1. Say what it does, then say it with style

Insight from Joey Noble and Devon Reynolds of Demand Curve.

Devon and I work on a lot of the same DC accounts, mostly through our landing pages and story system work. One framework keeps coming back up whenever we're using copy to lift conversion:

Conversion = Desire - Labor - Confusion

People convert when they want the outcome badly enough (high desire), don't feel like getting it takes too much work (low labor), and understand what they're getting (low confusion). There are more variables in reality, but as a way to think about maximizing conversion on a page, it holds up.

Labor and confusion are the easy ones to spot. Desire is trickier, and we think it's the most important of the three. Most conversion advice is about removing friction, which is just another way of saying labor and confusion. That work is worth doing. But if people don't want the thing in the first place, a perfect page won't save you. And if they really want it, you can get plenty wrong and still get the sale. Raise desire and the returns dwarf anything you'll get from cutting a form field.

So that's where we spend our time with clients, and the first move is always the same. Make the outcome easy to picture. It's hard to want something you can't see.

Which means a concrete outcome instead of a vague promise. Most pages start somewhere like this:

• A fitness app: "Unlock your potential."
• A productivity tool: "Work smarter."
• A music player: "The ultimate music experience."

These starting points are too abstract. You can't picture unlocking your potential, and you can't measure working smarter.

Compare that to Apple's "1,000 songs in your pocket." "1,000" is specific and measurable. "Songs" tells you exactly what you get. "In your pocket" is a physical image you can picture.

This relates to what psychologists call the concreteness effect. Concrete information is easier to understand and remember than abstract information, because people can form a mental representation of it. Which means they can picture themselves getting the outcome before they've clicked anything.

Here's the order I use. Say the outcome plainly first, before you try to make it sound good. Plainly, it's something like "carry your entire music library with you." That's more useful than "the ultimate music experience," because it names what you get. Then you go find the sharper version, which in Apple's case was "1,000 songs in your pocket."

Style can make a specific promise memorable, but it can't make a vague promise specific. So say what it does first, then rewrite it with style.

A few more, same two steps:

• "Take control of your inbox" becomes "Get to an empty inbox before lunch."
• "Streamline your workflow" becomes "See who's blocking what, on one screen."
• "Simplify your finances" becomes "Know your exact runway on the first of every month."
• "Best-in-class support" becomes "A real person replies in under five minutes."

Obviously, your specific promise needs to be something you can actually deliver on.

Go through your page and mark every promise you can't picture. Write the flat version of each one underneath it, then sharpen from there.

2. Choose the category your advantage is worth most in

Insight from Nick Costelloe of Demand Curve, inspired by Jeff Bezos at the 1997 Special Libraries Conference.

I’m not a fan of Jeff Bezos. But I do a lot of marketing research, so I came across this fascinating 1997 talk with a lesson that stuck.

Some quick context. In 1994 Bezos was working at a quantitative hedge fund in New York when he found the number that started Amazon: web usage growing at 2,300% a year. He wrote out a list of about 20 products you could sell online. Then … he picked one.

Why books? Because of a structural fact about the category: there were more than 3 million books in print worldwide at the time. Music was the runner-up, at about 200,000 active CDs. No physical store could hold a meaningful fraction of 3 million. As he put it, that many items lets you build a store online that couldn't exist any other way.

Here's what I found most interesting. Choosing what to sell is a growth decision.

It's a great example of product-channel fit. The Five Fits splits your business into the parts you can engineer and the parts you can't. Product, model, and story are yours. Market and channel are not. So the job is to shape what you control around what you don't.

Most of us run that backwards. We build a product, then go looking for a channel that will work for it. Bezos ran it in the right order. He started with the channel, an online store with effectively unlimited shelf space, and worked back to the product that fit it best.

A structural advantage isn't spread evenly across categories. Unlimited shelf space is worth almost nothing where there are a handful of SKUs, and it's enormous where there are 3 million. Books were where that gap between what he could do online and what a physical store could do was widest.

Sure, this is easier to act on if you're starting from zero. But even if you already have a product and a market, you can run the same test on whatever advantage you actually have. If your edge is speed, the category to attack is the one where the incumbent is slowest and customers feel it every week. If your edge is that you can serve a single customer profitably, go where the competition needs a hundred-seat minimum to bother. Instead of focusing only on what you're good at, ask where being good at that thing is worth the most.

3. Exclude the people who look like your refunders

Insight from Prash Brooks, who runs Paid Marketing for Demand Curve clients.

Some of my accounts have been getting a lot of low-quality traffic from Meta lately: bots and buyers who refund and return at high rates.

The old fix was to tighten who you target. That option keeps shrinking, because more spend moves into Advantage+ and Meta decides who sees your ads. Which leaves you with an odd split. You've lost most of your control over who gets in, but you still have all of your control over who stays out.

So I've been steering from the other side. Here's one thing I'd recommend trying if you're noticing high return rates. Build an audience of your existing refunders and returners. Then create a 1% lookalike off it, the people Meta thinks resemble your returners. Add both as exclusions. You're blocking the people who send products back, plus the ones who pattern-match to them.

I've been running this for several years across accounts. My ballpark is around a 5% improvement in ROAS. Not a huge transformation, but this can run in the background of every campaign once it's built.

One tip: keep the seed list small. Between 1,000 and 1,500 people is the range I aim for. Load in every refunder you've ever had and the lookalike comes back noisy and unfocused, which defeats the point.

News you can use

• YouTube now counts a view the moment a video starts playing, on long-form, Shorts, and live. Public counts will climb without anyone watching more, and the old metric stays in Analytics as "Engaged views." The switch is Aug 24, so if you report on video, move your dashboards to engaged views and treat that date as a break in the series. (TechCrunch)
• ChatGPT Ads switched Automatic Advanced Matching on for existing web pixels on Aug 17. It hashes email and phone numbers from your site forms to match more conversions, and it's on now unless you opted out before the deadline. If you're running ChatGPT ads, check Tools > Conversions > Data Source so it's your call and not a default. (Search Engine Roundtable)
• Google's generative AI report in Search Console is reaching a lot more properties, with a pop-up inside Search Console pointing people to it as of Aug 11. Google still says not every property has access. The report splits your impressions in AI Overviews and AI Mode out from regular organic, and there's still no click data, so track AI impressions on their own line instead of reading them like organic. (Search Engine Watch)