Custom factory software cost
A factory with 150 people across three shifts asks for a software quote. The vendor prices it per seat. Every operator who touches a terminal needs a license. That is not a purchase, it is rent, and it scales with the one thing a growing factory keeps adding: headcount. So the useful question is not what factory software costs. It is what you are renting, what you are owning, and how much scope you actually need.
A targeted custom tool, such as digital quality control or shift scheduling, runs roughly $10,000 to $30,000 and takes six to twelve weeks. A plant-wide execution system runs $60,000 to $130,000. Deep multi-plant machine integration starts around $150,000 and climbs from there. Those figures cover software only. Shop-floor hardware, tablets, scanners and floor Wi-Fi come out of a separate budget.
Takeaways
- Scope sets the price, not software quality. Workflow tooling and machine control are different projects with different risk profiles.
- Per-seat licensing punishes multi-shift factories. Custom shifts the spend from recurring rent to a one-time build plus flat hosting.
- Legacy machines without network cards are the largest single source of budget surprise.
- Solving one bottleneck first beats a full platform rollout, for both payback and operator buy-in.
What you are actually buying when you buy factory software
People use the phrase factory software to mean four different things. Separating them is the first step to reading a quote properly.
Enterprise resource planning runs the business. Accounting, purchasing, high-level inventory, customer orders.
Manufacturing execution systems, or MES, run the floor. An MES turns a production order into a digital traveler, records scrap and downtime, and tells the operator at each work center what to do next. Most people searching for factory software pricing want this layer, whether or not they use the acronym.
Supervisory control and data acquisition, or SCADA, talks to the machines. It reads temperatures, spindle speeds and fault codes straight from programmable logic controllers.
Point tools solve one problem. They replace a single shared spreadsheet with a secure multi-user application that several people can use at once without overwriting each other.
A point tool is not an execution system, and an execution system is not machine control. That distinction accounts for most of the spread in published price ranges.
The three cost tiers, and what each one buys
Published benchmarks for custom manufacturing software and MES work cluster into three tiers. Find the row that matches what you are actually asking for.
| Tier | What it covers | Cost | Timeline |
|---|---|---|---|
| Single-problem tool | QC logging, digital dispatch, scrap tracking, shift scheduling | $10,000 to $30,000 | 6 to 12 weeks |
| Plant execution system | Operator terminals at each work center, digital travelers, downtime taxonomy, scan on and scan off, two-way ERP sync | $60,000 to $130,000 | 12 to 16 weeks |
| Multi-plant platform | Dozens of PLC controllers, OPC UA and MTConnect, lot and serial genealogy, finite scheduling | $150,000 to $400,000+ | 6 to 12 months |
Between the first two tiers sits a common middle option: a single value stream or single-cell MES pilot, scoped to one line before anyone commits to a plant-wide rollout.
Where you buy changes the baseline. Industrial automation rate benchmarks put Southeast Asian development at roughly $40 to $100 an hour, Europe and the UK at $65 to $240, and the United States at $100 to $250. That is a different cost base, not a discount, and it is the main reason a studio working in Thailand can deliver a targeted departmental build for $10,000 to $25,000 (https://awcode.com/factories).
One honest caveat. If your project genuinely belongs in the third tier, with dozens of controllers and serial genealogy, you need an automation integrator with controls engineers on staff, not an application studio. Different scope, different kind of partner.
The build-versus-buy math for a multi-shift plant
Commercial off-the-shelf software earns its price in real situations. It wins when your processes are genuinely standard, when the vendor's compliance certifications are the point of the purchase, or when nobody internally has the time to specify requirements from scratch.
It also carries a structural cost. Mid-market commercial ERP and MES implementations regularly run $80,000 to $200,000 or more in upfront integration consulting, plus $30,000 to $50,000 or more a year in subscriptions and maintenance (https://www.symestic.com/en-us/blog/mes/mes-software).
Most of those vendors bill per named user. Call it the floor worker tax. In a plant with 150 operators across three shifts, the licence line grows every time you hire. Custom software moves that spend into a one-time build, with ongoing cost reduced to hosting and support, which can start from around $400 a month (https://awcode.com/factories).
Treat that comparison as an interpretation of published benchmark ranges rather than a like-for-like quote. Your own numbers depend on seat counts, module selection and how much of the integration work your team can absorb.
The five things that move your quote
Five variables decide whether you sit at the bottom or the top of your tier. Each one comes with a question worth asking out loud before you sign anything.
Data capture method. Barcode scanning and kiosk entry are cheap. Pulling data from legacy fieldbuses and analog retrofits is not. Ask whether data is typed, scanned or read from a sensor.
Machine integration depth. Modern protocols such as OPC UA and MTConnect keep costs predictable. Legacy serial links, custom PLC code changes and safety-rated validation do not. Ask whether anyone needs to touch machine code to make this work.
Hosting and infrastructure. Cloud hosting or a hybrid server on the plant LAN is standard. An isolated on-premise rack with manual failover is a different budget. Ask where the system lives and who patches it.
Interface and language. A single-language web interface is the cheap end. Ruggedized, glove-friendly screens running multiple languages for a mixed workforce cost more. If your floor runs Thai, English and Burmese, that is a line item, not a nice-to-have. Ask what localisation is included.
Compliance and traceability. Basic batch tracking and internal scrap logs are straightforward. Full lot and serial genealogy with audit-trail compliance for ISO or FDA 21 CFR Part 11 is heavy engineering. Ask exactly what level of audit logging the price covers.
What is not in the software quote
Software pricing does not include the physical equipment on the floor. Tablets. Ruggedized touch monitors. Barcode scanners and RFID readers. Optical sensors. Mounting arms, cabling and enclosures. Wi-Fi coverage that actually reaches the far end of the shop.
Budget those separately, and expect to buy most of them from a different supplier. Do not assume the developer is supplying the screens.
Why old machines are the budget wildcard
Connecting twenty-year-old machinery is where estimates go sideways. A press with no network card cannot simply report into a dashboard.
Getting a basic cycle count out of it usually means retrofitting sensing and wiring on the machine first, so there is controls work to pay for before a single line of application code is useful. Nobody can price that from a phone call. If a vendor hands you a fixed price for legacy machine integration without walking your floor, they are guessing, and the risk lands on you later.
Pay for a scoping visit first. Then fix the price.
How long it takes and what fast should mean
A targeted tool takes eight to twelve weeks. A plant-wide execution system takes twelve to sixteen. Multi-plant platforms take six to twelve months.
The date that matters is not go-live. It is the day a working prototype running real production data appears on the factory Wi-Fi, because that is when operators start telling you what is wrong with it.
Good scoping starts on the floor, not in a slide deck. Two days on site interviewing line operators and shift managers, a usable prototype deployed inside six weeks, full delivery between weeks eight and twelve is a reasonable shape for a tier-one build (https://awcode.com/factories).
Start with one bottleneck
Factories asking this question are usually stuck between two bad options. On one side, un-versioned spreadsheets, missed handoffs and quality data that arrives a week late. On the other, a quarter of a million dollars and twelve months spent reshaping the plant to fit a vendor's template.
The middle ground is narrow and deliberate. Pick the single worst bottleneck. Shift scheduling with conflict detection is a common first build. So is a digital QC checklist with defect photo capture, or structured downtime logging.
A small first build earns operator trust, produces real data you did not have before, and tells you whether deeper machine integration is worth funding.
Questions to ask before you sign
Print this and take it to the pitch meeting.
- Who owns the source code and the production data when the project ends?
- Is licensing per seat or flat?
- What happens to us if the development agency disappears?
- Is machine integration fixed price or time and materials?
- Who pays for connector maintenance when the ERP vendor pushes an update?
- Where is the data hosted?
- What happens on the floor when the internet drops?
- What is included in support, and what gets billed as new work?
FAQ
Is custom cheaper than a mid-market execution system? Often, over a three-to-five-year view, because you avoid the recurring per-seat licences that scale with headcount. Not always. If your processes are standard and you need certified compliance out of the box, the commercial option can be the cheaper path.
Can we connect machines later instead of now? Yes. Build the workflow layer first, scheduling, quality and downtime, then add machine data capture as a second phase once operators are using the interface daily.
Do we need to be online for the shop floor to keep working? That depends on the architecture, so specify it early. A hybrid setup keeps a local server running on the plant LAN during an outage and syncs to the cloud when the connection returns.
Who owns the code? You should. Confirm in the contract that ownership of the custom source code and your production data transfers to you.
What does ongoing support actually cover? Typically hosting, security patches, bug fixes and routine maintenance. New features are usually quoted separately. Get the boundary in writing.
The right number depends entirely on scope, and anyone who quotes a factory without walking it is guessing. If you want to replace shared spreadsheets with tooling built around how your floor already works, look at how AWcode approaches factory software (https://awcode.com/factories).