Nobody in your business decided to spend what you spend on software. It accumulated.
A tool bought for a project that finished two years ago. A tier upgraded for one feature somebody needed once. Seats for four people who have left. Two departments paying for different products that do the same job, neither aware of the other. A subscription on a personal card, reimbursed monthly, that nobody has questioned since.
None of that is carelessness. It is the predictable result of purchases being made individually, renewals happening automatically, and nobody owning the total.
The afternoon audit
You can find most of it yourself, and it is worth doing before anyone sells you anything.
1. List everything. The fastest route is not your memory — it is the card statement. Pull twelve months and mark every recurring software charge. Then check for annual renewals that fell outside the window, and ask people what they pay for personally and expense.
2. Get the annual figure for each. Monthly prices disguise scale. A tool at ₹2,000 a month is ₹24,000 a year, and seeing the year is what makes the question feel worth asking.
3. Open each admin console and compare seats assigned against seats active. This is where the real finding usually is. Most business tools show a last-login date. Anyone who has not logged in for ninety days is a seat you are buying and not using.
4. Note the renewal date and the notice period for each. You cannot act on anything you discover the week after it auto-renewed for another year, and vendors know this.
5. Circle the overlaps. Two products that both do file sharing, or project management, or e-signatures.
That is a genuine afternoon of work and it typically pays for itself several times over before you have considered replacing anything at all. The unused seats and the dead subscriptions are the easy money, and they require no migration, no retraining and no risk.
Then the harder question
For the tools that survive the audit — genuinely used, genuinely costly — there are three remaining options, and they are not equally good.
Renegotiate. Almost nobody does. Vendors discount to retain customers, particularly against an annual commitment or a competitor quote. The ask costs one email.
Switch to something cheaper, including a self-hosted open-source equivalent. The licence cost goes to zero. An operational cost appears in its place: updates, backup, monitoring, security, and someone accountable when it breaks. That is a real trade, not a free win, and whether it is worth making depends mostly on your seat count and how much of the product you actually use.
Build something custom. Usually the wrong answer. Here is why.
The arithmetic people get wrong
The comparison is almost always made like this: the subscription costs ₹6 lakh a year, a custom build is ₹8 lakh, so we break even in sixteen months.
That calculation is missing the entire right-hand side. Custom software has ongoing costs that a subscription silently includes:
- Maintenance — dependencies, browser changes, OS updates, the small breakages of normal life
- Security patching, indefinitely, because you now own that responsibility
- Feature work as the business changes, which the vendor would otherwise have done
- Hosting, backup and monitoring
- Someone who understands it — the cost that appears when that person is unavailable
Include those over five years and a lot of apparently obvious builds stop being obvious. The vendor is not only selling you software; they are selling you the amortised cost of maintaining it across thousands of customers. That is a genuinely good deal for anything you use in a fairly standard way.
When building actually wins
It does happen, and the conditions are specific. All of these, not some:
- You use a fraction of the product. You pay for an enterprise suite and use one module.
- Your use is narrow and stable. The requirements are not going to move much.
- There is no statutory churn underneath it. Nothing that changes every budget cycle.
- The scope is genuinely small. Weeks, not quarters. If the estimate is a year, the build will not win.
- Per-seat pricing has become the dominant cost and your headcount is growing.
We did exactly this for our own invoicing. We deployed the standard open-source platform first — five containers for a handful of export invoices a year — measured it honestly against what we actually needed, and replaced it with one container in a day. The write-up is here, including the feature we built and then deleted.
Note the condition that made it work: the scope was tiny. A business issuing hundreds of invoices a month with recurring billing and payment collection should have kept the platform, and we would have said so.
What to never rebuild
Worth stating plainly, because someone will eventually suggest it:
Accounting, payroll, tax filing and compliance reporting. The rules change every budget cycle and a vendor absorbing that churn is worth every rupee. Businesses that rebuilt these have regretted it, usually within two years and always at the worst time of the financial year.
Payments. Use a payment provider. The compliance surface is not yours to want.
Email infrastructure. Deliverability is a specialist discipline and running your own mail server in the year 2026 is a decision you will revisit.
The one habit worth keeping
Put a recurring reminder in the calendar, once a year, in the same week each time: review every software subscription.
Ten minutes per tool. Is it still used? By how many people? When does it renew? Could we drop a tier?
Most of what this note describes exists because that review has never happened, not because anyone made a bad decision. An hour a year, owned by one person, prevents nearly all of it.
We do this as a standalone review — see software cost review. It regularly concludes that a tool is worth what you pay for it, which is the finding that makes the others worth trusting.