If the hero image on SuClasico’s home page is a showroom, glass front, a row of cars inside, the sign above the door, it does a good job letting someone we know nothing about, know who we are. It says we're a real place, with real cars and a door you can walk through.

But if we know a little more about the visitor?

Someone searches for a 1967 Ford Mustang and lands on the homepage. Someone else searches for a 1968 Chevrolet Camaro and lands on that same page. They're on the same website, but they didn't arrive with the same question. Why should they see the same images and call to actions?

The website can change it’s personality according to what the visitor appears to be looking for. The Mustang visitor sees the 1967 Fastback from inventory. The Camaro visitor sees the 68 Camaro that's in stock. If the car isn’t in stock it defaults back to the home page image, or if we don't have a reliable signal about what they're looking for, we show the normal homepage instead.

And that matters because we don’t want to pretend we know something we don't.

There is also the problem that Google doesn't normally give the website the exact search phrase someone typed. We need a signal we can actually access, such as a vehicle-specific landing URL, a campaign link, or what the visitor has looked at on the site. Each tells us something different, but none of them give us proof of intent.

Now, I happen to like the showroom photograph. It's better composed, it's warmer, and it communicates something the car doesn't. But the question isn't which image I prefer. It's which image does the job this visitor needs it to do.

That's the difference between making a decision and having an opinion.

Every design decision has three parts: a purpose, a tradeoff, and a consequence. If I can't name all three, I'm not making a design decision. I'm decorating, and decorating is fine as long as nobody's paying for it.

The purpose. The hero's job is to help the visitor recognize that they're in the right place. If we have a reliable signal that someone is looking for a particular car, a matching vehicle can do that faster than a general photograph of the dealership. The visitor doesn't have to read a headline or work out where to go next. They see something relevant immediately.

The tradeoff. Personalization adds complexity. We need to identify a useful signal, match it to a vehicle in the inventory, and decide what happens when that vehicle is sold or the signal is missing. We also give up some consistency: two people visiting the same homepage may see different images. That's fine if the difference helps each person, but personalization for its own sake is just another thing to maintain.

The consequence. The homepage can no longer depend on one fixed hero image. It needs a sensible default and rules for choosing a different image when the available evidence supports it. If the Mustang sells, the site can’t keep presenting it as available. If the visitor's intent is unclear, we shouldn't guess. And if we personalize the image, the headline and next step should make sense with it. Showing a Camaro above a headline about finding your Mustang would only create a new problem.

Purpose: capture an email address by giving the visitor something worth saving. A configuration they built themselves, with their own engine and paint and wheels, is worth coming back to in a way that a newsletter signup never will.

Tradeoff: it's real development work, SuClasico is a small dealership, the tool becomes a line item the client has to agree to before I promise it, and it changes the VDP's job from answering questions to selling the tool. Consequence: if nobody maintains it prices change, parts go out of stock, the paint options shift and it becomes a liability rather than an asset, and a broken configurator is worse than no configurator.

Notice that the consequence isn't "the VDP changes." The consequence is "someone has to own this tool for as long as it's live." That's a staffing question, not a design question, and it belongs in the conversation before the design starts rather than after it. Naming it is how it gets into the conversation.

You can’t do this for every button color. Most decisions are cheap to reverse. A button color you can change in an afternoon. A hero image you can swap in an hour. The framework is for decisions that are expensive to undo: the navigation structure, the order of the pages, the tools you promise the client, the things that end up in a contract or in somebody's job description.

So the first question isn't "what's the purpose of this decision." The first question is "is this decision reversible." If it is, pick a sensible default and move on, because spending an hour on a reversible decision is an hour you didn't spend on an irreversible one. If it isn't reversible, write the three sentences.

That test is also how I separate a design decision from a preference, at least in the first instance. A preference is about me. "I like the showroom photograph." A design decision is about the visitor and the business. "When we have a reliable indication that the visitor wants a Mustang, showing an available Mustang gives them a more relevant starting point than showing the showroom." The second one has a purpose, a tradeoff, and a consequence. The first one has a mood.

The trap is that preferences are easier. They feel like decisions. They take five seconds, and something changes on the screen, and it feels like progress. And they accumulate. A site built from a hundred preferences looks like something. It doesn't do anything. A site built from twenty decisions, each with a named purpose and a named cost, does the job it was built for.

But here's where I have to be honest about the limits of what I've just given you.

Having three named parts is not the same as having a good reason. You can fill in purpose, tradeoff, and consequence for a bad decision as easily as a good one, and the form will look identical. The framework tells you whether you've thought about all three. It doesn't tell you whether the thinking is any good.

That's the harder question!