up in CAC: frequency rises, CTR falls, and the share of impressions going to people who've already seen the ad grows.

Pull these numbers:

• Spend-weighted creative age. What share of this month's spend went to ads first launched more than eight weeks ago? Past 50%, you're coasting on creative that's past peak.
• CTR by creative cohort. Compare CTR of ads in their first two weeks against the same ads now. Fatigue is the gap.
• Concentration. What share of spend sits on your single best ad? One winner carrying 60% of budget isn't a strength. It's a single point of failure.

The fix isn't "make more ads" as a volume ritual. It's a testing cadence with kill rules: a steady stream of new concepts (not just new crops of the same concept), a defined spend threshold at which an ad must beat the account average or die, and a bench of proven angles ready to take over when a winner fades.

Fatigue isn't an accident. It's a scheduling problem.

Gate two: the page freezes while the ads keep moving

The ad made a promise. The click is a person asking you to keep it. Most landing pages answer a different question than the ad raised, and the visitor leaves within seconds. That's message match, and it decays on its own: the media team ships new angles weekly while the page stays frozen for a quarter, so the gap between promise and page widens with every creative refresh.

Pull these numbers:

• Paid conversion rate vs site average. Cold paid traffic converting far below your blended site rate is normal. Cold paid traffic converting at a fifth of it means the page isn't doing its half of the job.
• Conversion rate by landing path. Segment new visitors by the page they landed on. A product page receiving cold traffic will usually lose to a page built for the ad's angle.
• Mobile speed. Measure largest contentful paint on a mid-range phone over cellular, not on your laptop. Slow pages pay full price for clicks and then throw a share of them away before the first screen renders.

The fix-first order on pages is its own essay, but the short version: speed, then message match, then offer clarity above the fold. Design polish comes last.

Gate three: your best prospects have seen this offer four times

Offers age faster than brands. The bundle that converted cold traffic eighteen months ago is invisible now, and the discount that once felt generous reads as the regular price.

Pull these numbers:

• New-customer AOV trend. Drifting down while CAC drifts up is margin compressing from both ends.
• Discount dependence. What share of new-customer orders used a code? When full-price conversion quietly dies and every acquisition rides a promo, your real CAC is higher than the dashboard version by the depth of the discount.
• Offer-level conversion. If you have run the same lead offer for over a year, test against it. Not a bigger discount: a different structure. Bundle vs single unit, trial size vs full size, free shipping threshold vs percentage off, subscription-first vs one-time-first.

The offer is the highest-impact test most brands never run, because creative is easy to iterate and offers require a meeting. One offer test can move conversion more than a quarter of creative refreshes.

Gate four: ad people call it web, web people call it done

Checkout is where the most expensive traffic you own quietly drops off. It's also the least-watched gate in the funnel, and the reason is organisational rather than technical.

Pull these numbers:

• Checkout completion rate. Sessions that start checkout and finish it. Track the trend monthly, not the absolute number. A two-point slide over a quarter is a real CAC increase that never gets called one.
• Mobile vs desktop completion gap. Paid social traffic is overwhelmingly mobile. If mobile completes 30% worse than desktop, your ad spend inherits that penalty on most of its clicks.
• Shipping reveal. Where does the customer first learn the true landed cost? Surprise at the last step is the classic self-inflicted abandonment.

Fixes here are unglamorous and fast: express payment options up front, shipping cost visible before checkout begins, dead fields removed, an actual working discount-code flow. Checkout work is the rare CRO that ships in days and pays on every order.

Gate five: retargeting creep is invisible from platform ROAS

The last gate doesn't lose orders. It loses information, and it hides the other four.

If you only watch platform ROAS, retargeting creep is invisible: the algorithm shifts budget toward warm audiences who were coming back anyway, reported ROAS improves while real acquisition decays, and the account looks healthiest exactly when it's buying the least new demand. If you only watch blended CAC, you can see that things got worse but not where, which is how operators end up re-briefing the creative team to fix what's actually a checkout problem.

Pull these numbers:

• Returning-customer share of paid conversions. How much of the revenue in your ads manager came from people who had already bought from you? That spend is retention, not acquisition.
• Branded search share. How much of "paid" revenue came through people searching your brand name? Some of that credit belongs to whatever created the demand, not the ad that harvested it.
• Blended CAC, computed honestly. All marketing spend, agencies and tools included, divided by genuinely new customers. Watch the trend monthly. It's the one number in this essay that can't flatter you.

Why single-gate fixes don't hold

Here's the uncomfortable pattern behind the "we switched agencies again" cycle. Each vendor owns one gate. The ad agency refreshes creative, CTR recovers, CAC dips for a quarter, and then the decay at Page and Offer catches up and the drift resumes. The operator concludes the agency got worse.

Usually nobody got worse. A one-gate fix in a five-gate system buys you a quarter, and the gates you didn't touch keep compounding.

This is also why the worked example above matters. A 46% CAC increase built from an 8% auction move and two 14% funnel decays can't be fixed by any amount of media buying skill. The arithmetic won't allow it.

Run the diagnosis this week

Everything above compresses into one afternoon of pulling numbers:

Then rank the gates by how much each has decayed, fix the worst one first, and re-measure four weeks later using blended CAC on new customers only. Not platform ROAS. The platform will report the fix generously either way; blended CAC will tell you whether it happened.

One caution: don't run all five fixes at once. You won't know what worked, and next quarter you'll be back to guessing. One gate at a time, measured honestly, compounds. That's the whole method.

Fix one gate, then re-measure

If you want a second pair of eyes on the diagnosis, this is literally what a CAC audit is: a free 30-minute review of your numbers, no deck, no pitch. You leave with the top three things to fix first, whether or not we work together.

And if you'd rather run it yourself, the table above is the audit. Start with Measure, because until that gate is honest, the other four are guesses.