Tests cheaper than a build

How to Validate a SaaS Idea Before Building: Please Do Not Hire Us Yet

Your brother-in-law thinks it is a great idea. So does the man at the gym. Neither has ever paid for anything in your trade, and both will be unavailable when the first invoice arrives. There are cheaper opinions to buy, and some of them come with a stranger's credit card attached.

November 19, 2026 · 10 min read

acmeflueslog.com · test results so far
Compliments are free, so count something else
Forty names, three cards
People contacted40
Took a call14(35%)
Described the problem unprompted9(64%)
Tried the manual service5(56%)
Paid, knowing it was refundable3(60%)

Most people never pick up, and that sets what getting a customer will cost later

Illustrative: made-up test results for a fictional chimney sweep

The short answer

Validate a SaaS idea by naming ten real buyers, asking them about the last time the problem happened, checking what they already pay for, running the service by hand for three customers, then asking strangers for a refundable payment on a plain page. Each test costs far less than a build. If nobody pays for the clumsy version, a polished one rarely helps.

On this page
  1. We sell software, and our advice is to wait
  2. Ten names on a page, before anything else
  3. Ten conversations about last time, not about next time
  4. Five tests, cheapest first, with what each one costs
  5. Run the whole thing by hand for three customers
  6. Taking money for something unfinished, without fibbing
  7. Reading the results without lying to yourself

We sell software, and our advice is to wait

We build web applications for a living, and a founder who shows up with a deposit and no evidence is, in the narrowest sense, good for us. We would still rather you did not. A failed product is a bad reference, and it costs you more than the invoice. So here is a set of tests you can run this month with a phone, a page and very little money, starting with some that cost nothing.

The idea behind all of them is simple. Opinions are free, so they carry little information. Every step below asks for something more expensive from the other person: their time, then their honesty about their current habits, and finally their money. People who love you will give you compliments. Strangers with the actual problem will give you a Tuesday afternoon, then a card number. Pay attention only to the second kind of signal.

If you are starting from a file you already use, from spreadsheet to SaaS adapts these tests to it. If you just need the tool for your own business, none of this applies, and the question of custom software is the one to ask instead. The rest of this page assumes you hope other businesses will pay you every month, which is what makes a small subscription product harder than it looks.

Ten names on a page, before anything else

Say a solo chimney sweep in Asheville, invented for this example, thinks other sweeps would pay for a tool that files inspection photos by flue and produces a report for the homeowner. Before doing anything else, he writes down ten actual people who run businesses like his. Names, phone numbers, how he knows them. He gets to six, and has to think hard about where to find four more.

Those missing four are the first finding. If you cannot name ten people who have this problem, you have either picked a market you do not know or one you cannot reach. Both are serious, since a small product has no marketing department to fix that for it. Where do they gather? Which association, trade show, forum or supplier do they all use? That place is your entire sales plan, so it had better exist.

Finding them also tells you what getting a customer will cost later. When the only route to your buyers is paid search, read Google Ads for very small budgets and low-competition keywords before you assume they will arrive by themselves. Search takes patience. Google's starter guide, updated December 10, 2025, says some changes might take effect in hours and others could take several months.

Ten conversations about last time, not about next time

Now call them. Do not pitch. As soon as you describe your idea, a polite person starts helping you feel good about it, and your research is over. Ask about their past behavior instead, since it is the only thing they cannot be nice about. Twenty minutes per call is plenty. Say you are trying to understand how people handle this, which has the virtue of being true.

Questions that find things out

  • When did this last happen? Walk me through it.
  • What did you do about it?
  • What did that cost you, in hours or money?
  • What do you use for it now, and what do you pay?
  • What have you tried and stopped using? Why?
  • Who else should I talk to?

Questions that produce flattery

  • Would you use an app that did this?
  • Do you think this is a good idea?
  • How much would you pay for it?
  • Would you want this feature?
  • If I built it, would you sign up?
  • Isn't it annoying when this happens?

Listen for money already moving. Someone paying for a clumsy product, a part-time helper or a stack of duct-taped apps has already shown you the problem is worth something. Someone who says "yeah, that's annoying" and then cannot remember the last time it happened has shown you the opposite. Write down their words, not your impression. You will want the exact words later, for the page in the next step.

Five tests, cheapest first, with what each one costs

Run these in order, and do not skip ahead because you are excited. The prices are list prices as of November 2026. The pass lines are our own rules of thumb, chosen to be hard to fool yourself with. They do not come from any study, and your market may justify different ones. What matters is writing yours down before you see the results, because a line set afterward tends to land right below whatever you got.

Validation tests in order of cost, vendor prices as of November 2026
What it costsA pass, by our rule of thumb
Competitor searchAn hourYou find products with paying customers and complaints
Ten conversationsEvenings. HubSpot's free CRM can track them, up to 1,000 contactsSix describe the problem unprompted, and three already pay to deal with it
Do it by handYour hours. Zapier's Free plan includes 100 tasks a monthThree customers come back for a second round
One page with a priceSquarespace Basic at $25 a month, or $19 billed annuallyStrangers who found it on their own ask how to start
Refundable prepayment2.9% plus 30 cents per card charge through StripeThree people with no reason to be kind to you pay
Dark green chalkboard with chalk arrows narrowing downward, a black-and-white halftone cutout of a stack of index cards and a telephone in front, no words
Each test narrows the list. The ones left at the bottom are the ones who paid.

A word on competitors, since founders tend to hope for none. A market with no rivals usually means nobody pays for this. Complaints in reviews of an existing product are the best news you can get: paying customers who are unhappy in specific ways. You do not need to beat that company. You need to be better for a slice of its customers it does not care about, which is where small companies can compete at all.

Run the whole thing by hand for three customers

This is the test that saves the most money. Do what the software would do, yourself, behind the scenes. Our invented sweep could have three colleagues text him photos after each job. He sorts them in a folder, builds the report in a document and sends it back that evening. No code at all. It is slow and a bit absurd, which is fine for three customers and would be fatal for three hundred.

  • You learn the real sequence of steps, and it never matches what you imagined.
  • You find out which inputs people get wrong or forget.
  • You see which part of the output they actually use. It is often a surprise, and often smaller than planned.
  • You discover what you can charge, because you can raise the price on the fourth customer and see what happens.
  • You get your first references, from people who have used the service rather than heard about it.
  • You have the rule book that later becomes a scope a builder can price.

Charge for this. A free trial tells you people like free things. Even a small fee sorts the curious from the serious, and it starts a real conversation about value. A template or spreadsheet can serve as a halfway step if your service runs on one. If you get tired of doing it by hand and customers would be upset if you stopped, you have just found the strongest evidence for building there is.

Taking money for something unfinished, without fibbing

The strongest test is asking for money before the software exists. It is also where a founder can do something shabby, so do it honestly. Say in plain words on the page that the product is not built, what you intend to deliver, and that the payment is fully refundable on request. People who know that and pay anyway have told you more than any survey could.

Stripe's documentation, read November 6, 2026, says Payment Links can sell a product or a subscription from a Stripe-hosted page without code, share by email or website, and offer no-code refunds. Keep the page plain. Say what it does, who it is for, the price, and that it is coming. Use your interviewees' exact words. Our page on landing pages with one job helps if you have not written one before.

Reading the results without lying to yourself

You will want the data to say yes. Everyone does. So reread the lines you wrote down before the tests, and be stricter with yourself than you think fair. Most ideas fail these tests, and failure at this price is a bargain. A surprising number of those failures also contain a better idea, usually a narrower one that someone mentioned in passing during call seven.

Keep going if

  • Strangers paid, and could have asked for a refund but did not
  • Customers complain when your manual service runs late
  • People forward you to others in their trade
  • Someone asks when the next part will be ready
  • You still want to do this after thirty conversations about it

Stop or change course if

  • All your yeses came from friends
  • Everyone loved it and nobody paid
  • People use a rival and shrug when you ask why
  • You cannot find more buyers after the first ten
  • It would only work at a price that cannot cover your hours
Failing these tests costs a few weeks of evenings. Failing after a build costs a good deal more, plus the evenings.

If the answer is yes, bring us what you learned: the customer's words, the manual process, what people paid and what they kept asking for. That is most of a scope. Web apps at DataCram start from $575, depending on complexity, and a real SaaS costs more, so we quote it after a conversation. What it costs to build and run has the arithmetic, including your support hours. Our custom software page has the wider view, our prices are public, and the hub for very small businesses has the rest. If you want to think out loud, (210) 346-0848 reaches a person who will happily talk you out of it.

Questions owner-operators ask

No magic number exists, and we would not trust anyone who gave you one. We use ten as a working minimum for a narrow market, because by then you usually start hearing the same things twice. Keep going until you can predict what the next person will say. If every conversation still surprises you, you do not yet understand the problem.

It is weak proof. An email address costs a person almost nothing, and many sign up out of curiosity or politeness. Use a list to find people to talk to, then ask those people for a refundable payment. Even a small one sorts them. Compare how many paid with how many signed up, and trust that smaller number.

In a small niche, this fear costs far more than the theft does. The risk that nobody wants it is much larger than the risk that a busy stranger drops everything to copy you. Execution and your knowledge of the trade are hard to steal. Keep truly confidential details to yourself if you have them, and ask an attorney if something is genuinely proprietary.

Yes, as a test. A rough demo can show people what you mean and draw out sharper reactions than a description can. Just do not treat a working demo as evidence of demand, and do not put real customer data or real payments through it until someone who can read code has reviewed it. A demo answers how it might look, not who will pay.

Charge something close to what the product would need to cost to be worth running, and not a token. Nearly anyone will pay $1. The cost arithmetic matters too: by our math at Stripe's list prices as of November 2026, fees take about 4.6% of a $29 monthly charge and about 6.9% of a $9 one. Test a price you could live on.

Keep reading

Free · No obligation

Tell us what your tests found

Send what you learned, including the disappointing parts. A person replies with whether it is ready for a scope, or which cheap test we would run next. There is no charge for that reply.

Which of these sounds like you? (pick any)

Rather talk now? Call (210) 346-0848.

CallGet a quote