Insights·May 5, 2026·generic

Replacing your Google Sheet operations with custom software

When a Google Sheet stops being a spreadsheet and starts being your operations system, the right move isn't a bigger sheet — it's a small, focused app. Here's the framework.

The Google Sheet starts innocently. One tab to track orders. A second tab to track which ones got delivered. A third for vendor invoices, because the accountant asked for it. Two years later the sheet has nineteen tabs, four people who know which columns are load-bearing, and a quiet rule that nobody touches column M before noon. Last week somebody filtered without unfreezing the header and the order list silently lost half its rows for ninety minutes.

This is the spreadsheet ceiling, and almost every Indian operating business hits it. The reflex is wrong in two equal and opposite directions. One reflex is to keep stretching the sheet — more tabs, more formulas, more Apps Script — because the sheet has earned trust and changing it feels expensive. The other reflex is to throw the sheet away and buy a SaaS product, only to discover that the SaaS product makes thirty assumptions about your business and seven of them are wrong.

The right move is narrower and cheaper than both. Pick the single workflow inside the sheet that is doing the most damage, lift just that workflow into a small custom app, and leave the rest of the spreadsheet alone. This post is a field guide to doing that well.

The four signals that you have hit the ceiling

We've replaced more than a few of these sheets. The pattern is consistent.

Signal 1: the order of operations matters. A spreadsheet has no opinion about what gets filled first. The moment your team has a verbal rule like "always update status before quantity, otherwise the formula in column R breaks", the workflow has outgrown what a sheet is built for. Software encodes order; spreadsheets don't.

Signal 2: a wrong cell edit can break the day. If one person typing in the wrong column means somebody's order is missed or somebody's invoice goes out wrong, the cost of that error has crossed the threshold where validation, audit logs, and scoped permissions are cheaper than the next mistake.

Signal 3: the sheet is the integration. Data is being copy-pasted into the sheet from WhatsApp, from email, from a payment gateway, from a CSV. Then it is copy-pasted out again into invoices, into a bookkeeping product, into a courier dashboard. The sheet has quietly become the integration layer, which is the most expensive job in any operation and the one a sheet is least equipped to do.

Signal 4: only one person fully understands it. When the founder or the senior ops person goes on leave and the team stops being able to make changes confidently, the sheet has crossed from being a tool into being institutional knowledge. That is fine for documentation and bad for daily operations.

If two of these four signals are true, the workflow is past the ceiling. The European Spreadsheet Risks Interest Group has been cataloguing real-world spreadsheet failures for two decades — financial misstatements, mispriced contracts, missed legal deadlines — and the through-line is always the same: spreadsheets are perfect right up until they are catastrophic.

The trap of the platform middle ground

Before we recommend custom software, we always ask whether a no-code or low-code platform — Airtable, Retool, Glide, AppSheet, Softr — would do the job. Often the answer is yes, and the engagement ends there with a recommendation rather than a build. We have no incentive to oversell custom software for a problem that a 2,000-rupee-per-month SaaS solves.

Where these platforms stop fitting is predictable:

  • Multi-step business logic. A real approval flow with conditional routing, SLA reminders, and audit trail is the breaking point for most no-code tools. They can be hammered into doing it; the result is brittle.
  • Transactional integrations. When the workflow has to talk to Razorpay, Tally, Zoho Books, an SMS gateway, a courier API — and especially when those integrations have to roll back together when one of them fails — custom software becomes meaningfully cheaper to maintain than the no-code equivalent.
  • Custom pricing or business rules. Rate cards that depend on six inputs, discount logic that varies by customer segment, GST handling for mixed B2B/B2C lines — these are expressible in formulas but become unreadable in a spreadsheet and unmaintainable in a no-code platform's expression language.
  • Anything customer-facing. The moment the workflow has to be touched by a customer or a vendor — not just internal staff — the polish bar goes up sharply, and platform UIs start to feel like platform UIs.

If those four conditions are absent, use Airtable. If two or more are present, the cost curve has flipped and a small custom app is the right call.

The replacement framework

Once the decision is made, the trap to avoid is scope. A spreadsheet replacement project that tries to do everything the spreadsheet does is the most reliable way to spend INR 30L on a project that should have cost INR 8L.

The framework we walk teams through is sequenced.

Step 1: name the one workflow

Not the sheet. The workflow. In a typical operations sheet there are three to seven distinct workflows happening on different tabs — order intake, inventory adjustment, dispatch, billing, vendor management, reporting. Pick exactly one. The right one is almost always the one with the most daily errors or the most copy-paste between tabs.

Step 2: mirror the current process before improving it

The build's first job is to do exactly what the team already does, with the same field names and the same step order. Improvements come in the second release, after the team has stopped opening the sheet for this flow. Imposing a "better" workflow in version one is the most common reason these projects fail.

This is the same principle we apply on our internal tools service: the first build replaces the spreadsheet, the second build improves the operation.

Step 3: keep the data model boring

A small relational schema — three to six tables — is almost always the right answer. Even if the spreadsheet has thirty columns, most of them collapse into a handful of entities once you draw it. The spreadsheet's columns are the symptom; the entities are the disease.

Step 4: integrations come last

It is tempting to wire up Razorpay, WhatsApp, Tally, and a courier API in version one. Don't. Ship the workflow first with manual exports and imports, let the team use it for two weeks, and then add integrations in priority order. Half the integrations the team thought they needed turn out to be unnecessary once the data is structured properly. The other half become much easier to scope. We write about this sequencing in our automation service guide.

Step 5: leave the spreadsheet alive on purpose

The spreadsheet doesn't die. It keeps running for the workflows you didn't migrate, and often it keeps running as a read-only export from the new app for the few people who genuinely think in spreadsheets. Killing the sheet entirely is a vanity goal. Reducing the number of operations that depend on it is the real one.

A worked example

A specialty F&B operator we worked with — small chain, four outlets, central commissary — was running ordering between outlets and the kitchen on a shared Google Sheet. Each morning every outlet manager filled in their daily orders on a tab named after the outlet. The kitchen manager consolidated those tabs into a master production sheet. By 11 am someone had usually overwritten somebody else's row, and twice a month an order was missed entirely.

The replacement was deliberately small: one app, one workflow, four user roles (outlet manager, kitchen manager, owner, view-only accountant). Outlet managers entered their daily orders on a phone. The kitchen manager saw a consolidated production view that updated as orders came in. The owner got a daily summary email. Total build: six weeks, INR 6.5L. The case study is on our work page. The other twelve tabs of that spreadsheet — vendor lists, recipes, monthly P&L — were untouched and are still in use today.

That is the shape of a successful spreadsheet-to-software migration. Narrow, opinionated about scope, deliberately incomplete. The goal is not to build the operating system for the business. The goal is to retire the most dangerous workflow and leave everything else where it works.

What to do this week

If you suspect your operation has hit the spreadsheet ceiling, the cheapest first move is to spend twenty minutes counting. Open the sheet. Count distinct workflows. Count how many small errors each one has caused in the last fourteen days. Rank them. Pick the top one.

That single workflow is the project. The rest of the sheet can keep doing its job for another year, possibly forever.

When you are ready to scope a replacement, start a discovery conversation. The first call is usually enough to tell whether you need custom software, a no-code platform, or to leave the sheet alone for now. We're equally happy to recommend any of the three.

Frequently asked

When does a Google Sheet stop being the right tool for an operation?

The honest signal is not file size or row count — it is when the sheet has stopped being a place to record information and become the place where work happens. Once three or more people are editing it concurrently, a wrong cell edit can break the day, and the order of operations matters (this column has to be filled before that one), the sheet is doing software's job. At that point you are paying the cost of software — incidents, training, repair time — without the structure of software. The right move is to lift that one workflow into a small app and leave the rest of your sheets alone.

Do we have to replace the entire spreadsheet with software at once?

Almost never. The most expensive way to make this transition is to declare a 'no more spreadsheets' policy and try to migrate everything to a custom app at once. The cheapest way is to identify the single workflow inside the sheet that is causing the most damage — usually order intake, inventory updates, or status tracking — and build a focused app for just that. The rest of the sheet keeps doing its job. Most operations end up with a small app for the high-stakes flow and a few sheets quietly handling everything else, which is the correct equilibrium.

What about Airtable, Retool, Glide, or AppSheet?

These are good tools and we recommend them when the fit is right. Airtable is excellent for structured records you mostly read and lightly edit. Retool is excellent for back-office UIs over a database you already have. Glide and AppSheet turn a sheet into a basic mobile or web form quickly. The point at which these stop being the right answer is when the workflow needs business logic the platform cannot express cleanly — multi-step approvals, transactional integrations, custom pricing, or data that must reconcile against an external system. At that point a small custom app becomes cheaper than fighting the platform, and meaningfully more reliable.

How much should the first version cost?

For a focused replacement of one workflow inside a spreadsheet, expect INR 4L to INR 10L for a real first version, four to eight weeks of build time, and a small team — one designer, one engineer, one product owner shaping scope. Anything more expensive than that for a single workflow is usually scope drift; anything cheaper is usually a prototype rather than something the team can run on. The maintenance budget afterwards lands around INR 15K to INR 30K a month for a small internal tool, which is most often less than the cost of the time the spreadsheet is currently eating.

What gets lost when we move off the spreadsheet?

Two real things: instant ad-hoc edits and the visual freedom to put anything anywhere. Custom software trades those for structure, which is the entire point. The thing that does not have to be lost — and is often lost by accident — is the team's mental model of how the work flows. The most common cause of failed spreadsheet-to-software migrations is software that imposes a workflow nobody asked for. The build should mirror the existing process first, then improve it once the team trusts it.

How do we know the migration actually worked?

Three concrete signals, in order. First: people stop opening the sheet for that workflow within a month. Second: the small daily errors that used to require a Slack message to fix — wrong status, missing field, duplicated row — drop to near zero. Third: someone new on the team can do the workflow on day one without being shown the secret column or the magic formula. If those three things have not happened in six to eight weeks of real use, the build did not solve the problem and the team should be loud about it.

Ready to talk about your build?

Book a discovery call See selected work