Bionic Ads: The $139,522 Facebook Ad
The “secret sauce” has always been research. But what if that’s not true?
We’d been at it all day.
For six straight hours, we scrutinized our ads, emails, and sales page.
We pored over the data.
Punched holes in our theories.
And evaluated copy ideas, headlines, spitball drafts, calls-to-action, proof, benefits, feedback and revisions.
Everything was put under the microscope.
Each question was answered with watertight facts and ironclad proof.
Team members, stakeholders and decision-makers all chimed in with their ideas.
But as the hours chugged away, frustration and exhaustion set in.
A group session in the conference room can only take you so far.
Sooner or later, you have to decide what copy you run with—and let the market decide what “converts” best.
At the end of the day, we settled for campaign copy that can best be described as dishwasher water.
The copy was a soupy mess, poured into a grimy glass, with olives-on-toothpick and an umbrella to mask the disaster.
All the creative juices weren’t worth the squeeze.
As the new campaign was launched on a Tuesday morning, at 7:47am…
I already knew it would be a spectacular flop. An embarrassment.
It was nothing but a colossal, total waste of time, energy and resources.
The ads zipped around the web, deployed across networks, fueled by an impressive media budget large enough to fund a small country.
The landing pages, videos and emails kicked into gear, ready to receive and convert the masses.
The sales team had been briefed. The support team was on standby.
Any minute now, the payment processor would smirk like Nicolas Cage, put on a pair of Ray-Bans, slide into the driver’s seat, turn the key, and roar like a 1967 Ford Shelby GT500 and take a victory lap.
And the ding-ding-ding notifications were surely about to dribble in like giddy chipmunks.
But the rest, as Hamlet said, was silence.
A bunch of clicks, a handful of page views. Zero sales.
Now fast-forward to right here, right now, as you’re reading this.
Do you know why the campaign failed?
Do you have any guesses as to why this marketing campaign sputtered, choked and died on the finish line?
It wasn’t only that too many team members had far too many opinions on the copy, and decided they should share their “thoughts”.
It wasn’t only that, all of a sudden, the developer, manager and designer were all experts on copywriting and marketing.
It wasn’t only that the company had outdated, incomplete or useless “brand” and “messaging guidelines”, based on nothing but someone’s opinion or unproven data.
It wasn’t only the usual “copy by committee”.
All these things happened.
Do you know what the biggest problem was?
Humans writing all the words.
Humans writing all the words.
Now before you choke on your bulletproof coffee.
Read this closely:
Humans writing ALL the words.
Let me explain.
You can fix all of the above and your copy could still fail to move the needle.
The problem goes deeper.
A few experienced copywriters already know that.
So their approach is simply to generate quantity—as many ads, emails, landing pages, and copy as they can come up with.
Because they know you have to write a lot of copy, and test it in the wild, to discover what converts.
That’s also why templates and headline formulas are so popular.
They’re like a primitive, lo-fi version of AI copywriting and GPT-3.
Just grab some frameworks and templates, let them guide your writing, fill in the blanks, and voila, you have copy that should convert.
But the next set of problems soon comes knocking:
If you’re too close to the product, research, and copy for too long, you’re actually less likely to spot anything that’s wrong or should be done differently.
You’re effectively “blind” to your copy.
It becomes harder and harder to separate yourself from what you’ve written, taking a step back and reevaluating.
You simply cannot see what’s wrong (or what’s right.)
That’s because your brain is working against you.
Paradoxically, this is even more debilitating if you’ve done all your research.
The “secret sauce” of successful copywriting has always been “do enough research.”
This is where you discover big ideas, hooks, the right language and words to use.
You spend hours, days and even weeks immersed in the features and benefits of the product or service you’re trying to sell.
You deep dive into customer personas, psychographics, do interviews, run surveys, and assess competitor messaging.
You do all of it. Because that’s where you (often) find the best messaging that turns into copy.
And this is all true. Most of the time.
The danger is that you’re doing all the research but don’t know what to look for.
So if a big idea is staring you in the face, you fail to recognize it.
Or someone says something in an interview that could’ve turned out to be the perfect headline. But once again you don’t realize it.
There’s such a thing as too much research.
That brings me to what’s likely the greatest leap forward for marketers and copywriters in a century.
(And, to be honest, for any writer of any form—poetry, fiction, screenwriting, and so on).
Much has been said and written about GPT-3 and AI-produced content.
Robots can now write anything—poetry, screenplays, fiction and ads—to some degree.
But the giant leap forward is not that these tools can write 50 ads or more in mere minutes.
GPT-3 is a key that unlocks unlimited creativity.
Back to the story:
As everyone filed into the conference room and sat down for the post mortem of the spectacularly failed campaign…
I quickly opened up a new tab, typed in a URL and logged in.
Everyone was expecting the usual 20-slide presentation.
But I had other plans.
You see, a few weeks before this meeting, late one night, as I was playing around with this new tool, GPT-3…
I stumbled on something that was so obvious, and so evident, that I was embarrassed to have not realized it sooner.
If there’s such a thing as dereliction of duty for copywriters, I was guilty as charged.
My team and I had experimented with various AI text-generators for several months, and had used GPT-2, as well as beta-tested GPT-3 before it was publicly released.
But we had always gone for the obvious play:
Get a ton of text, of varying degrees of quality, generated really fast. Make edits, changes and turn it into something usable.
Everyone was excited about producing 50, 100, 200 ads in minutes.
And reading the often hilariously weird output was part of the fun.
But that’s what amateurs do.
Everyone was seated, a few people joined the Zoom, and I turned on the projector.
As my browser tab came into focus, all you could see was a sparse looking text input field and a couple of buttons.
The big kahuna, the team manager (Monica), turned to me and said, “OK Sam, let’s start with your overview, a look through the data, and then--”
I leaned across the table.
And cut to the chase:
“Monica, how about we skip the data and speculation and jargon… and I just give you a demonstration of how we fix this?”
I could feel every head turn toward me.
Everything and everyone went silent.
Monica sat back, eyed me up, and said, “Uh, what do you have in mind?”
I gestured at the screen, typed a couple of sentences into the text field, and clicked one of the buttons.
In less than a minute, text started filling my screen. More came, halting, then smoothly.
I turned to everyone and said, “What you’re looking at is text generated by GPT-3. It’s doing this on its own, taking my input, and feeding back text that it thinks fits.”
Puzzled looks. Squints behind glasses. I continued.
“Out of all this output, and if I keep giving it input to get more, we can begin to identify raw copy for ads, emails, pages – basically any copy that we need.
Take a look at this text, for example…”
I singled out a blurb of text…
“This, right here, could become an ad. It’s a little weird, but honestly, at this point, what do we have to lose?”
I copied the text, opened up a Google doc, and pasted it in.
“Now, watch this…”
I took a couple of minutes to edit the text, rearranged some of it, and turned it into a Facebook ad.
I could tell that Monica wanted to jump in with something, anything, but she was a bit stumped.
I seized the opportunity to just continue with the demonstration.
With a few more edits and a read-through, it was done.
One of our Facebook media buyers was in the room.
“Mike, I’m going to Slack you this document. Can you please launch it together with one of our ad sets, right now?”
Mike opened his laptop, hesitated, and looked at Monica. She only had eyes for me, but not in a romantic way.
She turned to Mike, nodded and said “Sure, why not.”
Mike got to work.
About an hour later, I found myself hunched over my laptop with everyone in the room standing around me, looking over my shoulder.
Peering down at my screen, I waited.
A moment passed.
Then, slowly, I glanced up from the laptop.
“Do you all see this? We have our first sale.”
Now cut and fast-forward again to right here, right now.
Not only had we just turned a failed and dead campaign into something that could work…
And it did.
To the tune of $739,120 in new revenue over the course of 3 months.
Not only had we done that starting with just ONE new ad hauling in $139,522 on its own (and that spawned into 20, 30, 40 and more over several weeks).
We had done so in a way that’s not supposed to work.
Conventional thinking around advertising, copywriting and creativity demands absolute worship at the feet of research.
The Gods of media buying demands a multitude of sacrifices.
The Lords of management dictates that all copy shall be written by committee.
The ritual of research, writing, and testing shall (almost) guarantee successful outcomes, in the form of conversions.
We petition the Angels and Saints of advertising for a bountiful harvest.
And what I had just done was utter blasphemy.
Instead of going back to the mile-long research documents and folders bursting at the seams…
I had taken only a morsel and offered it up to GPT-3.
Instead of writing copy by committee, soliciting stakeholder buy-in, and reviewing it all…
I had picked raw text that looked interesting to me.
Instead of writing text on a Procrustean bed of ad templates and frameworks...
I took the raw text and turned it into an ad.
It worked. It converted. It made sales.
And this is the giant leap forward (that’s more of a side-step).
It’s a new way of being “creative” and writing copy.
It’s a different way of processing, formulating, and writing—anything. Not just ads.
The constraints are not the usual suspects of templates, formulas, research, copy by committee, and having clueless people give their “feedback.”
The constraint isn’t even testing enough ad or copy volume, hoping to find a winner.
The constraint has always been ideation—seeing things differently, getting the weird ideas, the unusual perspectives.
With GPT-3 and other AI writing tools…
If you use them correctly, you can generate ideas that can turn $1 into $100,000… and then $100,000 into $1,000,000.
And, at the same time, dig deep into our hearts and psyche in a way that changes everything.
I wrote a fiction novel with the help of GPT-3. It’s selling on Amazon under a pen name. I’ll show you some of what I did.
Stay tuned for next week.
In the meantime…
Thanks for reading Bionic Writer! Subscribe for free to get more on writing and thinking with GPT-3 and AI.
Wow, What? Crazy!
After A Lot Of Hype, (Useful) AI May Finally Be Here
Open APIs developed by open-source businesses like OpenAI, StabilityAI, and others have made it easier for businesses to use AI in their products.
It can be used for more than just the common marketing, copywriting, blogging, and visual arts applications.
AI is used by businesses like New York-based Lavender to improve the email writing skills of salesmen. Regie, a San Francisco-based company, uses it to support go-to-market teams. Call records are converted into structured data by Y Combinator-backed Pilot AI, which refreshes the CRM.
Other industries outside marketing, sales, and gaming will be impacted as it gets simpler to create things out of all the preexisting audio recordings, texts, and photos with AI.
More AI applications will probably be introduced, especially in the high-end financial, health, and legal sectors.
Twitter MEGA Thread on Amazing AI Tools
You just got to see this. Keep scrolling.
Or how about:
Do you want to see the Facebook ad?
If enough people want to see it, I’ll share it.