Somewhere in your business, right now, a subscription is quietly renewing for a tool nobody has opened since March. I know this because I have found that exact tool in every company I have ever audited, including my own, and I have never once found just one.
A tech stack audit is the process of listing every piece of software your business pays for, working out who actually uses it, and cancelling or downgrading whatever fails that test. If you want to know how to cancel the software you're not using, the honest answer is that the cancelling is the easy bit. The hard bit is finding it all, and then having the nerve to press the button.
The Short Answer
Pull every software charge from your last twelve months of bank and card statements, put them in a spreadsheet with an owner and a renewal date, sort each tool into keep, downgrade, merge or cancel, then work through the cancel and downgrade piles before the next renewal hits. Expect to find that roughly a third to a half of what you pay for is either unused or overprovisioned, because that is what the industry data keeps showing. Do the whole thing in a fortnight, put a recurring review in the diary, and you will not have to do the big painful version again.
That is the whole method. Everything below is the detail that stops you making the mistakes I made the first three times.
Why Nobody Cancels Anything
The numbers on this are consistently grim, and worth knowing before you start, because they will stop you feeling like the only person who let it slide. According to Zylo's analysis of unused software drawn from its 2026 SaaS Management Index, 46 percent of applications go underutilised or unused, licence utilisation averages just 54 percent, and the average company is managing around 305 SaaS applications. Zylo also found that business units now control 81 percent of software spend directly, with IT managing only 15 percent, which is exactly why so much of it slips through.
Zylo's headline figure of $19.8 million in annual waste per organisation is a US enterprise number and should not be read as a UK small business benchmark. It does translate in one sense though: the proportions hold. Vertice's analysis of unused SaaS applications, drawn from over $30 billion of processed spend, found that 66 percent of all SaaS licences are either entirely untouched or surplus to requirements. Vertice splits that into 15 percent pure shelfware, meaning platforms with zero activity, and 51 percent underutilised, which it defines as less than half of purchased capacity being used.
Shelfware is a word you will see a lot. It just means software you bought, integrated and then forgot about. Vertice notes these tools persist mainly because of auto-renewals and a lack of centralised visibility, and that matches my experience precisely. Nobody decides to keep paying for a dead tool. They simply never decide to stop.
On the consumer side the picture is the same. The Department for Business and Trade estimates that UK consumers spend around £1.6 billion a year on subscriptions they do not want, across roughly 155 million active subscriptions. If that is what happens to people managing a handful of personal subscriptions, imagine what happens across a business with forty tools and three people who can add a card.
As the team at Procr put it in their piece on SaaS licence wastage, buying is frictionless and reviewing is not. A new tool takes ten minutes on a company card. Removing it requires someone to notice, care, and find the login.
Start With the Bank Statement, Not the App List
The biggest mistake people make in a tech stack audit is starting by asking staff what they use. You will get an honest but incomplete answer, because people forget the tool they signed up for two years ago, and they do not know what a colleague on another card has been buying.
Start with the money instead. Export twelve months of transactions from every business bank account, every company card, and any personal cards staff use for expenses. Twelve months, not three, because annual renewals only show up once and they are usually the expensive ones.
Filter for anything that looks like software, including the payment processors that sit in front of smaller tools and any recurring charge in dollars or euros you cannot immediately explain. I always find at least two charges nobody in the building can identify on the first pass.
Build a spreadsheet with these columns: tool name, what it does in one line, monthly or annual cost in pounds, billing currency if it is not sterling, who pays for it, who owns it internally, renewal date, notice period, number of seats paid for, number of seats actually used, and a decision column that stays empty for now. Where a tool bills in dollars, record the sterling amount that actually left the account rather than the list price, because that is your real cost including the exchange rate.
Then, and only then, cross-check the spreadsheet against what people tell you they use. The gap between the two lists is your first savings pile.
The Four Piles
Once the list exists, every tool gets sorted into one of four decisions: keep, downgrade, merge or cancel. I make myself put every single row into one of them, even the obvious ones, because doing it row by row is what surfaces the surprises.
Keep is for tools that are used, that do something no other tool in the stack does, and that are on the right tier. Be strict. A tool being used is not the same as it being needed. If your studio's project tracker is used by two people once a month, that is a merge into whatever you already have, not a keep.
Downgrade is where most of the real money sits, and I will come back to it in detail. It covers tools you use but at a tier that gives you features nobody touches, and seat counts that include people who left in 2024.
Merge is for duplicates. The classic pattern is three note taking tools, two file storage services and two ways to hold a video call, each adopted by a different person who liked a different thing. Pick one, migrate, cancel the rest. Zylo's data on business units controlling most spend explains why duplicates breed: nobody is looking across the whole picture.
Cancel is for the shelfware. Zero logins, no owner, or an owner who cannot remember what it was for. This pile is usually smaller than people expect in count, but it is the most satisfying to clear.
The Questions I Ask Before Cutting Anything
Rather than agonising over each row, I run every tool through the same short set of questions. The answers do the deciding.
Has anyone logged in during the last ninety days? Most tools have an admin panel that shows last activity per user. If yours does not, that absence is itself a warning sign.
If it disappeared tomorrow, what would break? Be concrete. If the answer is nothing, or a mild inconvenience, cancel. If it is a client deliverable or a legal obligation, keep, and write that reason in the spreadsheet so next year's you does not have to work it out again.
Is there something already in the stack that does this? Video calls, chat, task lists, basic design work and simple automations are all covered by several suites most businesses already pay for. Microsoft 365 Business Standard, for example, bundles Teams, Bookings, Planner, Forms, Loop and Clipchamp alongside the desktop apps at £10.80 per user per month on the annual plan, ex VAT, according to Microsoft's UK business pricing page. If you have that, you may not need a separate scheduling tool, a separate task board or a separate lightweight video editor.
What does the exit actually cost? Check the notice period and whether you are on an annual contract paid monthly. This one question is where people get stung, and it deserves its own section.
Where the Money Actually Hides
The cancellations are the headline, but in every audit I have done, downgrades and seat clean-ups have saved more than outright cancellations. Here is where to look.
Seat counts first. Compare the number of licences you pay for against the number of active named users. The delta is almost always positive, because offboarding usually stops at disabling someone's email and never reaches the eleven other tools they had access to. As the guide from Gart Solutions on unused SaaS seats points out, granting a seat has a clear owner and trigger, while removing one usually does not. Remove the orphaned accounts, drop the seat count at the next billing cycle, and you have saved money without cancelling anything.
Tiers second. Almost every vendor sells a plan ladder where the middle rung is the default and the top rung is what the salesperson nudged you onto. Look at what the upper tiers actually add and ask whether anyone uses it. Microsoft is a useful worked example because its UK pricing is public. Business Basic is £5.40 per user per month annually, Business Standard is £10.80, and if your team lives in the browser and does not need installed desktop apps, Basic does the job. There are also versions without Teams at £4.20 and £8.30 respectively. Texaport, a UK managed service provider, notes that Business Premium held at £16.90 per user per month when Basic and Standard rose in the July 2026 price increase, which narrowed the gap and makes it worth re-checking whether the security features in Premium are ones you actually need rather than defaulting to Standard.
Bundled AI is the newest tier trap, and it deserves a mention because the regulator has noticed it too. In July 2026 the CMA opened an investigation into how Microsoft communicated changes to its consumer Microsoft 365 Personal and Family plans, after Copilot was added and customers were moved at renewal onto plans costing £25 a year more unless they actively switched to a Classic plan. The CMA's announcement of the Microsoft investigation is careful to say no conclusion has been reached. For your audit the lesson is simpler: check whether any tool you pay for has quietly added an AI feature and a price rise to the same renewal, and whether a cheaper non-AI tier still exists.
Annual contracts paid monthly are the third place money hides, and the nastiest. Adobe's creative subscriptions are the well known example. As Bristows explain in their note on the CMA investigation into Adobe's exit fees, the annual billed monthly plan commits you to twelve months, and cancelling after the initial fourteen day cooling off period triggers an early termination fee of 50 percent of the remaining annual cost while you also lose access at the end of the current billing month. The CMA is examining whether that is fair, with an update expected in autumn 2026, but until anything changes that is the deal you signed. If you are a musician or a small studio paying for a creative suite one person uses twice a year, do not cancel it mid-term in a fit of tidiness. Diarise the renewal date and cancel in the window instead.
Cancelling Cleanly Without Losing Data
Once the cancel pile is decided, there is an order of operations. Skipping it is how you end up locked out of two years of client files.
Export everything first. Most tools will let you download your data in some form, and most will delete it within a fixed window after cancellation. Microsoft's own terms state that when a business subscription is cancelled, all associated data is deleted, and it points to a retention document for the timings. Assume that is the norm everywhere. Get the files out, get them into a storage location you are keeping, and check they open.
Turn off auto-renew before you cancel outright, if the tool lets you. This sounds pedantic but it protects you if a cancellation request goes astray. Then submit the cancellation through the route the vendor specifies. Some want it done in the account settings, some want an email, some annual business contracts want written notice a set number of days before the renewal. Read the terms. The notice period column in your spreadsheet earns its keep here.
Get written confirmation. A screenshot of the cancellation screen and the confirmation email, saved somewhere you will find them. If a charge appears later, that is your evidence.
Cancel the card if you cannot cancel the tool. For a few stubborn vendors this is the only move that works, but it can put you in breach of contract inside a fixed term, so treat it as a last resort for genuine zombies.
Finally, remove the integration. Many tools are wired into your other systems through API keys, meaning the small tokens that let one piece of software talk to another. Leaving a dead tool's key live in your file storage or customer database is a security hole.
What UK Law Does and Does Not Do for You
This is where I see the most confident wrong advice, usually imported from American articles. So let me be clear about what actually applies here.
If you signed up as a consumer, meaning as an individual rather than in the course of a business, you have real protection. The Consumer Rights Act 2015 makes unfair or hidden terms non-binding, and the Consumer Contracts (Information, Cancellation and Additional Charges) Regulations 2013 give you a fourteen day cooling off period on most distance contracts. The Competition and Markets Authority enforces this, and since April 2026 it has been using new direct fining powers under the Digital Markets, Competition and Consumers Act 2024. White & Case note that the first such penalty, of £4.2 million, was for drip pricing, and Bird & Bird point out that those fines can now reach 10 percent of annual global turnover.
The bigger change is still coming. Chapter 2 of Part 4 of the Digital Markets, Competition and Consumers Act creates a new subscription contracts regime with a second fourteen day cooling off period after a free trial ends or a contract renews for twelve months or more, mandatory reminder notices before renewal payments, and a requirement that online subscriptions can be cancelled online without unnecessary steps. In the TLT briefing on the subscription regime being brought forward, Molly Efford explains that on 10 August 2026 the Prime Minister announced the go-live date is moving to January 2027, having previously slipped to spring 2027, and that the secondary legislation and guidance had still not been published at the time of writing. The government has said that providing an email address alone is unlikely to count as an online exit, and that cancelling a direct debit with your bank does not count either.
Here is the catch that matters for this article. All of that is consumer law. It applies when a trader deals with a consumer. If your limited company signed up for a tool, you are almost certainly a business customer, and none of it protects you. As SA Law's Clare Mackay put it in a piece for SME Today on auto-renewal clauses in business contracts, a business customer tied into an unwanted additional term because of an auto-renewal clause is unlikely to find any assistance from a court, because English law enshrines freedom of contract and it is no defence to say you did not read the terms. The Unfair Contract Terms Act 1977 does exist, but as the guide at MyRenewals on UK auto-renewal law notes, in a business context it mainly bites on liability exclusion clauses, not on renewal mechanics.
The practical consequence: as a business, your protection is your diary and your spreadsheet. The renewal date and notice period columns are not admin, they are the only legal remedy you have. Sole traders are a grey area and it depends on whether you were acting for purposes outside your trade, so if a large sum is at stake it is worth a conversation with a solicitor rather than assuming either way.
Should You Pay for a Tool to Manage Your Tools
There is a whole category of software, usually called SaaS management platforms, that exists to solve the problem this article is about. The irony is not lost on anyone who sells them. Whether you need one depends almost entirely on how many tools you have and how many people can buy them.
Under about fifteen tools and one or two card holders, no. A spreadsheet and a calendar reminder for every renewal date does the job, and adding another subscription to manage your subscriptions is exactly the behaviour you are trying to stop.
Between roughly twenty and seventy five tools, with several people able to spend, the maths starts to work. Cledara is the option I would look at first for a UK business, partly because it is built around a London based team and bills in pounds, and partly because its model is unusually practical for small firms. Each tool gets its own virtual card with a spending cap, so a vendor cannot quietly charge more than you allowed, and every subscription shows up in one place because the payment itself is the tracking mechanism. On Cledara's own pricing page, the Basic package is free if you pay for ten applications through its virtual cards in your first month, with a charge of £100 a month if you do not meet that condition. The Spend Optimisation add-on, which is the bit that flags duplicate tools, price rises and seat utilisation, is £200 a month or £1,500 a year, and the IT Management and Software Compliance modules are each £1,500 a year. Cledara says it does not add a markup on foreign currency card payments, which matters if half your stack bills in dollars. The trade-off is that it only sees what you pay through it, so anything on a bank transfer or a personal card stays invisible until you move it over.
Vertice is the other name you will see, and it is aimed at a different customer. Cledara's own comparison piece claims Vertice's plans start at around $30,000 a year, which is a competitor's characterisation and should be read as such, but it fits Vertice's positioning for companies with seven figure software budgets and dedicated procurement staff. If you are reading a guide on how to cancel software, you are not that customer yet.
I would steer a small business away from any platform that wants browser extensions on every employee's machine to discover shadow IT. For a fifteen person firm it is intrusive, it raises its own UK GDPR questions, and the bank statement method finds the same tools with less friction.
An Honest Reality Check
A tech stack audit will save you money. It will not save you as much as the first spreadsheet suggests, and it is worth going in with that understanding so you do not feel cheated by the end.
Some shelfware will turn out to be a client's requirement or the one place a crucial archive lives. Some annual contracts will be too expensive to break and will have to wait for the renewal window, which may be eight months away. Some downgrades will be reversed within a month because a feature nobody thought they used turned out to be load bearing for the accounts team. In my experience a realistic first pass recovers somewhere between a fifth and a third of annual software spend, front loaded on seat clean-ups and downgrades, with the outright cancellations being the smaller share. That is still a lot of money for two weeks of work.
The bigger risk is that you do it once and never again. Tools creep back in. People join, add things, leave, and the seats stay. The audit only holds if you do two things afterwards: put every renewal date in a shared calendar with a reminder before the notice period closes, and make one person responsible for approving any new software before the card is used. A message in the team chat saying what the tool is, what it costs and what it replaces is enough. It just has to exist.
One category sits outside the four piles: consumption based billing such as cloud hosting and AI credits, where you pay for usage rather than seats. The bank statement finds the charges, but judging waste there needs someone who understands the usage, so handle those in a separate conversation.
Something You Can Do This Week
Do not try to do the whole audit in one sitting. Here is the version that fits into a normal working week.
Monday, export twelve months of transactions from every business account and card and filter for software. Tuesday, build the spreadsheet with the columns above, and send a two line message to everyone asking which tools they used last month and which they can log into but have not opened this year. Wednesday, sort every row into keep, downgrade, merge or cancel. Thursday, for every row in the cancel and downgrade piles, find the renewal date and notice period and put both in a shared calendar. Friday, export the data from the first three tools in the cancel pile, cancel them, and save the confirmations.
The rest of the cancel pile, the seat clean-ups and the merges take another week or two of small tasks, and the tier downgrades happen at the next billing cycle. By the end of the month you will have a live list of everything you pay for, a calendar that stops any of it renewing by accident, and a smaller bill. Then set a reminder for the same week next year, and it will take you an afternoon instead of a fortnight.