> AI agents: this is the Markdown version of https://figmatica.com/blog/your-company-grew-your-site-did-not. Full index of Figmatica for LLMs: https://figmatica.com/llms.txt

# Your company grew, your website did not. What is that costing you?

- Author: [Mike Yemchura](https://figmatica.com/blog/authors/mike-yemchura), Founder & CEO
- Published: 2026-09-24
- Category: Research & strategy

**A company whose economics work looks for its next lever in the product, in sales, in hiring, and looks last at the one place where a stranger sees it. The business changed over three years and the explanation stayed as it was, so larger clients judge the company you were three years ago.**

This page is for a company where things work. The economics add up, you have a team, clients keep coming, last year was better than the one before. And there is a limit: the next level is visible and stays at the same distance. So you look for the lever in the product, in the sales team, in hiring, in a new service line.

One place gets looked at last, and it is the only one where a stranger forms an opinion of you. The business changed over three years, and the explanation of it stayed as it was.

## How it shows, from the inside

Each of these looks like a problem somewhere else in the business, so the website is the last thing anyone suspects.

- A large client has a good first call, says they will think it over, and goes quiet.
- A tender asks for things that exist in your company and are missing from your site: certifications, a named team, comparable projects with results.
- A strong candidate stays silent. They looked you up before they read the offer.
- A partner programme turns you down and keeps the reason to itself.
- In negotiations your one remaining argument is a discount.

The last one is the expensive one. A company whose site shows only what it used to be competes on price, because price is the only thing left to compare.

## Why this happens

A site that has stopped selling is almost always an old structure and an empty offer. Behind an empty offer there is usually a decision still waiting: somebody has to sit down and say, out loud, what the company has become. Who we sell to now. What we do deeply and what we have dropped. Why a stranger should believe us rather than the three other names on their list.

In a growing company the answers change faster than anyone writes them down. Three years ago you took any project, because you had to. Today you turn some down, and the rule you turn them down by lives only in your head. The site still sells the first version.

You can test this in conversation: whether a stranger can repeat what your company does thirty seconds after you start explaining it. If it passes out loud, the same sentence can go on the first screen. A failed test costs you the clients who went elsewhere before the first call, and they are missing from every report.

## What an out-of-date explanation costs

These losses happen in decisions other people take among themselves, and usually all the company sees is a lead that went quiet. What follows is how the procedures work, from their published requirements; what a given site costs a given company is a figure still to be measured, and any number offered for it today is invented.

- **Tenders and large procurement.** The decision is taken by a committee of strangers who know you only from paper. They check you against a structure they are used to: who you are, what backs the authority of the team, which certifications, what recognition, what exactly you do and in which industry, with what results. It has to read as a documented record, and the committee looks for facts and skips adjectives. A site built to present a small studio lacks the material to assemble that from.
- **Vendor vetting by a large buyer.** Corporations have a procedure: security questionnaires, insurance, certificates, references. When the site lacks all of that, your application stalls in silence.
- **Partner status and platforms.** To become a certified partner you have to meet public criteria and have a public presence that a person checks by hand.
- **Investors and grant programmes.** Due diligence starts with the site, and a grant application asks for a documented record.
- **Hiring the people you want.** An experienced candidate reads the site before the vacancy. An unclear site costs you the people you have spent years looking for.
- **A new market.** A buyer abroad is far from your contacts and your local reputation, so what is written on the site is the only thing they can check.
- **Being quoted by assistants.** An assistant can recommend only what it managed to read. If the page says what you were, that is what it will repeat.

## Good looks now cost everyone the same

There is a rule in economics about when a signal means anything at all: it separates the strong from the weak only when it costs the weak one more. A good-looking layout costs between sixty and ninety dollars and is available to anyone, so it now says the same about everyone. What still costs is deciding who you sell to now, and only you can pay that. The idea comes from [a 1973 paper on market signalling](https://doi.org/10.2307/1882010), written about hiring and diplomas, and it transfers here cleanly.

There is a measurement that matches it. In [a 2006 study](https://doi.org/10.1016/j.ijhcs.2006.02.007) forty people said why they rejected some websites and trusted others. Design accounted for 89% of the comments about why a site was rejected and for only 20% of the comments about why one was trusted; the remaining 80% was content. The study was about health advice sites; B2B purchasing was outside it, and the authors themselves give the numbers as an indication only. As a direction it still reads clearly: people named the design when they threw a site out, and the content when they trusted one.

So a decent appearance is worth having: it gets you through the first cut, and its work ends there.

## Three numbers you will be shown, and what is under them

That first cut is what the figures are sold on, and whoever sells you a redesign will bring figures. It helps to know how they are made, because the three commonest ones are made in three different ways and each only looks like a measurement.

- **A real number that measured something else.** "Forty-six per cent of people judge a company by its design" comes from [a Stanford study of 2002](http://web.archive.org/web/20050924204540/http://www.consumerwebwatch.org:80/dynamic/web-credibility-report-evaluate.cfm), in which 2,684 people compared live websites. The figure is the share of free-text comments that mentioned design. The share of people is a different number. The slip is in the original report itself, a few paragraphs from its own correct definition.
- **A number of unknown origin.** "Seventy-five per cent judge a company's credibility by its website design" is absent from every Stanford document. We searched the full report page by page. It is a separate number of unknown origin, with a history of its own.
- **A number that contradicts the same company's own experiment.** "Half a second of delay costs twenty per cent of traffic" is a conference anecdote from 2006, retold second hand. Three years later Google published [its own controlled experiment](https://research.google/blog/speed-matters/): 100 to 400 milliseconds of added delay reduced searches per user by 0.2% to 0.6%. Both figures are about Google. Only one has a method, and they differ by a factor of about thirty-four, and the test behind the anecdote changed the number of results per page as well as the speed.

The counterweight to those figures comes from the people who run more experiments than anyone. At Microsoft about a third of ideas improve the metric they were built to improve; at Bing it is closer to 15%, and at Booking.com, Google Ads and Netflix about 10%. The same authors write that across [tens of thousands of tests](https://doi.org/10.1145/3534678.3539160) the largest conversion gains they have seen sit far below the three hundred per cent that agency case studies advertise. Treat a promised percentage as a sales technique, including ours.

One thing in this area is measured properly and is worth using. In [an experiment of 2006 with 112 participants](https://doi.org/10.1287/isre.1060.0093), buying from an online store, a promise on its own left trust where it was. The same promise with figures behind it, and with an explanation of where the figures came from, moved it about two and a half times as much. On a site, that means writing how many years, how many projects, and what happens if it goes wrong, in place of "quality and professionalism".

## An old structure holds an old explanation

Many companies at this stage are living on a design that was bought or commissioned once, and buying it was the right call at the time. It was cheap and looked decent enough to start with. The trouble is what it was: a layout drawn to present an invented company in a demonstration. It has every outward sign of having been thought through, a first screen, three columns of benefits, a block for testimonials, a call to act. The decisions behind it were made about somebody else.

So the site has a slot for every answer, and the slots hold generic copy. And when the business grows, the new answers need sections the structure lacks: a specialisation, a scale, proof that appeared after the layout was drawn.

There are also costs that are invisible on the day of purchase. Support on the marketplace is [six months, extendable to twelve at most](https://themeforest.net/page/item_support_policy). The [licence](https://themeforest.net/licenses/terms/regular) covers one site. The page builder that comes with the theme leaves its own markup inside your text when it is switched off, which both [WPBakery](https://wpbakery.com/blog/7-common-wpbakery-issues-and-how-to-fix-them/) and [Divi](https://www.elegantthemes.com/blog/general-news/divi-5-and-the-move-away-from-shortcodes) describe themselves. And a theme's maintenance can stop: [The7](https://themeforest.net/item/the7-website-and-ecommerce-builder-for-wordpress/5556590), sold over three hundred thousand times, moved to a subscription in July 2026, and testing of the version people already owned against new versions of WordPress and PHP has stopped. Separately, the security report for 2025 counts [11,334 vulnerabilities in the WordPress ecosystem](https://patchstack.com/whitepaper/state-of-wordpress-security-in-2026/), and attributes the rise in the severe ones largely to paid components sold on marketplaces. [What the same site costs to run on Webflow](https://figmatica.com/compare/webflow-vs-wordpress) is a separate comparison.

## Three things a designer does before drawing

One caution, because it is easy to take this too far. Sites should differ where businesses differ: in the offer, in the order of the arguments, in the proof, in who is being addressed. Visual form is the place to stay familiar: in a study of 2012 people rated unusual layouts lower than familiar ones, so a navigation everyone already knows is worth keeping.

- **Asks the questions out loud, for the first time.** Who do you sell to now? What do you refuse, and why? What do you have that sets you apart from the other three names on the shortlist? Most of the work of a redesign happens here, before anything is drawn, and this is the part a bought design lets you skip.
- **Builds a structure for this business.** An order of sections that leads your buyer to your decision, instead of the order inherited from a demonstration site. This is also where the boundary goes: what you do deeply and what you have ruled out. In our own first meetings a stated boundary does more for trust than a longer list of services, though this is our own observation and a study on it is still to be found.
- **Chooses an address.** Tone, rhythm, how dense the text is, how much restraint, how much air on the page. All of it says who this is for. A bought design can be beautiful, and it is addressed to the invented company of its demonstration.

## Our own case, and the limits of what we claim

We are describing something we did to ourselves this month. Our old site had five pages. It had been made when the studio was small, and it explained that studio: we design websites, here is some work, here is a contact form. Meanwhile the company had grown a research practice, a migration service, an approach to AI search, and opinions about platforms that clients kept asking us for by email.

The weight and the requests are [measured and published](https://figmatica.com/research/figmatica-webflow-to-nextjs-2026-09-22.csv). The home page went from 20.88 MB to 1.74 MB and from 177 requests to 55. The explanation changed more: on 22 September the site went from five pages to eighty-six, and the text a machine can read before any code runs went from 1,066 words to 1,700 on the home page alone. Of the six checks we use for whether an assistant can read a site, the old one passed one and the new one passes six.

What it earned is still open. The site has been live for days, and the first conversion figure will come after a full quarter of work. Naming a figure now would contradict the section above, where we ask you to treat promised percentages as a sales technique. When it comes in, we will publish the raw data the same way we published this.

## Two companies whose explanation we rebuilt

Applica, a growth agency for mobile apps, asked us for a brand and a website from scratch. BloxFi, which builds a global API for currency conversion and settlement, asked us to redesign a site that had fallen behind the company. In both cases the work started with the questions from the section above, and the drawing came after.

- [Applica](https://figmatica.com/blog/remote-design-team-to-dedicated-design-force): Growth agency for mobile apps. Brand and website from scratch.
- [BloxFi](https://figmatica.com/cases/bloxfi): Global API for currency conversion and settlement. Website redesign.

What clients say about working with us, in their own words on Clutch.

> Every our client says that the main thing that makes them trust us and choose us is our website and its design. So, the result speaks for itself.
> Iryna Hurieva, Co-Founder, Clutch

> The guys over-delivered and addressed our request changes promptly, resulting in a smooth engagement.
> Artem Kuzmych, COO, Clutch

## Ten questions to put to your own site on Monday

Open your home page as a stranger who has your money and three other companies on a list. Each item on its own rests on our judgement; together they are what a buyer looks for and finds missing.

- Does the first screen say what you do and for whom, in a sentence a stranger could repeat?
- Is the buyer named, or is it written for everyone?
- Can a reader see what order of price you work in, or what the price depends on?
- Is it written how the work goes, step by step?
- Is there proof: projects with results, named clients, figures?
- Is there an answer to "why believe you": the team, certifications, years, guarantees?
- Is there a boundary: what you do deeply and what you turn down?
- Does the list of services and projects on the site match what you have sold in the last six months?
- If a tender committee opened it, would they find certifications, a named team and projects with results?
- Is every promise on the page backed by something, or is it an adjective?

If the list shows more gaps than answers, the decisions about who you sell to and why almost certainly still live only in your head, and the site shows that to people who keep their conclusions to themselves.

We do this as a service, and you should read the page knowing that. You can also do the first half yourself, with the ten questions above. The part that is hard to do alone is the second: hearing how your own answers sound to somebody who is hearing them for the first time.

> **About Figmatica.** Figmatica is a design studio that builds [AI-native websites](https://figmatica.com/services/agentic-website). Our clients include WWF Ukraine, KFC Ukraine, the Ukrainian Startup Fund, BRDO, RIEL and TAF Industries. We [redesign websites so that you keep your Google traffic](https://figmatica.com/services/legacy-migration) and [start getting found and recommended by ChatGPT, Claude and Perplexity](https://figmatica.com/services/demand-generation). Yes, this is an ad at the end of an article, but a short one, so [book a call](https://figmatica.com/book) and we will show you on concepts what your company's website could look like.


## Frequently asked questions

### Our site looks fine. Is that enough?

It is enough to get past the first glance. At the next step a buyer looks for what you do, for whom, and with what proof, and there appearance has done its work. A good-looking layout costs between sixty and ninety dollars, so everyone has one.

### Can we just rewrite the text on the pages we have?

Sometimes yes, and it is the cheapest thing to try first. It stops working when the answers need sections the structure lacks: a structure built for a small studio holds only what existed when the layout was drawn, and a specialisation, a scale or new proof need room of their own.

### How long does this take?

The part that decides the result is the interviews and the decisions, and that depends on you rather than on us: it takes as long as it takes the founders to agree on who they sell to now. Drawing and building afterwards is the predictable part.

### What if we redo it and everything stays the same?

That is a real outcome, and a promise of the opposite is a sales technique. At the companies that run the most experiments in the world, between one idea in ten and one in three improves the metric it was built for. A rebuilt explanation changes something narrower: whoever checks you against a list now finds what they are checking for. Whether that turns into contracts depends more on the business than on the pages.


## Related

- [Storytelling websites](https://figmatica.com/services/storytelling-websites): Sites that win traffic and people.
- [Agentic website design](https://figmatica.com/services/agentic-website): Design and development of a site your team runs with AI agents.
- [Legacy migration](https://figmatica.com/services/legacy-migration): Agents move your site fast, traffic held.