Finance operations
A finance team needed daily sales and returns across several marketplaces and was waiting minutes per query, then asking an engineer for extracts anyway. The fix was in the queries and the data model, not the hardware.
2–3 min → 2–5 sec
query time, on the same hardware
8.5 million
rows produced by a single join, returned in seconds
19 million
rows in the largest table involved
5 reports
across generic, Myntra-Omni and TataCliq variants
Swipe to see the full screen
The finance team’s daily reports were technically available and practically unusable. A query took two to three minutes, exports crashed the browser, and the workaround — asking an engineer for a CSV — put a person in the middle of a daily routine.
The underlying data was large and awkward: a time-partitioned detail table running to over a thousand daily chunks, a join that fanned 2.4 million rows out to 8.5 million, and a 19-million-row ledger behind it.
A self-serve reporting application over the existing warehouse, with the query layer rebuilt around how the data is actually shaped.
Daybook
Sales & returns, daily
Returns
36,466
▼ 1.8%
Return value
₹2.41 Cr
▼ 2.2%
Return rate
7.8%
▼ 0.4%
QC failed
4,118
▲ 0.9%
Reason for return
QC outcome
Swipe to see the full screen
The choice, why it was made, and what it means for you. Skip the middle part if you like — the point still lands.
01
The choice
The work went into how the data is joined and grouped rather than into a bigger machine.
Why
A slow report is almost always a query that asks for more than the answer needs — a count nobody reads, a join evaluated before a filter, a sort over rows that will be discarded.
What it means for you
Two to three minutes became two to five seconds on the same hardware, which means no infrastructure spend and no migration.
02
The choice
Pagination reports whether more rows exist rather than computing a total count.
Why
On a 19-million-row table an exact total costs more than the page being displayed, and nobody reads the total.
What it means for you
The first page arrives immediately regardless of how much data sits behind it.
03
The choice
The report SQL the finance team supplied was deliberately preserved rather than rewritten, with the optimisation applied around it.
Why
Their formulas encode agreed definitions. Rewriting them to be faster would have made the output faster and wrong, and the disagreement would have surfaced weeks later in a reconciliation.
What it means for you
You get the speed without having to re-audit the numbers.
If your existing reports time out, your exports crash the browser, or a page takes minutes, that is usually fixable without replacing anything. This applies to any reporting layer sitting on tables that have outgrown the queries written for them.
Screenshots show the real interface with sample figures. Timings are measured on the client’s own hardware, before and after.
Twenty minutes on a call is usually enough for me to tell you whether software is the answer and roughly what it would cost. If it isn’t, I’ll say so.
Available for freelance builds alongside senior engineering work. One or two projects at a time.