Nobody publishes these numbers, which is why the question is so hard to answer from the outside. Agencies quote in the tens of thousands, freelance marketplaces quote in the hundreds, and neither figure tells you what a working replacement for the tool you're paying for actually costs. Here are real bands, what moves a quote within them, and three examples with the arithmetic shown.
Short version: a single well-defined workflow runs $5,000–$12,000. A system a department depends on runs $12,000–$35,000. Below $5,000 you're buying a script, not an application — and that's sometimes the right purchase.
The four bands
| Band | Cost | Timeline | What you get |
|---|---|---|---|
| Automation | $1,500–$5,000 | 1–2 weeks | A script or integration that removes manual work between tools you're keeping. No interface of its own. |
| Single workflow | $5,000–$12,000 | 3–6 weeks | One real application: logins, a handful of screens, your data model, reports. Replaces one subscription outright. |
| Departmental system | $12,000–$35,000 | 6–16 weeks | Multiple connected workflows, roles and permissions, integrations with accounting or payments, real reporting. |
| Company platform | $35,000+ | 4+ months | The system the business runs on. Usually built in stages, with the first stage delivered in the band above. |
The top band is open-ended — $60,000 is shown to keep the chart readable, not as a ceiling. Company platforms are normally built in stages, with the first stage priced in the band below.
Most small businesses land in the second band. That surprises people who expected either $800 or $80,000 — the first is a freelancer who will disappear, the second is an agency pricing in account managers.
What moves the number
Things that add real cost
- Users who aren't you. Software used only by staff can be blunt. Software your customers log into needs to be obvious, forgiving and good-looking, and that roughly doubles the front-end work.
- Migrating history. Moving five years of messy records out of an old system is frequently the single largest line item. Budget for it explicitly rather than discovering it in week six.
- Integrations with anything old. A modern REST API is a day. A vendor's CSV drop folder, a legacy SOAP endpoint or a system with no API at all is a week or three.
- Money and compliance. Anything touching payments, PHI, or regulated records carries requirements that are not negotiable and not fast.
- Offline and mobile. "It has to work in a basement with no signal" is a genuinely hard requirement, not a checkbox.
Things that cost less than people expect
- More screens of the same kind. The tenth list-and-form screen is far cheaper than the first, because the machinery already exists.
- User count. Ten users and two hundred users cost almost the same to build. This is exactly the assumption per-seat pricing charges you for and custom software doesn't.
- Reports, once the data model is right. Most of the cost of a report is having clean data underneath it.
Three worked examples
1. A trades business replacing per-technician scheduling
Fourteen field staff on a dispatch tool at $65 per technician per month — $910/month, $10,920/year, and every hire adds $780/year. The replacement covers job scheduling, crew assignment, a mobile job sheet and an export into their accounting system.
| Build (single workflow, upper end — mobile use) | $11,500 |
| Data migration from the old tool | $1,800 |
| Hosting and support, year one | $1,200 |
| Year-one total | $14,500 |
| Previous year-one cost | $10,920 |
| Break-even | Month 19 |
Year two costs $1,200 instead of $10,920, and the gap widens with every hire. Over five years they spend about $19,300 instead of $54,600.
2. A 9-person agency replacing project management
$24 per user per month across 9 seats is $216/month — $2,592/year. This one is a closer call, and it isn't really about the subscription: the deciding factor is that three people spend an afternoon a week moving data between the tool and a billing spreadsheet.
| Build (single workflow, plus time tracking and invoicing hooks) | $9,000 |
| Hosting and support, year one | $900 |
| Year-one total | $9,900 |
| Subscription saved per year | $2,592 |
| Labour saved (3 people × 4 hrs/mo × $45) | $6,480/yr |
| Break-even | Month 14 |
On subscription savings alone this project would take nearly four years to pay back and I'd advise against it. The labour is what makes it work — which is why an honest audit counts hours, not just invoices.
3. A retailer who should not have rebuilt
A 6-person shop paying $140/month for inventory and point-of-sale, all in. The tool did the job; the complaint was two missing report columns.
| Replacement build (estimated) | $16,000 |
| Annual subscription | $1,680 |
| Break-even on subscription alone | Year 11 |
| What was actually built | $1,900 |
The right answer was a small reporting tool reading the existing system's export — two days of work against the $16,000 rebuild nobody needed. Roughly a third of the audits I do end somewhere like this.
The ongoing cost nobody quotes
A build is not the end of spending, and any quote that implies otherwise is hiding something. Budget 10–20% of the build cost per year for hosting, patching, dependency updates and small changes. On a $10,000 build that's $1,000–$2,000/year — usually still far below the subscription it replaced, but it is not zero. The ownership guide breaks that down properly.
How to sanity-check a quote you're given
- Ask what happens to the source code. If the answer isn't "it's yours," the price is not the whole price.
- Ask for the estimate broken into workflows, not phases. "Discovery, design, development, QA" tells you nothing about what you're buying.
- Ask what's excluded. Migration and integrations are the two things most often quietly left out.
- Be suspicious of very low quotes. A $2,000 quote for a $10,000 job usually ends in an abandoned project and a second bill.
Once you have a number, put it against your actual subscription in the break-even calculator — that's the figure that decides it, not the build cost on its own.