How I 10x'd New Downloads And Cut Cost Per New User by 40x For A $XX Million AI Startup
"All happy companies are different." - Peter Thiel
I started working with an AI devtool startup with $XX million in funding.
Goal? Growth.
Problem? Cost.
The Problem
The current advertising was far too expensive and they didn't have the bandwidth to fix it. We reduced the UAC (User Acquisition Cost i.e. the cost to get a new user) by 40x (not a typo).
To put that 40x cost reduction in real numbers: If you're spending $500,000/yr on acquisition and reduce the UAC by 40x, you save $487,500. I can't disclose the numbers, but the startup saved a Scrooge McDuck vault of money.
This allowed us to grow rapidly and 10x the number of new downloads by January.
Also - we're talking US users. Not pumping up the numbers with users from developing countries who seldom convert to paid plans.
I'll try to explain how we did this best I can.
The Big Picture Is Everything
The first thing I noticed was that we weren't advertising our difference - our big idea. This happens with most companies when they copy competitors blindly, or the marketing team doesn't have the bandwidth, or use an agency that doesn't put in the hours it takes.
This creates a problem. If we say the same words and look and feel the same as everyone else - they will think our product is like everyone else’s. Unless we're different, we won't escape mediocrity.
If you’re not sure what's your big idea, start with a comparison table. Are we 1.3s faster than competitors? 4x cheaper? Safer? Used by half of Fortune 100 (and soon - the other half)? Or the best kept secret of early-stage startups?
"Unless your advertising contains a big idea, it will pass like a ship in the night." - David Ogilvy
So what makes this AI devtool startup different? Their product runs in the background and sees all the work you're doing for up to 9 months. Then you can access all that important info with an AI. Very nice.
You can ask questions like:
- What's the link to that retention dashboard Alice sent a few weeks ago?
- Can you write an update for everything I worked on last week and my tasks for this week?
- Can you sum up and link all the research papers I read on arxiv today?
We tested a bunch of different ways to say this. What worked best: "AI that remembers everything you work on."
Simple. Immediately understood. That was our big idea.
Author's note: I should repeat this point twice, three times over. If we didn't get the big idea right, no little ads optimization would help.
Now. Where could we find notoriously expensive-to-reach developers (our ideal avatar)?
Untapped Opportunity
Google Search was expensive. But display ads? Many startups ignore them because they think they don't work. They usually don't, but we'll get to that later.
Retargeting was especially good. Someone visited but didn't download? Maybe they got distracted at work or visited our page during their commute - we could get them back super cheap.
So, so many ads
When I analyzed their current advertising one thing stood out - they didn't have the bandwidth to make enough creatives. We needed to implement a regular creative and testing cadence. Big picture is key, but so is iteration.
"The most important word in the vocabulary of an advertiser should be TEST" - David Ogilvy
"'You miss 100% of the shots you don't take.' - Wayne Gretzky" - Michael Scott
What worked best:
- Actually funny developer memes relevant to the big idea
- Falsifiable proof - 9 months of memory
- Graphs (who doesn't like graphs??)
- Before/after comparisons
- Short form videos
A good framework I like to keep in mind when making ads:
- Can I visualize it? - eg. dangle a Volvo truck from a crane to show how strong the truck’s hook is
- Can I falsify it? - eg. “Not a drop is sold till it’s seven years old - Jameson”
- Can nobody else say this? - eg. “Your car has five numbers on the speedometer. Volvo has six.”
Having that much variety meant we could:
- avoid people getting sick of ads
- convince a much larger percentage of the TAM by explaining our product in a way compelling to many subsets of users.
The Weird Quality Problem
So we make these changes and the number of downloads grows 10x in two months. Great, right?
But I noticed tons of people would download but never install. Our ads were showing on garbage websites. Like roblox cheats and free app sites.
I would've thought excluding these using Google's website categories in campaign settings helps. It doesn't. This is exactly why many can't get Display Ads to work for cold traffic.
Solution: website exclusion lists. I made a list of 70,000 "bad websites" and blocked them all. Next day, 3x more people were finishing the install.
The Numbers
I won't show real numbers because of the advertiser-client privilege (it's a thing), so lets simulate data.
If you're spending $500,000/yr and you reduce UAC by 40x, that's $487,500 saved while getting the same number of users! And we got more.
Turning the right gears
Advertising has 9,999 gears to it. Knowing which 4/9,999 most important ones to change gives you most of the returns in the fastest time. Alas, speed is key with startups.
Unfortunately it takes years and 10,000 iterations to know how to spot those 4 gears among the 9,999 and how to fix them.
“A blind pig can sometimes find truffles, but it helps to know that they are found in oak forests." - David Ogilvy
But I do this kind of thing for tech startups - find the marketing opportunities everyone else misses. I might be able to do the same for you. Send an email at hello@lukakalajzic.com (I reply faster if you include a star wars quote [none from the VII-IX movies though])!