
Jurij Tokarski
Should I Build Custom Software or Buy?
A handful of spreadsheets and tools that don't talk, and an offer to build one system. Buy by default; build only what no tool fits or what clients pay for.
We run our clients, bookings and invoices out of several spreadsheets and two tools that don't talk to each other. A developer offered to build us one custom system. Is that worth it, or should we keep buying tools?
Buy by default. Build custom software only for the part no tool fits, or the part your clients pay you for. Most sheets should stay sheets until they break.
One custom system for clients, bookings and invoices would rebuild a lot that tools already do well. I'd split it instead. Keep what's standard, connect what doesn't talk, and build the one job left over.
Buy the Glue, Build the Edge
Sign-in, payments and storage come off the shelf. Only what makes you different is worth building.
Founders order developers to build custom sign-in, spending 20 to 100 hours on it. Firebase Auth does the same job at no cost up to 50,000 monthly users. Payments, file storage and notifications go the same way.
Domain-driven design gives the split. The core domain is what makes the business unique: build it, and spend most of the effort there. The generic domain is sign-in, payments, storage, email and notifications: buy it, always.
Find the Core is the principle behind it. One capability creates the advantage, and everything else is bought or deferred.
Owning generic software means maintaining it forever. Every security update and every bug becomes your problem. Buying hands that problem to someone else.
For a small business, that puts invoices on the buy side. Accounting, payroll and email stay bought, because their rules change and the vendor keeps up. A cheap tool one person uses stays too, because a rebuild would never pay back.
The build side has two signs. A subscription charges per seat for a simple job, like a form, a list or a client tracker. Or a plan caps your own data.
Your client list is the first place to test the split.
A Small CRM Is a List
A small business can build its own CRM, after trying the free tools.
What a small business means by a CRM is a client list, notes and follow-ups. A sales team's full CRM is a bigger build.
The CRM worth replacing is the one billed per seat while the team uses two of its features.
Building for your size now doesn't lock you in. An ordering tool I built ran for over three years. Then the business moved to a full CRM, copied the data it needed, and I switched mine off.
Your client sheet may not need even a small CRM yet. That depends on whether the sheet has broken.
Sheets Become Apps When They Break
A sheet breaks when two people edit it at once, or when a wrong cell costs money.
Plenty of client lists never leave a spreadsheet, and most shouldn't. The free step comes first: a shared Google Sheet with a form in front, or Airtable's free plan.
A free tool stops paying off when the rules outgrow its template. On one order desk, none of the no-code platforms I tested held the pricing rules. Basic changes sat behind premium plans, and building was faster than configuring.
Other signs come from the work itself:
- Only one person can do it right, so the work waits for them.
- The same data gets typed twice.
- A wrong price or product turns into a refund.
A warehouse I built a tool for in 2020 had two signs at once. Only its most experienced staff could match the Excel files to the boxes, and one wrong product meant a return.
The Team Shapes It or Ignores It
The people who do the job shape the tool as it lands, or they won't use it.
So they try each screen on real work and comment before the next task starts. The tool bends to how they already work, paper included. Warehouse workers who preferred printed lists got printed reports next to the phone app.
Training is then short. On a parcel desk, it took 30 minutes, and packages stopped waiting for managers.
On your team, the person who keeps the bookings sheet today is the one to test the first screen.
Hours Times a Rate, Plus Hosting
The build costs hours times a rate. Running it costs a few dollars a month.
Web developers average $45 to $75 an hour, in Index.dev's May 2026 data. What a freelance web developer charges has the other ranges. Set that against the subscription's yearly bill, not its monthly one.
Running it is the small line. A tool for one team sits near the free tiers, on your own accounts. The parcel tool runs on about $3 a month.
Upkeep is the bigger line, as what an app costs after launch shows. So the comparison for the system you were offered is the build plus upkeep against a year of the tools it replaces.
How I Replace a Tool
I start with a map of the subscriptions, tool by tool. Buy what's standard and changes often. Build what's simple, yours and priced per seat.
I'd rather say "keep" in week one than sell a rebuild. The map with its top replacement fits in one week, and so does a single form or a first dashboard. A small CRM's first screen fits in a week too, and a full one grows a week at a time.
When a sheet does have to go, I start with the job, not the sheet. Someone walks me through the work once, and I write down the rules they apply without thinking.
Then only the columns that matter move into a database. In the warehouse files, 3 to 4 of 10 to 15 columns told the products apart.
Next comes one screen for the worst part, built around how the team works. On the parcel desk, that was a keyboard-first form where a few letters of a client's name filled in the rest. Checks, sign-ins and a dashboard follow one task at a time.
The map is where custom software for small businesses starts. A sheet's replacement is an internal tool. Each is one week at the weekly price on the homepage, and each next replacement is another week.
What the Owner Should Do Now
Connect the tools you have first, and build only the one job left over.
Start with the two tools that don't talk. Try connecting them before replacing them. n8n, Make or Zapier covers which fits.
Keep the invoices in the accounting tool. Keep every sheet that one person runs without errors.
Then look at what's left. If one job still lives in someone's head, or a wrong cell in it costs money, that's the one to build. The rest of the custom system you were offered can wait until something else breaks.
Got thoughts on this? Reply by email.

Jurij Tokarski
I run a solo business and produce software. Usually I'm shipping for clients or building for myself. Sometimes I pause to write about it.