Per-employee SaaS spend climbed from $7,900 in 2023 to $9,100 by the end of 2025, according to Zylo’s annual SaaS Management Index. That’s a 15% rise in two years, on top of inflation that wasn’t doing anyone any favours. Subscription software was supposed to be the affordable alternative to building things in-house, and in some cases it still is. But the price line and the alternative-build line have moved in opposite directions long enough that custom is starting to look reachable for businesses that had previously written it off as out of budget.

This isn’t an argument for ripping out everything, but it is one for doing the proper maths.

What SaaS actually costs a UK SME in 2026

A mid-sized UK business now runs somewhere between ten and twenty separate SaaS subscriptions across sales, marketing, finance, HR, and operations. Cost-per-employee figures vary by source, but the direction of travel is consistent: spend is up, headcount has not changed proportionally, and the curve has steepened in the last 18 months.

There’s also a UK-specific waste figure that’s worth pausing on. Research published by SME Today put the annual cost of unused SaaS subscriptions at up to £10,000 per business, with around 39% of organisations admitting they don’t fully use the software they’ve paid for. Roughly 40% have at least one or two redundant tools running in parallel. None of this is new, exactly. What’s new is that the cost of carrying that bloat has materially increased.

The reason most businesses don’t audit and consolidate is that cancelling subscriptions takes time. Multiplied across twenty subscriptions, it becomes quite a tedious task, and there’s always something more pressing. So the spend keeps on creeping up.

Why prices are rising faster than people expected

Vertice’s SaaS Inflation Index for 2026 puts SaaS price inflation at 12.2%, which is roughly 4.5 times the general inflation rate across G7 economies (2.7%). That’s the headline figure most procurement teams have already encountered. The detail underneath it is more interesting.

Standard annual increases sit in the 8 to 12% range. Aggressive vendors are pushing 15 to 25%. The effective rise can be higher still once you account for what analysts have started calling shrinkflation: feature tiers being reshuffled so what you used to get on a mid-tier plan now requires the enterprise tier, usage caps appearing on previously unmetered functions, and AI features bundled in at a markup. CFO Dive reported that 28% of SaaS contracts experienced shrinkflation during 2024 alone. The SaaStr 2025 price-surge analysis pegged the proportion of vendors using AI bundling to mask underlying price increases at around 60%.

Vendor consolidation has reduced competitive pressure in several categories. Switching costs are high enough, and migration painful enough, that paying the increase is often the path of least resistance. Vendors know this. Pricing has adjusted accordingly.

What custom-built software costs by comparison

The numbers here are well-documented for the UK market. A recent breakdown of UK pricing for bespoke software puts internal tools and minimum viable products at £10,000 to £30,000, customer-facing platforms at £30,000 to £150,000, and enterprise-scale solutions at £150,000 and above. Specific examples in the same range: customer portals £30,000-£80,000, e-commerce builds £50,000-£150,000.

Year-one ownership cost adds £5,000 to £20,000 on top of the build, covering hosting (typically £50-£500 per month), third-party API services, and data migration where it applies. Annual maintenance is usually budgeted at 15-25% of the original build cost.

Two things stand out when you compare these figures to subscription totals. First, the costs are predictable. Once a fixed-price quote is signed, the build cost doesn’t get re-priced because the vendor decided to bundle AI. Second, the running cost after year one is modest, especially relative to what subscription bills do over the same period. A £40,000 build with £8,000 of annual maintenance behaves very differently to a £30,000 SaaS bill that grows 12% a year.

The crossover point

When does the maths actually flip? The honest answer is that it depends on three things: how many users you have, how much workflow customisation you need, and how strategic the data is to your business.

The cost guide above includes a worked example that’s useful to anchor against. A £50,000 system saving 20 hours of weekly admin time at £30 per hour blended cost gives you a 1.7-year payback period, with a five-year return on investment of around 188%. Industry benchmarks for custom builds tend to land between 200% and 300% over five years, so the example is on the conservative side. The figure that matters is whether a UK bespoke software development team can deliver something that recovers the build cost inside two or three years and continues paying out after that.

Seat-based pricing is where the comparison becomes most painful for growing businesses. A 30-seat licence at £40 per user per month is £14,400 a year. At 100 seats it’s £48,000. The vendor’s cost to serve a hundred seats isn’t four times the cost to serve thirty, but the price almost always is. Custom-built systems generally don’t price by seat, which means the cost curve flattens once you’ve paid for the build.

Data ownership and integration are the other two pieces. If your operational data is locked inside a SaaS platform that doesn’t expose it cleanly, you’re paying for the privilege of borrowing your own information. If three of your most important workflows depend on data flowing between four separate SaaS products that don’t natively talk to each other, you’re paying twice: once in licence fees, once in the integration work needed to make them cooperate.

What this doesn’t mean

Custom isn’t the right answer for everything, and pretending otherwise would be misleading. SaaS is the right tool for commoditised functions. Email, video conferencing, payroll, accounting, project tracking at the simpler end. The product is mature, the workflow is similar across thousands of businesses, and the per-seat cost is justified by what you’d otherwise have to build and maintain.

Custom makes more sense where the workflow is the differentiator, where seat-based pricing has reached uncomfortable scale, or where data ownership and integration have strategic weight. Early-stage businesses with unproven processes are usually better off on SaaS until the workflow has settled enough to be worth codifying. There is no universal answer.

A practical way to evaluate

The framework worth running once a year looks something like this. Count your tools. Count your seats. Project five-year SaaS cost at a realistic inflation assumption (12% annually is currently the defensible number, though some categories are higher). Compare that against an amortised build cost: the build itself plus five years of maintenance and hosting. Add back the labour cost of every workaround and integration script your team currently maintains because the SaaS stack doesn’t quite fit. Factor in the value of owning your data outright.

What surprises businesses past the 50-seat mark, or businesses with workflow-specific needs the SaaS market doesn’t serve well, is how often the build option comes out cheaper over five years than the subscription option. Not always. But often enough that the comparison is worth running. The cost of running the numbers is a couple of hours of someone’s time. The cost of not running them is whatever the next 12% increase turns out to be.