Custom system or off-the-shelf: which one costs less in the end

Off-the-shelf is cheaper at the start, a custom system over the long run. That is the main thing to keep in mind when choosing: the price on the pricing page shows the entry ticket, but what you pay for is ownership — over years, for every employee and every bit of customisation. A ready-made SaaS product has a low entry cost and a bill that grows with your team and goes up every year. A custom system has a large first invoice and an almost flat tail once it is live.

Below: how to calculate the cost of ownership, why per-seat subscriptions keep getting more expensive, where the break-even point between the two models sits, and when the honest answer is to buy off the shelf.

What is cost of ownership, and why does the price on the pricing page mislead?

Total Cost of Ownership (TCO) is everything a system costs over its lifetime, not the figure on the pricing page. Acquisition, rollout, customisation, integrations, training, downtime, the cost of leaving the platform — all of it together. Gartner introduced the concept in the late 1980s, and its analysts make a separate point of it: indirect costs, such as bending the product to fit your processes and dependence on the vendor, are usually underestimated at the moment of choice.

The pricing page shows only the entry ticket. A monthly subscription looks cheap because it does not say three things out loud: you pay for every new person, you pay extra to make the product fit your processes, and you agree to an annual price rise the vendor never asks you about. A custom system works the other way round: it frightens you with the first invoice and then barely grows. Comparing the two models by entry price is like choosing a car by the cost of the first tank of fuel.

What makes up the bill for an off-the-shelf product?

An entry fee plus a fee for every person, forever. Boxed SaaS charges by the number of users per month, and that is the main, endlessly growing line item. Then come customisations for your processes, separate modules for integrations with 1C, the accounting and ERP platform most Russian companies run on, along with telephony and warehouse systems, training your team on someone else's logic, and the "exit cost" — the point at which your data and your habits are tied to a single platform.

Russian vendors use different models. amoCRM, a Russian CRM, charges per user: its Professional plan is around 1,699 rubles per person per month (the price for new customers from 1 September 2025). Bitrix24, a Russian CRM and collaboration suite, sells packages of users within a plan rather than strictly per seat. The low-code platform ELMA365 starts at 500 rubles per user per month in the cloud, and its licences went up by an average of 20% on 1 January 2025. Exact prices change quickly — check the vendors' own sites — but the model matters more than the figure: the bigger the team, the bigger the bill, every single month.

A separate line item is vendor dependence, which never appears on a pricing page. The deeper a platform grows into your processes, the more expensive it becomes to leave: data, integrations and employee habits turn into switching costs. There is a separate piece on why even a system you have paid for often fails to take root — why a CRM "doesn't stick".

What makes up the bill for a custom system?

A large first invoice and an almost flat continuation. Custom development is capital expenditure: you pay once to design and build a system around your process, and after that you carry only support, further development and infrastructure. There is no fee for each new user, no annual price rise from a vendor, no risk of "they raised the price" or "they left the market".

Support is usually estimated at 15–20% of the development cost per year — a rule of thumb from contractor practice, not an independent standard. You control the server or the cloud yourself. The key difference is elsewhere: you own the code and the data. Nobody can switch them off from outside, you can hand them to another team, and no one else's price list applies to them.

The flat tail after the first year is a property of the model itself. Whether it pays off is a question of scale. Which is where we go next.

Why does a per-seat subscription get more expensive every year?

Because the subscription software market is growing faster than inflation, and AI features are sold as a surcharge on every user. According to SaaStr analysis citing Vertice, SaaS prices rose roughly 11.4% year on year in 2025 — almost five times the rate of inflation in the G7 countries. The average business spends about $7,900 per employee per year on subscriptions alone, 27% more than two years ago.

A phenomenon of its own is the "AI tax". Vendors build AI features into the product and raise the price at renewal by 20–30%, whether or not you use that AI. Microsoft 365 Copilot costs $30 per user per month as an add-on to the core licences: for a team of 50 that is an extra $18,000 a year for a single add-on. Salesforce has pushed its top plan to $500 per seat per month — twice what it was five years ago. Slack added 20% in a year; Adobe raised its photography plan by 50%.

The Gartner forecast quoted by Rimini Street reads: "Certain costs to own and operate software, including support, will rise 35% by the end of 2025." The meaning for an owner is simple: the monthly model is a bet that the vendor will not raise the price and will not tie you to itself. Over the past two years that bet has not paid off once.

Want a system built around your process instead of bending your business to fit someone else's price list? Custom development at IncubeAi starts with working out where you are losing money and designing a system for your niche — under contract, with data held in Russia and support after handover. The team has 300+ automation and AI projects behind it.

Where is the break-even point: when does custom cost less than off-the-shelf?

Where the rising subscription line crosses the almost horizontal line of a custom system. The mechanics are not in doubt: the per-user fee scales linearly with your team and with the annual increase, while the cost of a custom system barely grows after the first year. Sooner or later the first curve overtakes the second — the only question is when.

The specific thresholds people quote vary, and all of them are estimates from an interested party: custom development contractors, not independent research. By those estimates the break-even point for small companies falls somewhere around 15–30 users, with payback in 1.5–2 years. You have to check against your own numbers: multiply the per-user price by the size of your team and by a horizon of 3–5 years, add the annual price rise, and compare that with one-off development plus support.

The logic is backed up by a large case of the same nature — renting versus owning. The company 37signals (which runs Basecamp and HEY), led by David Heinemeier Hansson, moved off the AWS cloud onto its own servers. The bill dropped from $3.2 million to $1.3 million a year, and the projected saving, in Hansson's words, came to "well over ten million dollars over five years — with faster hardware and far more storage". At scale, renting loses to owning — the same arithmetic as subscription versus custom system.

The decision breaks down along three axes:

When is off-the-shelf the right choice?

When the process is standard, the team is small, and the system is needed only for a while. If you have routine bookkeeping, a simple sales funnel and a dozen employees with no plans for multiple growth, an off-the-shelf product is almost certainly cheaper and faster. In that scenario capital spending on development simply does not have time to pay off, and paying extra for something "of your own" is a straight loss.

An off-the-shelf product is a proper working tool. The mistake is usually not in choosing a ready-made product but in trying to stretch it over a non-standard process through endless customisation. Custom development fails just as predictably: when there is nobody on the client's side who accepts the work and answers for the result, the project turns into exactly the "lottery" people fear. One author on Habr, a Russian tech community site, puts it bluntly: "Every failed project I have worked on has one thing in common — communication with the client was never properly set up."

The honest conclusion: if the three axes from the previous section point to an off-the-shelf product, buy one and don't overpay for engineering you don't need yet.

When does an off-the-shelf product become a trap?

When the process is non-standard, the team is growing, and dependence on the vendor becomes a business risk. In those three situations the low entry cost turns into a growing "tax", and the longer you stay inside the product, the more expensive it becomes to get out.

• A unique process. If the way you work is a competitive advantage, bending an off-the-shelf product to fit costs two or three times more on top, and you still run into its ceiling.
• A growing team. The fee for every new employee turns into a permanent "tax on growth": the more successful you are, the more you pay for the mere fact of scaling.
• Critical independence. When data, core integrations or sanctions risk matter for survival, lock-in outweighs any arithmetic. Federal Law 58-FZ (April 2025), a Russian law on critical infrastructure, requires operators of critical infrastructure to move to domestically produced software; sole traders were exempted from those requirements, but the direction has been set.

The trap closes quietly. While the team is small, per-seat pricing goes unnoticed; by a hundred employees it is already a budget line, and data and habits have grown into the platform so deeply that moving costs as much as a new system. If you have reached the stage where your various services no longer talk to each other, there is a useful piece on how to connect 1C, CRM, your bank and Telegram into a single loop.

How much does custom development cost, and how long does it take?

A large first invoice — yes, but "slow and unpredictable" is the result of a badly run project; the format itself has nothing to do with it. A budget turns into a lottery where there are no clear requirements and no engaged client. A managed project is built differently: it starts with a goal stated in money, then the architecture appears, and only then the code — in stages, each of which can be accepted and checked.

For us that is four steps:

1. Understand the task: get into your processes, find where money and time are being lost, and fix the goal in money terms.
2. Design the system: an architecture for your niche and scale — what we are building, how we integrate it, in what order.
3. Build and roll out: assemble the system, connect it to your services, migrate the data, train the team.
4. Support: develop the system after launch, because the business changes and the code has to change with it.

The fair comparison for the first invoice is the sum of all subscription payments over the years the system will live. Then capital expenditure falls into place: it is the purchase of an asset that stays your property, whereas rent has to be renewed forever. Similar logic applies to moving off spreadsheets — we covered it in the piece on when a business should leave Excel behind.

How do you choose without overpaying?

Work out the cost of ownership over 3–5 years for both options and check it against the three decision axes. Counting the full sum over the system's lifetime is what reveals the real difference that the entry price hides.

1. Add up the TCO of the off-the-shelf product: price per user × team size × 12 months × horizon in years, plus the annual price rise, plus customisation and integrations.
2. Add up the TCO of a custom system: one-off development + support (15–20% a year as a guide) + infrastructure.
3. Map it onto the three axes: number of users, how non-standard the process is, lifetime of the system.
4. Assess risk separately: does owning your data and staying independent of a vendor matter to you? If it does, that outweighs the plain arithmetic.

If the process is standard, the team is small and the horizon is short — buy off the shelf. If the process is non-standard, the team is growing and the system is needed for years ahead, building your own pays off both in money and in control. There is no universal answer; there is your calculation.

If you would like to work through that calculation together, tell us about your task. We will go through your processes, work out the cost of ownership for both options honestly, and if the off-the-shelf route turns out to be cheaper, we will say so plainly. We work under contract, with data held in Russia and support after handover.

Sources

• Gartner IT Glossary — definition of Total Cost of Ownership
• Microsoft — official pricing for Microsoft 365 Copilot for business ($30/user/month)
• SaaStr — The Great SaaS Price Surge of 2025 (SaaS inflation +11.4%, the "AI tax", Salesforce, Slack, Adobe)
• Rimini Street — Gartner Predicts 2022 forecasts on software ownership costs rising 35%
• ELMA365 — announcement of a 20% average licence price increase from 2025
• Data Center Dynamics — 37signals on leaving the cloud: around $2 million saved per year and over $10 million across five years
• Habr — "Why custom development turns into a lottery" (anti-patterns and communication)