Costs and decisionsUpdated 12 min read
How much does custom software cost? Ranges and the math

In this article
- The short answer: price ranges with a real source
- Why two quotes for the same system look nothing alike
- Salary, rate and project price are three different things
- What the price is made of: hours per component
- The five things that move the number most
- The cost nobody puts in the proposal: maintenance
- Why small projects turn out better
- Set the appetite before the design
- Three-year cost: three ways to pay for the same system
- 12 questions for comparing quotes
- When custom software isn't the right call
- How Efidis charges
The short answer: price ranges with a real source
Clutch, the software vendor directory, publishes a pricing guide built from verified reviews of projects on its platform. Two numbers from that guide sum up the problem: most projects cost between $10,000 and $49,999, but the average is $132,480, running about 13 months at roughly $10,209 a month (Clutch, 2026).
The gap between those two numbers is the first lesson. A few very large projects drag the average up. If your company has 10 to 50 users and needs to get sales, inventory and collections under control, your benchmark is the common range, not the average.
| Source | What it measures | Number |
|---|---|---|
| Clutch (2026) | Most common budget range for reviewed projects | $10,000–$49,999 |
| Clutch (2026) | Average cost per project | $132,480, about 13 months |
| Clutch (2026) | Hourly agency rates in the U.S. and Mexico | $50–$99 and $25–$49 |
| Accelerance (2026) | Senior developer rate from Latin American vendors | $60–$75 an hour |
These figures come from agencies that sell to companies of every size and country, and Clutch doesn't publish its sample size. Use them to get your bearings, not to set a budget.
Why two quotes for the same system look nothing alike
Here's a made-up example with round numbers. You ask two vendors for "an order and inventory system." One quotes $8,000 and the other $60,000. Odds are neither one is ripping you off. They priced two different systems that happen to share a name.
The $8,000 vendor heard a product catalog, orders, a single warehouse and an initial product upload from a spreadsheet. The $60,000 vendor heard three warehouses with transfers, a sales rep app that works without a signal, an integration with your invoicing system, five years of migrated history, and different permissions for sales, the warehouse and management.
Before you compare prices, compare scope. If the two proposals don't spell out what's included piece by piece, you have nothing to compare.
Salary, rate and project price are three different things
A lot of the confusion comes from mixing up three numbers that measure different things.
- SalaryWhat the developer earns. In the U.S., the median was $133,080 a year in May 2024 (BLS, 2024). That's what you'd pay for one person on payroll, before benefits, equipment and software, and with no backup when they leave.
- RateWhat an agency charges per hour. On top of salary, it covers unbilled time, project management, testing, tools and margin. Latin American vendors charge $60 to $75 an hour for a senior developer, and those rates dropped 7.1% in 2025 (Accelerance, 2026).
- Project priceEstimated hours times the rate, sometimes with a buffer for the risk that scope grows. For example, 500 hours at $60 comes to $30,000.
That's why the rate alone tells you little. Accelerance, which tracks rates every year, warns that the hourly rate is a poor measure of real cost. A team at $40 an hour that needs 900 hours costs more than one at $60 that gets it done in 500.
What the price is made of: hours per component
There's no reliable study that tells you how many hours an inventory module takes. Every vendor estimates its own based on experience and how it works. What you can demand is the breakdown. A serious quote lists hours per component, and this table tells you what to ask about each one.
| Component | What it covers | What drives the cost up |
|---|---|---|
| Customers and catalog | Customer records, products, prices | Multiple price lists, customer or volume discounts |
| Orders or sales | Creating, approving and shipping orders | Multi-level approvals, partial shipments, deposits |
| Inventory | Receipts, shipments and stock levels | Multiple warehouses, transfers, lots, expiration dates |
| Receivables and collections | Balances, due dates, reminders | Payment plans, partial payments, bank reconciliation |
| Field app | What sales reps or technicians use on the road | Working offline and syncing when back online |
| Each integration | Invoicing, online store, payment processor | Vendors with poor documentation or changing rules |
| Data migration | Moving customers, products and balances from the old system | Duplicate, incomplete or scattered data |
| Reports | Dashboards and lists for management | Metrics that are calculated by hand today and nobody has defined |
With this table you can run the math yourself: hours per component times the rate. And if a component isn't in the proposal, it isn't included, no matter what the system is called.
The five things that move the number most
- IntegrationsEvery outside system brings its own rules, errors and changes. An integration isn't done the day it works. Someone has to watch it for as long as the other system exists.
- Offline modeAn app that saves orders without a signal and syncs later has to decide what happens when two people change the same record. It's one of the most expensive things to build and test.
- Your own business rulesTiered commissions, price lists by channel, approvals based on amount. Every exception you handle in your head today becomes code and test cases.
- Migration and data qualityMoving 2,000 clean customer records takes an afternoon. Moving 2,000 records with duplicates, typos in tax IDs and balances that don't reconcile is a project of its own.
- Roles and permissionsA system where everyone sees everything is cheaper than one where reps only see their own accounts, the warehouse can't see prices and management approves discounts.
The cost nobody puts in the proposal: maintenance
In 2001, Robert Glass collected the facts of software engineering that people keep forgetting. One of them: maintenance eats up 40% to 80% of a system's total cost, 60% on average. And 60% of that maintenance is enhancements, not bug fixes (Glass, 2001). It's known as the 60/60 rule.
Run the numbers on the earlier example. If development costs $30,000 and that's 40% of the total, you'll spend another $45,000 maintaining and improving the system over its life. That's not an overrun. It's what happens with software people actually use, because your business changes and the system has to change with it.
Many agencies quote maintenance at 15% to 25% of the development cost per year. That's an industry rule of thumb, not a study, but it gives you a ballpark: on $30,000 of development, $4,500 to $7,500 a year.
Whatever the vendor calls it (maintenance, support or license), a monthly fee should cover:
- The server the system runs on, plus backups.
- Bug fixes, at no extra charge.
- Security updates and updates to the libraries the system depends on.
- A support channel with a response time in writing.
- Which small changes are included, and where a change starts being billed as an enhancement.
Why small projects turn out better
Research on software projects agrees on one thing: risk grows with size. McKinsey and the University of Oxford studied more than 5,400 IT projects. Those over $15 million ran 45% over budget and 7% over schedule on average, and delivered 56% less value than expected (McKinsey, 2012).
Flyvbjerg and Budzier looked at 1,471 projects. The average cost overrun was 27%, but 1 in 6 blew up: in those cases the average overrun was 200%, and schedules ran almost 70% long (Flyvbjerg and Budzier, 2011).
For ERP rollouts, Panorama Consulting's 2026 report found that 30% went over budget. The two most cited causes were additional technology (54.9%) and more scope than planned (51.0%) (Panorama Consulting, 2026). The sample is small and skews toward large companies, but the cause shows up at any size: scope grows along the way.
The practical takeaway for a small business: don't price out the whole system at once. Start with the process that costs you the most today, on a fixed timeline, use it for a month, and then decide what's next.
Set the appetite before the design
Shape Up, the method 37signals uses to build Basecamp, has an idea that works for any software purchase: appetite. Before designing a solution, the team decides how much time the problem is worth, then designs within that limit. In their words: "Estimates start with a design and end with a number. Appetites start with a number and end with a design" (37signals, Shape Up).
Applied to your business: before you ask for quotes, decide what solving the problem is worth to you. If three people spend half a day each week reconciling inventory in a spreadsheet, put a price on those hours for a full year. Then tell the vendor: "I have $15,000 and eight weeks. What can you deliver that fixes this?"
Instead of a number you can't judge, you get a scope proposal you can actually compare. Budget and timeline stay fixed; what flexes is how much gets built in the first release.
The rule: appetite first, design second. A fixed one-to-three-month timeline, like the one Efidis works with, is an appetite. Whatever doesn't fit goes into the next release.
Three-year cost: three ways to pay for the same system
The first invoice is misleading. Always compare over 36 months. This example assumes 15 users and $30,000 of development, which falls inside Clutch's common range. The off-the-shelf license prices are published list prices for two well-known systems on annual plans.
| Item | Traditional custom build | Off-the-shelf, per user | Efidis |
|---|---|---|---|
| Development or implementation | $30,000 (example) | Partner implementation, quoted separately | Based on scope, delivered in 1 to 3 months |
| Licenses: 15 users for 36 months | $0 | $4,833 (Odoo Standard, $8.95) to $43,200 (Business Central Essentials, $80) | $4,500 ($50 platform fee plus $5 per user, per month) |
| Maintenance and bug fixes | 15–25% of development per year: $13,500–$22,500 | Updates included; your own customizations billed separately | Servers, backups, security patches and bug fixes included; major upgrades are quoted |
| Three-year total | $43,500–$52,500, plus hosting | $4,833–$43,200, plus implementation and add-on modules | Development based on scope, plus $4,500 |
This table doesn't say custom software always wins. If your process fits Odoo out of the box, its licenses are cheap and it's a serious option. What the table shows is where the money goes in each model. With a traditional build, it's maintenance. With off-the-shelf software, it's per-user licenses that go up every time you hire someone. And with either one, it's whatever the quote leaves out.
Also check for licenses you pay for and nobody uses: at the large companies Zylo tracks, only 54% of purchased licenses get used (Zylo, 2026). A subscription audit gets you your own number in an afternoon.
License prices checked on October 22, 2026, on annual plans, before tax. The example leaves out the manual work that remains with each option: add it in if you know it.
12 questions for comparing quotes
Bring these to every vendor meeting and ask for the answers in writing. Two quotes are only comparable when they answer the same questions.
- What components does development include, and how many hours is each one?
- What is explicitly out of scope?
- How much is the monthly fee, and what does it cover: hosting, backups, bug fixes, support?
- What's the support response time, and through which channel?
- How are changes I request after signing billed?
- When will my team have a first version it can actually use?
- What happens if the project runs late because of the vendor?
- Are payments tied to deliverables I can test?
- Who migrates my current data, and what happens if it's messy?
- How many people get trained, and how many hours are included?
- Is my data mine, and in what format can I export it if I leave?
- What happens to the system and the code if I stop paying the monthly fee?
When custom software isn't the right call
Martin Fowler splits software into two groups: utility, which just has to work, and strategic, the kind that sets you apart from competitors. You buy utility software and adapt your process to it; strategic software is what's worth building (Fowler, 2016). To apply it to a specific process, use the 10-question scorecard for custom vs. off-the-shelf. In practice, custom isn't worth it when:
- The process is the same at every company, like accounting or payroll. Off-the-shelf software keeps up with tax and labor law changes, and you don't pay for that work.
- You have few users and a well-kept spreadsheet works for you. The trouble starts when several people edit the same file.
- You don't know yet how you want to work. If the process changes every month, stabilize it with a simple tool first.
- You need to start next week. Even a short build takes weeks.
- You can't budget for the monthly fee. A system with no maintenance decays, and in two years you have a more expensive problem than the one you solved.
How Efidis charges
In two parts, as laid out on our custom software page. Development is quoted in writing after a call where we learn how your operation works, and it's delivered in 1 to 3 months depending on scope. It can be a single module or a full system, in one web and mobile app.
The monthly license is a $50 platform fee plus $5 per user. The platform covers server and infrastructure maintenance, backups and quick patches, such as fixes for security vulnerabilities. It also includes an extended warranty (we fix any bug at no cost while the license is active), support over WhatsApp and replies in under 24 hours. Major upgrades, like a new module or moving the system to a new version of its underlying technology, are quoted separately.
Sources
- Clutch, Software Development Pricing Guide (2026) (opens in a new tab)
- Accelerance, 2026 Global Software Outsourcing Rates (opens in a new tab)
- U.S. Bureau of Labor Statistics, Occupational Outlook Handbook: Software Developers (opens in a new tab)
- Robert L. Glass, Frequently Forgotten Fundamental Facts about Software Engineering, IEEE Software (2001) (opens in a new tab)
- Bloch, Blumberg and Laartz, Delivering large-scale IT projects on time, on budget, and on value, McKinsey (2012) (opens in a new tab)
- Flyvbjerg and Budzier, Why Your IT Project May Be Riskier Than You Think, Harvard Business Review (2011) (opens in a new tab)
- Panorama Consulting Group, 2026 ERP Report (opens in a new tab)
- 37signals, Shape Up, Chapter 3: Set Boundaries (opens in a new tab)
- Zylo, 2026 SaaS Management Index (opens in a new tab)
- Martin Fowler, UtilityVsStrategicDichotomy (opens in a new tab)
- Odoo pricing (checked October 22, 2026) (opens in a new tab)
- Microsoft Dynamics 365 Business Central pricing (checked October 22, 2026) (opens in a new tab)