Hannah Parvaz, Founder, Aperture first appeared on a Turing stage at Turing X SYSTM Growth Summit in November of 2024. She gave an awesome, inspiring talk full of takeaways for why you should focus on your creative and experimentation to get the real bang for your buck on social media.
Naturally, we had to have her speak at Turing Fest 2025. Here is that talk:
Stop Asking AI to Write Your Ads (and start using it to find your audience)
In a world obsessed with AI-generated content, most companies are missing AI's true power: its ability to validate and scale real customer insights. The best marketing doesn't start with prompts - it starts with customer psychology. This session reveals how to combine deep customer understanding with advertising algorithms' intelligence to find product-market fit and build sustainable growth engines. Learn how to turn customer conversations into testable hypotheses and let AI validate what resonates at scale.
Auto-generated transcript - may contain errors.
Tap a timestamp to jump the video.
Everyone, hello, hello. Great to be here. Thank you for having me. This is my QR code. I can see one brave person scanning there. That's early adopter energy. So I love that. The rest of you, I think at the end, you'll want to scan that because it's got all of the slides, it's got how we reduced CAC for some companies by up to ninety three percent.
So, I'm Hannah, as Helda said, so thank you for the intro. I run a growth agency called Aperture. So half the business is performance. The other half is focused on product strategy, full funnel optimization overall. We work with brands, apps, SaaS companies, and don't just spend their money, but help them scale very profitably.
Before that, I spent a decade in house running growth at four different companies, scaling them all up to millions of users. And now I've worked with over three hundred companies. And most of the things that I learned came from either fixing things that were already on fire or just building them myself.
So the old playbook is broken for paid. The last few years have been like building on quicksand with everything changing constantly. And if you have been advertising for a long time, and I've been advertising for about twelve years now on these platforms, and actually I got started on Twitter, but we don't really talk about Twitter anymore.
But if you have been running ads for that long, you will know that things have changed. So just so I can get some context, who here was running any ads before twenty twenty one? Okay, okay, okay, okay. So you remember probably when ads used to just work.
And if you had your hands up and you're still running ads now, who are you? Since twenty twenty one. Okay. So you guys are the survivors. You stared into the algorithmic abyss and said, not today, I'll be fine. So if you've been in the game that long, you'll see, you know, your CPMs have going up.
They have tripled in the last couple of years. Companies like TEMU have been putting in billions into these platforms. So just entry point to where you're starting has increased three, four x in certain geos. That means your returns are down. That means you've got less control.
You're doing a lot more guessing. And on top of that, traditional targeting is dying. And there has been a slew of different reasons for that, from Apple's fun changes to privacy in twenty twenty one to Cambridge Analytica making everyone shit the bed, to AI.
And ultimately, what this means is the old way of running campaigns is dying too. But here's the secret. It isn't just the algorithm. That is one of the reasons the platforms have been restricting the hell out of things. Like you literally can't do the targeting that you used to be able to do.
But one of the reasons that everything is changing is because, well, people are getting really bored. They have seen your ads over and over again and they have just stopped giving a shit. They don't care anymore. Whenever we think we know the algorithm better than whenever we think we know what to target better than the algorithm, we are usually wrong, at least in the long term.
And that is because really the algorithms have been built for this. They are there to do that one job, to do that targeting, and it is up to us to put in what is going to work. We are not just fighting the algorithms.
We're fighting for human attention. And so last year, I want to just break this down. Last year I took over one ad account that had one hundred and forty two live campaigns in it. And if you don't really realize how many that is, that is an unbelievably crazy insane amount of campaigns to have in one account, for that much spend.
And this company had been never had been profitable in the whole many years that it had been alive. Never. And so what were they doing? They were just taking those same creatives and launching them with new audiences again and again and again and again.
They actually hadn't added any new creatives for over six months. It was I did whiplash. I had whiplash when I was looking at it because it was just so crazy to see this. So what we did was we cut this down. Know, more campaigns does not mean better results.
It just means usually more confusion, more reporting, more time wasted thinking maybe something might work here. And to be honest, most of the time this is not going to work. You can't outsmart bad creative with your campaign structure and you can't optimize your way out of a relevant creative.
So what we did was we cut that one hundred and forty two campaigns down, we chucked all of them in the bin and we relaunched with four, and within three months only, they were profitable for the first time in their entire business history.
So, that was not all we did. We didn't just cut it down, but instead of spending all that random budget across dozens of random audiences, we were really focused. You know, we were then able to focus on the creatives. So that was not what we did and I'm gonna take you through it.
So this was a supplement brand. So let me show you why this shift mattered so much. Most companies are targeting with one fixed message. They are just saying the same thing over and over again. They're pushing with just one message and when you're doing that, you're only going to reach people who care about that one specific message.
So take a supplement brand like this company. Whenever we started working with them, all their ads just said, we have vitamins, we have vitamins, we have vitamins, we have vitamins. And was and actually people don't care about the vitamins, they care about what it's going to how it's going to impact them.
So whenever we start thinking about these other messages, you know, better skin, better immune system, have better sleep, this product has vegan nutrition. This is where we're going to be able to tap into other audiences and expand. So here's the thing, most of your potential customers aren't ignoring your product, they're ignoring your message.
Anybody could be your customer, anybody, but only if you speak to what actually matters to them. You can segment your audiences all day long, but if you're only saying one thing, you're only going to reach one slice of that audience. And motivation, this is what we have to tap into these different motivations and this is what unlocks scale and your motivations lie inside your customer conversations.
So what do we do when our traditional targeting stops working? We stop chasing audience hacks and we start focusing on our creative. So what makes our creative effective? And now that we have got incredible AI at our fingertips, what I'm seeing now is the same kind of dry, dead, lifeless copy coming out again and again and again.
And it can help get you some really good scale. But actually, it's only as powerful as what we're putting in. So what is making our creative effective? It's not prompts, it's not chat GPT and it's definitely not vibes, right? Like we actually have to talk to our customer, have customer conversations and get our customer insights.
You can't prompt your way to product market fit. So before we start worrying about creatives or targeting, we need to make sure we're focusing on finding or already have our product market fit. But if you're not on your way there, none of this is going to matter anyway because when you're getting the right Even if you're getting the right people in, they're not going to convert.
First, if you don't know what product market fit is, you're probably in the wrong room, but just in case, let's break it down. So Marc Andreessen said in two thousand seven when he coined that term that product market fit is being in a good market with a product that can satisfy that market.
But I said product market fit is when you found a need in the market and a solution that customers actively want to buy. And the important word here is actively. Oh, it disappeared. They're not just clicking. They're not just saying they like it in a survey.
They're pulling out their wallets, their credit cards, their Apple Pay, and they're actually buying it. So next, this is the shift that most teams miss. Your ads are not just ads, your creatives are not just ads. Each one is a tiny little experiment that you can run to find something out.
It's a tiny customer interview disguised as a video or as a headline because whenever you run performance, you're not just testing your performance when you're running your creative, you're testing resonance as well. And here's the thing, Meta for example, this works across all of the advertising platforms, but Meta for example in twenty fourteen, so eleven years ago, ran a study of eighty six thousand people that tested if they engaged with three hundred things on Meta on Facebook at at the time, if it would know you better than your spouse.
And it did. So that was eleven years ago. So now what does it why is it gonna take like a few likes on something, a couple comments, one real, you know, the TikTok algo gets to know you after just a few swipes, what you're going to be into.
So us thinking that we're going to be able to guess what interests are most relevant doesn't work. We're targeting broad. How does Meta, for example, know this? It has hundreds of billions of data points. It's tracking every user, every action across years, across platforms.
It's not just listening, but it is learning from you constantly. So if we break that down for a second, Meta is literally tracking and these platforms are tracking what you buy, everything you click on, whatever you're watching, how much of that you're watching, how you behave online because the all of these websites have the Facebook pixel on it.
So it can see what you're doing on every single website. What you say near your phone is listening. And also what you're commenting on captions and on other people's videos and everything. Actually, I even saw an article recently that said that Meta has been sending people beauty ads whenever you delete a selfie, which is crazy.
So this is the power that the platform has because it knows that people are going to convert better at that time. So we can use these platforms to answer which customer problems resonate strongest, how can we best articulate our solution and what benefits will actually drive purchases.
We can use it to test our strategy, our messaging, our positioning. And we've rebranded entire companies because of their ad performance when they've been at an early stage. So how do we flip this switch? We stop obsessing over targeting and let the creative do the targeting for us.
We let the platforms do what we are built to do and we feed them with inputs that actually work. We stop asking who do I want to reach and ask ourselves what message will land with the person that I want them to resonate with.
We can think of this as well like jobs to be done. So here is what happened when we stopped chasing audiences for one of our companies and focused on the messaging. For this company, we followed the process that I've been walking you through, but I'm also going to take you through some real examples.
Over a few months, we reduced CPA by sixty seven percent while maintaining a consistent investment spend. No hacks, no budget increases, just better creatives and stronger signals. This was another one, there's a little quote from someone. This time we dropped CPA by seventy nine percent and this company wasn't able to maintain that reduction.
And we didn't just reduce the costs there, built a really strong foundation that we've been able to continue iterating on. It was continuing to trend down. And again, it didn't start with targeting, we removed all targeting. It started with our better creatives. And I'm going to take you through now for a financial education product, a FinTech product, how we reduced their CAC by ninety three percent with all of the actual examples that we used in production.
So from the first insights that we discovered to the actual creatives and how we scaled this all step by step. So here is the framework. Feel free, take pictures. I love it when people take pictures. So go ahead. It's really simple. Three steps.
It's bit more complicated than this in practice, but you can break it down like this. So one, develop your themes based on your real customer insights. Simple, talk to people. Two, test those themes really creatively. And three, iterate and then scale what works.
We do not treat creative like it's a one off production task. We treat it like it's an insight in learn, refine, repeat. It's a growth loop. So step one is insight. And this is where most teams honestly go wrong. And we have so many companies coming to us saying, oh, we've actually never spoken to one of our customers, literally.
And I've done a Over the years I've been keeping track of this, and over seventy percent of the founders I talk to have never spoken to a customer, especially in the last year. So this is where most teams go wrong. They want to focus on the features a lot of the time instead of the benefits of their product, and they skip talking to their customer.
But if you want to find creatives that actually land, you need to know what your customers are actually thinking before they see your ad. So before we start testing anything, we start by listening, and that means real conversations, not AI prompts. You don't need hundreds of conversations or hundreds of interviews to find something useful, even five to ten very deep conversations.
And sometimes these people end up crying in these conversations because we go so deep, not because I'm a dick to them though. They will start to uncover patterns. What you're listening for is really simple. What do they want? You know, we're thinking about Ryan's talk before, where are they and what are they trying to achieve?
What's holding them back and what finally made them act? And this is where your best creative comes from. It's not from product features. Actually, people don't care about your features. They care about how it makes them feel. These creatives should come from their lived experiences.
So after that, we literally just put them into two columns, start mapping everything into goals and struggles, into motivations and barriers. And from there, what we do is we pick the themes that feel the most distinct and actionable. The ones that we can actually build a message around.
And then we get into our test and learn section. So we will need per week, and usually we start with a few different weeks of testing at the beginning just to refine our messages. We need around three to four at least of these different distinct messages.
So I'm always wanting to start with kind of eight to ten messages at least. So I've got two weeks of concept testing to begin with. And I'm gonna show you now exactly how this plays out in practice. These are some of the rules if you're on meta, but they vary based on different platforms.
So fifty conversions per variant, seven days minimum for your testing period. So here is exactly what we tested with one company. This was a FinTech product called Juno. All we changed here was the messaging. So you can see there, exact same message, learn about X and get money confident in just five minutes a day.
And all we changed was one or two words, salary negotiation, investing, budgeting, same format, same layout, same offer, just different, slightly different angles on top based on what we had heard in the research about why people were already using this product. We didn't guess, we didn't say, let's try a fun meme, which is what sometimes there's the creatives that we get delivered.
And this made us drop CPA by ninety three percent. It was significant. Not just these creatives didn't drop ninety three percent, but I'm gonna show you the flow. I always say start with statics for these messaging tests because they're super easy to put together, you don't really need to think too hard about them, you can knock them out really quickly and all you need to do is change the copy on one design.
And here is how to set that test up in platform. We start with one campaign, one ad set and one ad per ad set. So you've got an ad set per your creative theme and this way every single message gets equal chance. And we can clearly see which one performs best.
This is where a lot of people mess up their tests. They will often put all of the creatives together in one ad set, and that is not what we want to do. We want to, at this stage, remove spend as a variable. If one ad is getting more spend than others, your results are going to be skewed before they even start.
So to be clear, one campaign, several ad sets, and one creative per ad set. The only thing that should vary is your message. So now we've got the results. So here we can see that the creative with budgeting reduced the average CPA cost by twenty seven percent.
So this was our winner. One of them was slightly better, slightly less performing, less efficient, the salary negotiation one. And then investing actually was not what people were interested in at all. So this is the moment at which we stop testing these creatives, that message specifically, and start scaling them.
So we take that winner and we put it into an evergreen always on campaign and allow that one to start optimizing. And in parallel, we take that creative and we start then moving it into the iteration phase. So we take it and this is where this real kind of compounding value kicks in.
We start iterating with different formats, different angles, different designs, eventually into different funnel steps, whatever helps us build around that core message. And eventually that could be landing pages and things like this too or flows inside your product. So we're not guessing, we're refining here based on what the audience has told us with their clicks and with their conversions.
Here is what this looks like in practice. So we took that message and very simply just put it onto a few different designs to understand, okay, if we get a new design, will that help us with our performance? And will that give us a kind of new creative container to be our baseline creative container to start with?
So we're testing similar styles and tones, but the message is consistent. That's what we've already validated. So, this is how we structure the next round of testing. So you'll notice now we've got this control section and that is using that control creative there.
So again, we've removed spend as a variable at this stage because we're trying to get very clean learnings, and we're adding in our original winner back in as control, and that allows us to compare performance. So our best performing previous creative against our new ones.
So to see if our new ideas are actually better, not just newer. So whenever we did this, we found that this one reduced the CPA even further by twenty three percent more, and the one where the woman is holding a jar of money increased the CPA nearly double.
So sometimes these are just not going to work, and we can't guess. Like if we would have just put all of these into our always on campaign, it could have tanked it. So we want our always on campaign to be predictable, repeatable, guess, like we wanna feel calm and confident about that one while we're doing our experimentation in a controlled environment.
So we're not looking for something that just hits the benchmark ideally, we want to hit something that outperforms it. So twenty three percent there was really great. Once we have validated that visual, then we went back into some messaging testing. So we took that winning creative.
Again, we tested it with the control in that structure. So, we had four ad sets this time. Same design, same format, but with copy. So, each of them are speaking to slightly different emotional angles, empowerment, a little bit of humor, rebellion. And this is what happened this time.
So, we found another winner this time and I was writing all this copy, but I don't actually I didn't write that one. I have to say I stole it from Cher. But it worked and it reduced CPA by twenty seven percent. So, it spoke to the emotional hook that mattered to the audience and it performed so well.
Next, we brought it back to full circle. So, we went back to the customers. We were starting to understand exactly what they were looking for through our ads. So we ran a quick survey just via email with a type form, people about how much more they had been earning, about if they'd been saving, about the impact of the product on their lives.
And we pulled out a few different metrics and this was the average. People who used the product were making four thousand four hundred pounds more per year, self reported. And people who were using the product saved seventeen hundred pounds more per year, self reported.
So we took that real stats, put that into our ads, and we reduced CPA by thirty eight percent further. So this was working extremely well. That second variation about the savings, we actually thought this one would perform better because we were talking about budgeting and so on.
But this is why it's so important to test these in this controlled environment before putting them into your main campaign. Once we validated in static, and I have included this because everyone always asks, how do we get into video after this? Should we just start with video?
I never start with video because it's higher production costs than just making a quick static. You can actually just write into chat GPT now, make me the static with this copy and it will do a great job. But what we do is we move into more complex formats after we've started to validate things, but the process stays the exact same.
We take our winning themes and we build three to four hook variations around it. I'm sure you all know what a hook is, but just in case, that's the first kind of three seconds at the beginning of a video that gets people to stop.
And actually, more than ninety percent of people are usually gonna be only watching that portion of the video. So, need to spend our time focusing on this. It really doesn't matter what's going on in the body or the call to action at the end of the video at this stage because ultimately no one is going to get there anyway.
And that makes it also really easy to isolate what's working. So, we've got our three hooks or sometimes we're actually briefing in like ten hooks literally because that is the most important bit so that we can start to understand. So these are some examples.
These two videos use the same creator as you can see, same format, same script, same body, and the only thing that we changed was the hook at the first few seconds, and this is for this company as well. One of them cut CPA in half, Bless you.
And the other more than doubled it. So I am so broke. I literally can't make my money last. That's the hook. I'm so broke I can't seem to make my money last. And the other one? I'm so broke. I've been trying to save money this past month and I haven't been able to.
That's it. They even start with the same couple of words like I'm so broke. But one of them huffed the control cost and the other doubled it. Does anyone guess why that happened? A little bit longer, but I'll play it again. I am so broke. I literally can't make my money last.
Okay. This is the one that halved the CPA. This one doubled up. I'm so broke. I've been trying to save money this past month and I haven't been able to. Yeah, both of these were conversation style UGC videos. So, she like changed her outfit between them.
The reason that second one performed worse is because she starts by going and by the time she's done that, everyone has already swiped away. Like, no one cared at that point, because you're not starting with something very interesting, even though it goes into the same video content essentially afterwards.
So that's why it's so important to get those first, even second right. And one thing that we do now also, I haven't included this here, but as we always start with some text on the screen describing what the video is and who it's for, And we keep that there and we've seen that improve performance dramatically, because it gives someone a preview in their minds of what's going to be working.
So, this is how powerful hook testing is. So, once you found a winning creator and hook, don't move on. You wanna go deeper here. I always want to work with that same creator again, and again, and again. Because we want to build deep relationships with them.
I want them to feel really positive about the brand, and even post about it on their organic pages. Don't know why I said pages. This is where we build real performance consistency. So by creating hooks, by using the same structure, and building a reliable bench of creators who convert.
So, it isn't about making more, more, more, more content all of the time. Creative velocity is a very important thing, but it's not just about more content, it's about scaling the message and the messenger. So this is how you go from performance spikes to repeatable, predictable results.
And if you do want to introduce a new creator, you take that winning script, that winning hook, and then you start to introduce it with a new face. So our top performing script. I am so broke. I literally can't make my money last.
And then test it with a new face. Same words, same energy, new person. So broke, just can't seem to make my money last. And if it's going to perform at similar or better rate, then we can start to introduce those people into creatives.
So this one with this lady, she's Scottish, dropped CPA by seven percent. So this is how it looked, the full journey fully laid out from start to finish. I'll pause because I see lots of people taking pictures, but we didn't reinvent anything, you know, we didn't reinvent the wheel here.
We kept asking just what worked, what didn't, what have we learned, what should we try next? We move from messaging to design, to hook, to creators, and every time we tested something, we let that feed into the next round. So this is what creative iteration actually looks like.
It's not just guesswork, it is structured momentum. And here is the result. Over this time, this testing loop drove a reduction in CPA of ninety three percent without any huge spikes in spend or any black magic audience hack because we scaled with clarity, with the message format face hook, all tested, all proven.
If you're thinking this might only work for consumer products, because I mentioned matter a few times or DTC, no, we've worked with all kinds of companies across cybersecurity, kids products, supplement brands, gaming, meditation, B2B SaaS, sleep products, all of this kind of stuff, wearables.
The system still works because this is not about product category. It's about creative strategy. And that means you. You don't need a huge budget. You don't need a ten or a hundred person creative team. Like we're a tiny company and we're doing this for multiple companies because we have got our system set up.
We're using automation where we can to do the building and so on as well and the reporting. You just need to stop guessing and start testing and build around what works. So the main takeaways, if you just follow these six things, then you're going to have some good results.
So talk to your customers and identify eight to ten themes. Simple, create static images with unique differentiated messages. Give them the same budget, very important, and test them. Analyze, identify winners and iterate. Use a control to measure performance, very important. And more campaigns does not mean better results.
It does not. Consolidate. Consolidation is king. So stop pouring money away without learning anything meaningful. So, we build ads, we build systems, we scale what works. And if you want these slides, if you scan that, I will send them to you. So, I'm gonna be doing a round table in that next room straight after this.
So, if you have any questions about how you can apply this or anything like that, come say hi, or you can send me an email at any time, and that QR code has the slides on it. So, thank you so much for having me.
It was a pleasure, and I will see you in the other room.
