Why 90% of professional services SaaS ideas fail - and why that number is wrong

You have seen the statistic. Ninety per cent of software ideas fail, quoted with the confidence of something everybody knows, usually just before somebody explains what you must do differently. It is a striking number, and it has a way of settling the question before anyone has looked at your situation. So it is worth asking where it came from, because the answer is more interesting than the number.

It matters because of how the figure makes people behave. Believe that nine in ten ideas die and you conclude, sensibly enough, that you must be exceptional to survive: build an audience, run validation experiments, launch a landing page and gather email addresses from strangers to prove there is demand. That is a reasonable programme for somebody starting with nothing. It is a strange way to spend a year if you have spent the last decade being paid by the exact people you intend to sell to.

In this blog, Elliott Prince, Managing Director at Ferrous Labs, takes the failure statistic apart, explains what the real numbers say, and sets out why an established consultancy is holding the proof of demand that everybody else has to go out and manufacture.

Where the ninety per cent actually comes from

Trace the figure back and it leads to a Startup Genome study of venture-backed startups, where the thing being counted was not companies going under but companies failing to return ten times an investor's money. Those are different events. A business that supports its founders comfortably for a decade is a failure by that definition and a success by any other, and quoting it as a death rate flatters the drama while losing the meaning.

The official numbers are less alarming and rather more useful. The Office for National Statistics puts the five-year survival rate for UK businesses born in 2019 at 38.4%, published in its Business Demography bulletin in November 2025, alongside 317,000 business births and 280,000 deaths in 2024. So roughly three in five UK businesses do not reach their fifth birthday, which is sobering enough without inflating it to nine in ten.

Even that is not the number the original claim pretends to be, because it counts registered businesses rather than ideas. Nobody measures ideas. There is no register of the product somebody sketched on a train and never mentioned again, which means any statistic about the failure rate of ideas is an assertion wearing a decimal point. The useful question was never what happens to ideas in general. It is what happens to yours, given what you actually bring to it.

Those numbers describe somebody who is not you

Survival statistics average across everything: the restaurant, the one-person trade, the venture-backed company burning capital in pursuit of a market that may not exist. What almost all of them share is that they began without customers. The founder had a conviction and had to go and find out whether anybody agreed, which is expensive, slow and the single largest reason early businesses run out of money before they run out of ideas.

An established consultancy is in a materially different position, and the difference is not confidence, it is evidence. You have been running a paid experiment for years. Every engagement was somebody deciding your judgement was worth money. Every repeat client was them deciding it again with full information. Every project you turned down for lack of capacity was demand you measured and declined to serve. None of this is in a spreadsheet labelled market research, which is precisely why it gets overlooked, but it is the highest-quality demand data that exists: people paying, repeatedly, with their own money, for the thing you are proposing to productise.

Set that against a landing page with an email capture on it. The landing page tells you that some proportion of strangers, arriving with no context and no obligation, were willing to type an address into a box. That is a weak signal collected at real cost. You already hold a strong one collected over years, and it is sitting in your invoices.

What your track record already proves

It is worth being specific about what that history can and cannot tell you, because the argument is not that a consultancy needs no validation at all. It is that most of the expensive validation has already happened, and what remains is a narrower question.

Your record establishes that the problem is real, that it recurs, that people pay to have it solved, and roughly what they pay. It tells you which parts of the work clients query and which they accept without comment, which is a good guide to where the perceived value sits. It tells you what you have said no to, and a pattern of declined work is a description of unmet demand in your own market. Most usefully, it tells you where your own delivery is repetitive, because the tasks your team finds tedious are generally the ones a product can take.

What it does not establish is whether anybody will buy the solution without you attached to it. That is the genuine open question, and it is the one worth spending money to answer. A methodology can be inseparable from the person delivering it, and plenty are. So the validation that remains is not "does this problem exist" but "will somebody who has never met me pay for this in software form", which is a far tighter question and one you can test against the people who already know you before you test it on anybody who does not.

Iconic Digital already knew what their clients were struggling with

We are building Growth Gorilla with Steve Pailthorpe and the team at Iconic Digital, turning a services-only marketing agency into one with its own product line. The case study describes the starting point in a phrase that has stuck with me: the opportunity sat with their own clients.

They had watched small business owners try generic AI tools, struggle through onboarding, get content back that did not sound like them, and drift back to doing it by hand. No survey produced that insight and no landing page tested it. It came from delivering marketing services to those owners and paying attention, and it is considerably better evidence than a conversion rate, because it explains the mechanism rather than just registering an outcome.

That understanding shaped the product rather than merely justifying it. Knowing that authenticity was the sticking point is why the build centres on a voice engine that captures how an owner actually speaks. Knowing that onboarding friction kills conversion is why there is a deliberate path for people unwilling to sit through it on day one. An agency starting from nothing would have discovered both of those the expensive way, some months after launch.

Start with what you already know

The failure statistics are not irrelevant to you, and I would not pretend that a client list makes a product inevitable. Building software is still hard, it still takes longer than you expect, and a proven service can absolutely become a product nobody wants. The honest version is that your odds are better than the average because the average includes a great many people answering a question you settled years ago.

So before you build anything, spend a week with your own records rather than a landing page. Go through the last two years of work and mark which engagements were substantially the same job wearing different names. Look at what you have quoted for and lost, and why. Ask your team which parts of delivery they would automate first if anybody let them. That exercise costs a few days, uses evidence nobody else can access, and will tell you more about what to build than a quarter of audience-building ever will.

Then go and answer the one question your history genuinely cannot: whether the value travels without you in the room. If it does, you are not starting a business at ninety per cent odds of failure. You are extending one that already works, which is a different proposition entirely, and rather a good one.