How We Added 230,000+ Registered Canva Users in French-Speaking Markets for as Little as $0.03 Each

By Kevin Veitia

Client

Canva, the online design platform

Markets

French-speaking countries in Europe and sub-Saharan Africa

Channel

Meta Ads (Facebook and Instagram)

Conversion

Registration: install, open, then sign up

Timeframe

2017 to 2018, a one-year contract

The challenge

In 2017, Canva hired us for global user acquisition on its mobile app. Paid social wasn't a big part of its growth yet. The testing it had done was mostly in the US, and the best it had managed was roughly $1 to $1.25 per user.

The US is one of the most expensive ad auctions on the planet. So the question was simple: is there a market with valuable users and less competition? Canva picked the French-speaking world, in Europe and sub-Saharan Africa, and gave us a year to find out.

The first thing we did was kill the vanity metric. An acquisition wasn't an install. It was someone who installed the app, opened it and registered. Cost per install makes campaigns look great, because a big chunk of installs never turn into users. Cost per registration only counts the people who actually showed up.

What we did

Built audiences from what people actually use Canva for

Canva does a hundred things, and "design anything" speaks to nobody. So we skipped the ad account and went into the product analytics. We pulled the specific jobs people came to Canva for, like birthday cards and CVs, and built user segments around each one.

Then we built 1% lookalikes from each segment. This was 2017, years before Apple's privacy changes, and lookalikes were the best weapon Meta had. Seeding them with people defined by a real use case gave the algorithm a much cleaner signal than "everyone who signed up."

Wrote every ad for one job

Nobody wants design software. They want a CV that lands an interview, or a birthday card that makes someone's day. So every audience got creative written for the one job it was hiring Canva to do.

We call it targeting–messaging fit: the right job, in front of the people who have that job, in words that sound like them. Get it right and the ad stops looking like an ad. It looks like the thing they were already trying to do.

Tested like a maniac, killed without mercy

Over the year we ran somewhere between 20,000 and 30,000 ads. Volume without rules is just burning money, so the rule was set before anything launched: once an ad hit about 3,000 impressions, if click-through was under 1% and cost per registration was over target, it died. Next variant in.

Ads that haven't found their people by then almost never come back, and every day a loser keeps spending is a day a winner doesn't. A rule also ends the arguing. Nobody gets precious about their favorite ad when the numbers have already made the call.

Every time we hit the target, Canva moved it

First target: $1 per registration. We hit it fast. Canva dropped it to $0.75. Then $0.50. Then $0.25. Then $0.10. Every time we cleared one, a lower one showed up.

Each reset meant going back to the audiences and the creative, because what gets you to a dollar doesn't get you to a dime. The best result of the year was $0.03 per registration, and it hit on several different ads. That's how we knew it was a repeatable formula, not a fluke.

The data didn't care what we assumed

Some results surprised us. For French-speaking markets in sub-Saharan Africa, our instinct was to cast local models. The data said otherwise: the creative that converted best featured a model who read as French. That looked like aspiration, not identification.

We're not going to pretend that's comfortable. Creative that leans on aspiration can echo some ugly history, and it has to be handled with respect. But we don't get to throw out a result because we don't like it. Our job is to find out what actually moves people, then make responsible calls with that knowledge.

Results

• 230,000+ registered users added, counted at registration, not install
• Cost per registration as low as $0.03, hit on several different ads
• Target cost per registration lowered from $1 to $0.10 within the year, with every target met
• More unique viewers reached than watched the 2018 Oscars broadcast

Compared to the $1 to $1.25 Canva had been paying in the US, a three-cent registration is about 97% cheaper. And that's for a user who installed, opened the app and signed up.

So why did it stop?

Everyone asks. Canva's organic search was bringing in registrations for under a cent, so the company went all-in on its cheapest channel and switched paid off.

Looks smart on a spreadsheet. In real life, channels feed each other: people see an ad, then show up through search later. Kill one channel and you change the math on all the others. When paid came back into Canva's mix later on, it ran at many times the cost per registration we'd gotten it to. We call this min-maxing: optimizing every channel on its own. It looks brilliant for a quarter and expensive for a year.

"OuiGrowth did an exceptional job for Canva on our Meta Ads campaign, reaching CPA as low as 0.03 cts. They're always enthusiastic and dedicated to reaching our ambitious goals. We will definitely get in touch again for our future social paid and organic campaigns. Highly recommended marketing specialists!"

Pierre Girardot, Canva

What you can take from this

• Measure the first event that proves a real user. Installs lie. Registrations don't.
• Your best targeting insight is sitting in your product analytics, not your ad account.
• One ad, one job. Someone making a CV and someone making a birthday card need different reasons to download the same app.
• Write your kill rule before launch: impression threshold, click-through floor, cost ceiling. Then actually follow it.
• Hit the target? Set a lower one. What got you here won't get you there.
• Don't min-max your channel mix. The "cheap" channel usually depends on the "expensive" one.

Want a second set of eyes on your account? Book a free 30-minute consultation and we'll walk through what's working, what isn't, and where the next win is. Curious what it costs first? Our pricing is public.