Run a flight search on a travel site like Kayak and, while it works, it shows you a little readout: searching Delta, searching United, checking hundreds more sites. It could just hand you the results. It shows you the search instead.

That choice runs against almost everything we tell ourselves about building software. Shave the milliseconds. Kill the spinner. Make it instant. Most of the time that instinct is correct. But there is a specific, well-studied case where the instant result is the worse product, and it has a name.

The study that put a number on it

In 2011, two Harvard Business School researchers, Ryan Buell and Michael Norton, published a study in Management Science called The Labor Illusion. Across five experiments built around online travel and online dating, they found that people often preferred a website that made them wait while it visibly did work over one that returned the identical results instantly. When the site showed its effort, participants rated the service roughly 8% more valuable. Same answer. Higher perceived worth, purely because they watched the labor behind it.

The reason is not patience. It is reciprocity. When you can see someone working on your behalf, you assume they care, and you value the result more. An instant answer triggers the opposite suspicion: did it actually check anything, or did it just hand me whatever was already sitting in a cache? Buell later widened this into an idea he calls operational transparency in a 2019 Harvard Business Review piece: deliberately building windows into your process so customers can appreciate the work being done for them. The finding held up. People end up more satisfied, more willing to pay, and more loyal when they can watch the hidden work happen.

You have already felt this

You have felt this even if you never named it. The Domino's pizza tracker does not make your pizza arrive one second sooner. It turns the wait into a service instead of a void. The skeleton screens that flash on Facebook and LinkedIn before real content loads are doing the same job, signaling that something is being assembled for you rather than that the app has frozen. A progress bar that names what it is doing (compressing image, uploading, finalizing) reads as competence. A blank pause reads as a bug.

The AI "thinking" animation is a trust device

The place this is playing out most visibly right now is AI. Open almost any AI chat tool and it shows you a running account of its reasoning: searching, thinking, checking sources, drafting. Some of that readout is genuinely useful. A lot of it is the labor illusion doing exactly what Buell and Norton described. An assistant that pauses and narrates a few steps feels more trustworthy than one that fires back an answer in a blink, even when the fast answer is just as good. The thinking animation is not only a status indicator. It is a trust device.

Where transparency turns into manipulation

This is easy to misuse, so it is worth being precise about the line. The study is not permission to add fake delays and pretend to work. Two things separate operational transparency from manipulation. First, the labor has to be real. If you stall a checkout for two seconds and animate a spinner over nothing, you have not built trust. You have added friction and lied about it. Buell's transparency works because the work exists and you are revealing it, not inventing it. Second, there is a ceiling. The effect helps most when a wait already exists and you are giving it meaning. Past a point, more delay is just delay, and the goodwill curdles into irritation. Nobody thanks you for a loading screen bolted onto a button that should have been instant.

Design the wait on purpose

For most of the sites and tools we build, the practical lesson is narrower and more useful than "make things slower." It is this: the moments when your software is actually working are design opportunities, not dead air to be minimized. A search that takes a beat can say what it is searching. A contact form can confirm it is checking something before it says thanks. An AI feature can show the step it is on instead of freezing on a gray disc. A quote tool can narrate the inputs it is weighing. None of that requires slowing anything down. It requires deciding that the wait is part of the interface instead of an accident to hide.

That is the part that usually gets skipped. When a team builds a feature, the happy path gets the attention and the loading state gets a default spinner dropped in at the very end, if it gets anything at all. But the loading state is where a real chunk of your user's trust is won or lost, which makes it a design question as much as an engineering one. We build it that way. The same care that goes into what your site says when everything works belongs in what it says while it is thinking, because that pause is a first impression too. We made a related argument before about how the first impression your website makes isn't your homepage. This is a cousin of it. The impression a product makes in its in-between moments does quiet, constant work on whether people believe you, and belief is usually what separates a site that gets clients from one that just gets visitors.

The instinct to make everything instant is a good instinct that occasionally lies to you. Speed is a feature. So is the visible evidence that something worthwhile is happening on the other side of the screen. The skill is knowing which one the moment calls for, and building the wait on purpose instead of leaving it to a default. A site that clearly does more can be worth more, even when the output is identical. Buell and Norton put a number on it. The rest is design.