SaaS vs custom build: how to compare the real costs as prices rise

Software is getting more expensive on both sides of the build-or-buy question. SaaS prices are rising through straightforward increases, renamed plans, features moving up a tier and AI add-ons. Meanwhile, the hosting, licences and build pipelines behind a custom system are moving too, some up and some down. The fair way to decide is to cost both options over five years using what you actually pay, then judge fit separately from price.

What has changed in software pricing this autumn?

In October 2026, Atlassian raised prices across several of its cloud products, according to a price tracker on dev.to. For teams of up to 100, the monthly per-user prices moved as follows:

• Jira: $9.05 to $9.70 (+7.2%)
• Confluence: $6.70 to $7.15 (+6.7%)
• Bitbucket Premium: $7.25 to $8.00 (+10.3%)
• Jira Product Discovery: +5%

None of those rises is dramatic on its own. But they land on tools that are hard to leave once your projects, documentation and code history live inside them. Small yearly rises on sticky tools compound.

Salesforce took a different route. In September it renamed its editions and raised prices at the same time. Core replaces Enterprise and goes from $175 to $195 per user per month. Advanced replaces Unlimited and goes from $350 to $395, with AI credits included, as tracked by SaaS Price Alerts. When the plan name changes and the bundle changes, comparing this year's bill with last year's gets harder.

That pattern is becoming common. A Hackernoon piece from 20 October by someone who tracks SaaS pricing pages lists the quieter ways prices go up: features moving from the middle plan to the top one, plans being replaced, grandfathered prices expiring, units being renamed and AI features sold as separate add-ons.

Step 1: What are you really paying for your SaaS tools today?

Start with invoices, not pricing pages. Pull the last two or three years of bills for your main subscriptions and put them in a single spreadsheet.

For each tool, record:

• The total annual cost, not just the per-user price
• The number of seats you pay for, and how many people actually log in
• Any add-ons, including AI features, storage or extra integrations
• The year-on-year change in total cost

The total matters more than the list price. A 7% rise on a per-user price can turn into a much bigger jump if you were also pushed onto a higher plan.

Step 2: Where are the hidden rises in your plans?

Go back through the same invoices looking for the patterns in the Hackernoon list. Did a feature you rely on move up a tier? Was your plan retired and replaced with something more expensive? Has a discount or grandfathered rate ended, or is it about to?

Note AI add-ons separately. If you're paying for AI credits or features that nobody uses, that's money spent on the vendor's roadmap rather than your needs. This is also where you'll start to see whether a tool's price is drifting steadily away from the value it gives you.

Step 3: What would a custom build cost to run for five years?

This is where honest comparisons often fall down. A custom system isn't free once it's live. It has its own running costs, and they change too.

A few recent examples show both directions:

• Microsoft licences through resellers. From 1 October 2026, annual CSP subscriptions billed monthly rise 5% at renewal, covering SQL Server, Windows Server, Client Access Licences and System Center, according to ToolRelief's tracker. Paying annually upfront, or on a month-to-month term, isn't affected. How you pay changes the bill.
• Hosting. Fly.io raised shared-CPU machine prices by about 13% and extra RAM by about 20% from 1 October 2026, per PriceHiked.
• Costs going down. Oracle cut Java SE prices for organisations under 3,000 employees in its 8 October price list, and GitHub cut hosted Actions runner prices by up to 39%, as logged by Licenseware.

For any custom option, list five years of costs: the initial build, hosting, database and server licences, build and deployment pipelines, security updates, support and planned improvements. Include the billing choice for each licence, because the same SQL Server licence can cost different amounts depending on how it's paid for.

If you're not sure what a sensible architecture and hosting setup would look like, an architecture review before you commit is cheaper than finding out later.

Step 4: Are you comparing like for like?

Once you have both columns, check they describe the same thing. A SaaS plan with bundled AI credits isn't directly comparable with a custom system that has no AI features, and a custom system that only covers your core workflow isn't comparable with a suite you use a fraction of.

Ask two questions of each SaaS tool:

• Can we export all our data, in a usable format, whenever we want?
• What would it cost, in time and money, to move away in three years?

If leaving would be painful, the tool's future price rises are effectively guaranteed costs. That should count against it in your comparison.

Step 5: How well does the tool actually fit how you work?

Price is only half the picture. Score fit separately, because a cheap tool that doesn't fit your process can cost far more in staff time than it saves on licences.

Signs that fit is the real problem:

• Staff keep spreadsheets alongside the tool to fill its gaps
• Data is typed into two or more systems by hand
• You pay for a higher plan to get one feature
• Your process has been bent to suit the software, not the other way round

Where those signs show up and SaaS costs keep climbing through plan changes and add-ons, a bespoke system built around how you work starts to pay off. Where a tool does a standard job well and its rises are modest compared with the cost of building, hosting and maintaining your own, keeping it is usually the sensible call.

Step 6: Could a mix of SaaS and custom work better?

Build-or-buy rarely needs to be all or nothing. Standard jobs such as accounts and email are well served by established products. The part of your operation that makes you different, such as quoting, scheduling or a customer portal, is where a custom build earns its keep.

The two can be linked so data flows between them without double-keying. Business process automation is often the glue that makes a mixed setup work. Building in short sprints also means you see working software early and can stop, adjust or expand based on real results rather than a big upfront bet.

What should you do next?

Work through this checklist before your next significant software decision:

• Gather two to three years of invoices for your main SaaS tools
• Mark real rises, including plan renames, features moved up a tier and AI add-ons
• Count paid seats against active users
• List five years of running costs for any custom option, including licences, hosting, pipelines and support
• Check how each licence is billed, as monthly-billed CSP renewals now cost more
• Confirm you can export all your data from each SaaS tool
• Score fit separately from price: workarounds, double-keying and unused features
• Consider keeping standard tools and building only the part that sets you apart

The goal isn't to prove one side right. It's to see both costs clearly, so the decision rests on evidence rather than a headline price.