Data-driven in 90 days on a small analytics budget

Most Indian SMBs are not short of data. They are short of a decision about which five numbers matter, which is why the GA4 property installed two years ago gets opened once a quarter and closed in ninety seconds. The cost lands hardest on founders, finance and ops leads at 30-200 person companies, where nobody is a full-time analyst and the person "handling data" is also closing the books or chasing deliveries.

What should be true by day thirty?

By day thirty you should have one screen with five to seven numbers, and a named person who looks at it every Monday morning.

Work backwards from decisions, not tools. Write down the three decisions you made last month that you wished you had evidence for — which lead source to fund, whether to hire a second delivery person, whether a discount is actually closing deals. Turn each one into a number. Then check whether the events behind those numbers are even being captured; in our experience most small B2B sites have analytics installed but no working conversion events beyond pageviews. A free site check will show you whether tags and basic tracking are firing before you plan anything on top. Instrument those five numbers only, and deliberately ignore the other forty the tools offer you.

Which tools are worth paying for?

Pay for the warehouse and the dashboard layer, and keep everything else free or cheap until you cross roughly fifty people.

For most SMBs a single Postgres database plus a BI tool covers about 80% of what a six-figure annual analytics stack does. Typical monthly costs we see:

• Behaviour: GA4 and Search Console at zero, or Plausible or Matomo at roughly ₹1,000-2,500 a month if you want simple, cookie-light reporting
• A place to join data: managed Postgres at about ₹1,500-3,000 a month
• Dashboards: Metabase open source on your own server, or its cloud plan at roughly $85 a month
• Revenue truth: UPI settlements, bank statements and your invoice register reconciled weekly in one table

If you would rather have the pipeline wired once than maintain it yourself, a fixed-scope dashboard build is usually cheaper than six months of part-time tinkering by whoever is least busy.

What does a 90-day budget actually look like?

A realistic first-90-days spend for a 40-60 person company is ₹60,000-1,20,000 all in, including a one-time build.

Say ₹18,000-24,000 for tools across three months, a one-time build of 40-60 hours, and three to four hours a week from an internal owner — about 40 hours of someone's quarter, which is the part people forget to budget. Compare that to an analyst hire at ₹8-12 lakh a year and the arithmetic is not close. Scope creep is what breaks the number: a build that starts as "five questions" quietly turns into a customer 360 project. Write the five questions down before anyone writes SQL, and check what a fixed-scope engagement looks like on our pricing page.

Why do these projects stall around week six?

They stall because nobody owns the numbers, not because the tools fail.

Three things go wrong in predictable order. Definitions drift, so "qualified lead" means one thing to sales and another to marketing, and the dashboard quietly stops matching reality. Pipelines break silently when a form field is renamed or a webhook expires, and nobody notices for a month. And the Monday review gets cancelled twice, then never comes back. The fixes are cheap: a one-page definitions document, one owner with thirty minutes a week, and a freshness alert that emails you when a table stops updating. While you are at it, keep personal data out of dashboards where you can — the DPDP Act expects you to know where customer data sits, so aggregate at source and keep consent records in their own place rather than in a CSV that gets forwarded around.

What changes between day sixty and ninety?

Between day sixty and ninety the work shifts from reporting to running one experiment at a time.

By now you should know your funnel baseline: of 100 enquiries, how many get a quote, how many convert, how long each stage takes, and what an average won deal is worth. Pick the leakiest stage and test one change for two weeks. If 100 enquiries produce 40 quotes and 7 wins, the quote-to-win step is your leak, and moving first follow-up from day four to day two is a testable change — you will know within about six weeks whether it moved 7 to 9. Retire any metric nobody has opened in a month. A dashboard with five live numbers that change behaviour beats a beautiful one nobody trusts.

Where to start

Pick your five questions this week, name one owner, and put a standing thirty-minute review in the calendar before you buy anything. If you want the plumbing built once, documented, and handed over without platform lock-in, Maysan Labs does exactly that kind of small fixed-scope work — but the five questions are yours to write, and they are the hard part.