Off-the-shelf software is cheap while your process matches the vendor's plan. The moment it stops matching, the subscription keeps its price and you start paying the difference in staff time, workarounds and fees that were never on the quote. The signals are specific and you can check every one of them this afternoon.
Most of what a small business runs on should be bought, and the good products are very good. The skill worth having is noticing the point at which a tool stops serving the process and starts defining it, because that point arrives quietly and the invoice never mentions it. Each vendor figure below was read on that vendor's own published pricing page in September 2026, and plans change, so confirm the current terms before you decide anything on them.
Eight signals that the fit has gone
- Your bill tracks a unit that has nothing to do with your revenue. Zapier bills by task, counting a task whenever it successfully completes a unit of work for you, with failed actions excluded. The free plan carries 100 tasks a month and Professional starts at 750. That model is fair for a tool that works on your behalf. It turns awkward when one order fires a long chain of steps, because the automation cost grows with order volume while your margin per order stays flat.
- A plan cap has started shaping the work. Xero's UK entry plan, Ignite, is priced per organisation and covers quotes and 20 invoices sent, with 10 bills entered; past the cap you move up a plan. Caps like that suit a very small business. Watch for the moment someone starts holding invoices back to next month, or raising one consolidated invoice to stay inside the allowance. At that point the plan is designing your billing process and the upgrade has already been earned.
- Growth in your list is growth in your bill, and the ratchet only turns one way. HubSpot Marketing Hub is priced by seat and by marketing contact tier, with 1,000 marketing contacts included at Starter, and its pricing page states that once you select a contact tier you cannot downgrade to a lower one until your contract renews. Every campaign that builds the list moves you toward the next tier. Nothing moves you back until renewal, which makes list hygiene a commercial task.
- Access to your own system is rationed by seat. Shopify's UK plans include no additional staff accounts on Basic, up to five on Grow, up to fifteen on Advanced and unlimited on Plus. Seat limits are a legitimate way to price software. The failure mode is what people do about them: a shared login, one browser session that three people use, and an audit trail that can no longer tell you who changed the price.
- A percentage sits on every transaction because you picked your own provider. Shopify applies third-party transaction fees when you use a payment provider other than Shopify Payments: 2% on Basic, 1% on Grow, 0.6% on Advanced and 0.2% on Plus. If your acquirer, your bank or your accountant has good reasons for the provider you use, that percentage is a standing charge on trading, and it scales with success.
- Two systems hold the same record and neither one is right. The customer exists in the shop, the accounts package and the mailbox, with three spellings and two phone numbers. Nobody can say which is authoritative, so everyone checks all three. It is the most expensive item here, and the cost hides in small pieces across everybody's day. Here is the test: ask three people which system holds the correct phone number for your largest customer. Three different answers means you are paying for the disagreement every week.
- The workaround has a name, a rota and a spreadsheet. Somebody exports on a Tuesday, reformats the columns, and imports somewhere else. It works, it has worked for two years, and the business now has a single point of failure with a holiday entitlement. When the workaround appears in a job description, it has become part of your operating model.
- The export is the only integration you have. A tool with no usable API leaves you moving CSV files by hand for as long as you own it. Before you renew anything, check two things: whether there is a documented API, and whether you can run a full export of your own data without asking support. Those two answers decide whether the tool can ever be connected to anything else.
The cost that never appears on the invoice
Add up the subscriptions and the number looks manageable. The larger figure is the one nobody bills you for, so count the minutes. Time the re-keying for one week, then multiply it by the people who do it and the weeks you trade. Add the errors that come out of the re-keying, the calls to fix them and the credit notes at month end. Add the decisions nobody makes because the report needed to make them takes too long to assemble.
That figure is the honest budget for change. It is usually larger than the build people are afraid to quote for, and it is already being spent, quietly, every month.
How to compare a subscription with a build
Do the comparison over three years, because a build is a capital-shaped cost and a subscription is a rental. One year flatters the subscription and five years flatters the build, so three is the fair middle.
On the subscription side, list every tool the change would touch, its current annual cost, and the growth built into its pricing model. Growth is the part people leave out. A per-seat tool grows with hiring. A contact-tier tool grows with marketing success. A per-task tool grows with order volume. Take last year's actual growth in the relevant unit and apply it forward, using the vendor's published tiers.
On the build side, list the build cost, the hosting, the maintenance and the support arrangement, and be honest that a custom system needs all four for as long as you run it. Add a line for the second phase, because there is always a second phase once people see the first one working.
Then put the workaround cost on both sides of the ledger. It stays on the subscription side for as long as the tool stays. It comes off the build side on the day the build goes live. That single line moves more comparisons than anything else, and it is the one line most quotes never mention.
What to check before you commission anything
Start with the root cause. Several workarounds usually share one, and it is almost always two systems holding the same record with no agreement about which one leads. Fix the ownership of the record and much of the list disappears.
Then check what you can take with you. Every tool you plan to keep needs a documented API and a full export you can run yourself. Every tool you plan to replace needs the same, because a migration you cannot perform is a migration that never happens.
Then decide what you want to own at the end. Source code, data, hosting account and domain, in your own name. Open-source foundations make that easier: WooCommerce describes itself as the open-source commerce platform for WordPress, and offers full control of your checkout, your data and your costs. A build on foundations like that is yours in a way a subscription never is.
When off-the-shelf remains the right answer
Plenty of the time. Accounting, payroll and tax filing are the clearest cases, because the vendor tracks legislative change on your behalf and takes on the burden of keeping you compliant with it. The same goes for email, calendars, video calls and card payments, where the standard product is excellent and your version would be worse.
The rule that holds up: buy the things that are the same in every business, and build the things that are the reason customers choose you. Your quoting logic, your production workflow, your service routine and your reporting are usually in the second group. Your general ledger is not.
There is a middle option that gets overlooked. Keep the tool and buy back the part of it that is costing you, by connecting it properly to whatever it should have been talking to all along. An integration is cheaper than a replacement, it preserves the training your team already has, and it removes the workaround, which was the actual problem. Reach for it before you reach for a rebuild.
Where a build actually starts
We build software, web and apps around the way a business already works, which usually means an integration layer, a shared database or one interface over tools that are staying. The first session walks the workarounds, the rota and the spreadsheet, until the root cause is on a whiteboard and can be priced. Feature lists come later, once the cause is agreed.
If you recognise three or more of the eight signals above, you have enough evidence to start that conversation with numbers. Bring the invoices, the export files and the name of whoever runs the Tuesday spreadsheet. Those three things describe the problem better than any requirements document written from memory.