Build or keep renting

When a Small Business Needs Custom Software, and the Six Tests to Run First

Most of the time it does not. A very small business should rent software until a rented product fails a specific, repeated job, and an ordinary month of use will usually prove whether it has. Here are the tests, cheapest first.

November 11, 2026 · 10 min read

acmeautoglass.com · build or rent review
Six tests before anyone writes code
A quoting tool with one real problem left
  • PassJob fits in one sentence
  • PassTask timed for two weeks3.5 hours a week
  • PassProcess stable for six months
  • PassRented product tried and failed30-day audition
  • WarnSum clears price and upkeepclose
  • FailNamed person to fix itnobody yet
Illustrative: a sample review for a fictional two-van glass shop

The short answer

A small business needs custom software when it has tried the rented products, one repeated job still fails, the process has held still for months, and the hours or lost sales outweigh the build and its upkeep. If a rented product does the job well enough, keep renting. DataCram's web apps start at $575, depending on complexity, and we will say when not to buy.

On this page
  1. The answer nobody earns a commission on: buy nothing
  2. Give a rented product a 30-day audition on real work
  3. Six tests a build has to pass before anyone writes code
  4. Jobs a small company often cannot rent
  5. Bad reasons that sound expensive enough to be good
  6. Three outcomes, three very different bills
  7. Ask us, and expect us to talk you down sometimes

The answer nobody earns a commission on: buy nothing

Software salespeople sell seats. Developers sell builds. Nobody collects a commission on the sentence "keep the free plan." So you will not hear it often, and you should hear it first, including from a company that builds software for a living. That is us. We would like the work. We would like it more from an owner who has already ruled out the cheaper answers, because that owner knows exactly what the thing has to do.

There is a practical reason to look on the shelf before you look at a developer. Companies your size are not an odd corner of the market. The U.S. Census Bureau's 2022 business tables count 6,395,635 employer firms, and by our own arithmetic about 79.2% of them had fewer than 10 employees. The Census Bureau's May 2026 small business article adds 30,427,808 nonemployer establishments in 2023. That is a very large crowd of customers, and vendors have noticed. Someone has probably already built a plan for the job you are thinking about.

So the question at the top of this page has two halves. Is the job truly yours alone? And have you actually tried the alternatives, or only looked at their pricing pages at 10 p.m. and felt tired? The custom software overview lists the rented tools we would try first, with prices. This page is about how to test them, and how to tell when they have failed you for real.

Give a rented product a 30-day audition on real work

A demo proves that a product works for the person selling it. An audition proves whether it works for you, on a Tuesday, with a customer waiting. Thirty days is an arbitrary number that happens to include a billing cycle and at least one bad week. Run it on one product and one job, and do not move the rest of the company while you are at it.

  1. 1

    Pick the single job

    Not "the office." Something like "send a quote the same day I see the house." If you cannot pick one, you are not ready to choose software of any kind.

  2. 2

    Use real customers, not sample data

    Enter the next ten actual jobs. Sample data has no misspelled street names, no second addresses and no customer who changes their mind in the driveway.

  3. 3

    Keep a tally of every workaround

    Each time you leave the product to finish the job somewhere else, make a mark and write down why. Marks on the same reason are your evidence. Scattered marks usually mean you are still learning the tool.

  4. 4

    Ask the vendor in writing

    Describe the gap in two sentences and ask whether the product can do it. Support may know a setting you missed. That reply costs nothing, and it may save you a build.

  5. 5

    Test the way out

    Before the 30 days end, export your records and open the file. You need to know whether your data leaves in a usable form, whichever way you decide.

At the end you hold one of three results. No marks: you are done, and congratulations on your new subscription. Scattered marks: you have a product and a learning curve. Thirty marks with the same reason: you have a specification, written in your own hand, for the only piece worth discussing with a builder. Often that piece is not a whole system. It is one connection between two tools you already pay for.

Six tests a build has to pass before anyone writes code

Now the part that decides it. A project should pass all six of these. Passing four means you have an interesting idea and should write it on the wall. A failed test does not kill the idea forever, but it usually tells you what to do next, which is rarely hiring someone.

The tests, with what a pass and a fail sound like in practice
A pass sounds likeA fail sounds like
Can you say it in a sentence?"Price a windshield job from the vehicle and the distance.""Run the whole business better."
Have you timed it?Two weeks of tallies, as in timing a task before you automate it"It feels like it takes forever."
Has the process held still?The same steps for six months or moreIt changed twice this quarter and may change again
Did a rented product fail?A named product, a named gap, a vendor's written replyYou have never tried one
Does the sum clear?Hours times their value beats the price plus upkeep, by a margin you would bet onIt only works if every hour saved turns into a paid hour
Who fixes it on a Tuesday?A named person, in accounts you holdThe nephew who moved to Denver
Dark green chalkboard with chalk arrows splitting into two paths, a black-and-white halftone cutout toolbox in front, no words
Most projects should stop at the first fork, and that is a good outcome for the person paying.

The test people skip is the third. Software pours your current process into concrete. If you are still working out how you quote, how you schedule or who does what, the tool will hold the old version in place just as you learn the better one. A process that changes every few weeks belongs on paper, or in a spreadsheet that still does its job, where changing your mind costs nothing.

Jobs a small company often cannot rent

All that caution has an exception, and it deserves plain words. Rented products are built for the average customer in a category. Where your work departs from the average, in a way customers notice and pay for, the rental stops fitting. These are the places the argument tends to hold up. Each still has to pass the six tests.

The pricing rule that is yours

Say a two-van mobile windshield shop in Tyler, invented for this example, prices by vehicle, glass type and distance. No quoting tool knows its rule. A small estimate calculator might.

The gap between two good products

Each tool is fine. You are the cable between them, retyping at night. That is a Stripe, QuickBooks and CRM problem, and it does not call for a new system.

The thing customers must see

Job photos, status, documents, in your trade's terms. If your rented software has no such page, a customer portal may earn its place.

The page of numbers you assemble by hand

Three logins and a calculator every Monday. Five figures on one dashboard can end that chore, and reading data is the cheap kind of software.

The back office in your trade's shape

When no field product fits how you track work, a small internal app can replace the shared sheet. Check the field service page first.

Notice what is missing from those cards. Invoicing. Payroll. Email. Calendars. Accounting. Those are solved problems with large companies maintaining them and tax rules changing underneath them. A one-person business that commissions its own invoicing system has bought a second job. Rent those for as long as you are in business, and spend your custom dollars, if any, at the edges where your work is different.

Bad reasons that sound expensive enough to be good

These come up in nearly every conversation about building something. None of them is foolish. All of them lead to software that nobody needed, and to an owner who is now also the product manager of that software.

  • "I want to own it." Owning means owning the outages too. Ownership is a real benefit when the tool holds something you could not stand to lose. On a contact list, you can have it for the price of an export.
  • "The monthly fee adds up." It does. Intuit lists QuickBooks Essentials at $85 a month as of November 2026, which is $1,020 a year by our arithmetic. Now price rebuilding accounting software and keeping it current. The fee is the bargain.
  • "My competitor has an app." Ask whether anybody opens it. Then read whether you need an app at all.
  • "It would be AI-powered." A sticker on a contact form changes nothing about the form. If a machine could really take over part of the job, the AI automation page says which parts, and which to keep human.
  • "We could sell it to others someday." Maybe. That is a second business, with customers, support and billing. Read spreadsheet to SaaS before you let the idea justify the cost of the first version.
  • "Someone quoted me a great price." A low quote for the wrong thing is still the wrong thing, and it is not the owner's fault that it looked like a bargain.
A one-person business that commissions its own invoicing system has bought a second job.

Three outcomes, three very different bills

If the tests send you to a build, the size of the build matters more than the decision itself. Most very small companies should land in the first two rows. The prices are DataCram's, as of November 2026, and the pricing page has the rest. Ask anyone else for the same three numbers.

Rent, connect or build, with DataCram's published starting prices, November 2026
What you payWhat you are now responsible for
Keep rentingThe vendor's plan, often nothing extraLearning it, and exporting your data now and then
Connect what you rentAPI integration from $575, depending on complexity; live the same dayKnowing who to call if a vendor changes something
Build one small toolWeb app or portal from $575, depending on complexity, plus running costsHosting, backups, logins, and a named person to fix it
Build a phone appiOS or Android from $3,750, plus store feesAll of the above, and store rules that change
Hire a developerA median $135,980 a year per the BLS, May 2025, before benefitsBeing a software company, in addition to your company

The last row is there to make a point. Almost no business with fewer than 10 people should have one on staff. Buying a small tool as a project is how companies this size get software, and the real number comes from what moves a web app's cost. If the build is small and the budget is tight, web application financing runs at 0% with 25% down, up to 12 months. Do not stretch for a tool that fails the sum, though. Cheap money does not fix a thin case.

Ask us, and expect us to talk you down sometimes

Here is what you would be buying from. DataCram is a web studio in San Antonio, Texas, building for the web since 1999. We built and run our own software, including Athena CMS and Kairos, and we would not claim a shelf of case studies from small companies, because we do not have one. See the work we can show and judge it. The wider argument is in the section for very small businesses.

Questions owner-operators ask

Rarely for a very small business. A subscription bundles upkeep, security and new features, while your own tool has running costs and someone to pay when it breaks. A build can win when you pay for a large product and use one small piece of it, or when per-user fees climb. Run the sum on your own invoices, over several years, before you trust anyone's claim.

A connection, in most cases. If two products you already like refuse to talk to each other, an API integration ends the retyping without replacing anything. It starts at $575, depending on complexity, and goes live the same day. It is also included once in every DataCram website package, so check that before buying it separately.

Write the steps down, then ask whoever else does the job to read them. If they disagree, or you changed the steps in the last few months, it is too early. Software locks in whatever it is given. Run the process on paper or in a spreadsheet until it stops changing, then decide. That waiting costs little and improves any later estimate.

It happens, but usually in one narrow place, not everywhere. The usual signs are the same workaround every day, a report you rebuild by hand, or customers who need to see something the product will not show. Fix that one spot with a connection or a small tool. Replacing the whole product is a much larger decision and rarely the right first move.

Yes, and we would rather do it early. Describe the job in a sentence and tell us what you tried. If a rented tool or a spreadsheet would do it, you will hear that from us, with no bill for the conversation. If the case is thin, we will say so. Neither answer depends on you having a budget ready.

Ask where it will run, whose accounts hold it, who gets the code, what it costs each month, how you get your data out and who fixes it next spring. Ask what they would try before building, too. A builder who has never suggested a cheaper answer is telling you something. Our page on choosing a technology partner has a longer list.

Keep reading

Free · No obligation

Find out whether this one is worth building

Tell us the job in a sentence and what you already tried. A person at DataCram replies with the rented tool we would try, the small build we would quote, or a plain no.

Which of these sounds like you? (pick any)

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

CallGet a quote