How to lower your cost per acquisition: lessons from an 86% drop

Rising cost per acquisition is the most common reason a growing brand's marketing stalls. The instinct is to spend more. Usually that's the wrong move: it buys the same expensive customer, faster. The more reliable way is to check that you are counting the right thing, match the channel to how your product earns trust, and test many more ads than feels necessary. That is how one brand's cost per acquisition fell from S$25.30 to S$3.49.

Cost per acquisition, or CPA, is simply what your ads cost for each new customer, enquiry or sale. If you spent S$1,000 and won 40 customers, your CPA is S$25.

Why spending more rarely brings it down

Ad platforms find your cheapest customers first. When you raise the budget, they have to reach further: people a little less likely to buy, searches a little less exact, audiences a little less like your best customers. So the extra money usually buys customers who cost more than the ones you already had.

That is why a flat or rising CPA is rarely a budget problem. Something else is holding it up, and there are four places to look.

First, check that you are counting the right thing

Before changing anything, make sure the ad platform can see the sales you care about. It can only steer towards what it can count.

On Meta, that means the pixel on your website and the Conversions API from your server; Meta recommends both, with each sale counted once. On Google, it means the Google tag and, where sales close off the website, enhanced conversions. If many of your sales close on WhatsApp, by phone or in store, those sales need to reach the platform too, or it will keep finding people who click and never buy.

A surprising share of "high CPA" turns out to be a counting problem: the platform is optimising for clicks or form fills that never become customers.

Second, ask whether your channel catches demand or creates it

Search ads catch people who already want what you sell. Someone types "ship from US to Singapore" and you're there. That's valuable, but the pool is finite. Once your ads already show for most of those searches, more money mostly pushes up what you pay for each click, because every competitor bids on the same words.

Video ads on Facebook, Instagram or TikTok create demand: they reach people who weren't looking and show them why they'd want it. If your growth has stalled on search, the cheaper customers are often among people who don't know yet that they need you.

Third, ask whether your product is searched for or shown

Some products are easy to search for: a plumber, a spare part, a tax adviser. Others only make sense once you see them working. If yours is the second kind, a text ad on a search page is the hardest possible way to sell it, and your CPA will show it.

Fourth, count how many ads you are really testing

One or two ads, however polished, is a bet. Ten versions, each with a different opening, run side by side with the money moving to whichever brings customers in most cheaply, is a test. In our experience, on Meta especially, the ad itself now does much of the work of finding the right people: an ad that stops them is cheaper to show.

Ads also tire. When the same people have seen an ad several times, its results slip, so new versions need to be ready before the old ones fade.

The case: search had hit its ceiling

Buy&Ship, a cross-border shopping and shipping service, was running acquisition in Singapore and Malaysia on its own Google Search ads. Search is excellent at one thing: capturing demand that already exists. But only so many people search for a service they may not know exists. Its cost per acquisition had settled at S$25.30 and wasn't coming down.

The shift: match the channel to how the product earns trust

The product was something you have to show, not a keyword you bid on. So we moved its advertising to video on Facebook and Instagram in the same two markets, keeping the markets identical so the comparison stayed fair.

A short video can walk through the whole loop: shop at a store abroad, ship to a local address, receive it at home. A text search ad never can. And Meta reaches the much larger group of people who'd never think to search for the service, but want it the moment they see it.

What moved the number: many video ads, tested side by side

We treated the ads as the main lever, not the bids. Several video angles ran against each other: the frustration of a store that won't ship to you, the savings, the moment the parcel arrives. Each was cut for the feed and for Reels. Money moved to the versions that brought requests in most cheaply; the rest were switched off quickly, and each new round started from what the last one taught us.

Cost per acquisition fell from S$25.30 to S$3.49, 86% lower. With the same budget, Buy&Ship got about seven times as many requests. The budget didn't change. The channel did, and so did the number of versions we tested. The full story is in the Buy&Ship case study.

Give the change time to show

Both platforms need a run of results before their numbers settle. Meta's ad system settles once an ad set has had about 50 results in a week. Google suggests judging a campaign that bids towards a target cost over at least 30 conversions in the last 30 days. Change everything every few days and the platform never finishes learning, which keeps your CPA high on its own.

How to apply this to your account

Ask four questions, in this order:

1. Can the platform see the sales I care about, including the ones that close off the website?
2. Is my channel catching demand or creating it, and which does my growth need now?
3. Is my product something people search for, or something they need to be shown?
4. Am I testing enough versions to find a winner, and replacing them before they tire?

Lower CPA almost never comes from one clever tweak. It comes from counting the right thing, matching the channel to how your product earns trust, and testing enough ads to find what works.

Questions, answered