Ask a seller how business is going and you will hear order count and gross revenue — “crossed 600 orders, ₹4.2 lakh this month.” Ask what the bank actually received per order after deductions, how much of last month's money is still stuck unpaid, or how many hours of SLA headroom Tuesday's manifest had, and the room goes quiet. The first numbers are applause. The second set runs the business.
This post is the second set: twelve ecommerce seller KPIs, grouped into four clusters — money, returns, operations, concentration. For each one: why it matters, the formula, and a healthy band. The bands are illustrative, drawn from sellers we have sat with at 5–25 orders/day on AJIO and Meesho; your category and COD mix will shift them. The formulas are not negotiable — they come from your own statements.
The money cluster: ecommerce seller KPIs that touch the bank
1. Settled margin per order. Formula: (bank credit per order − product cost − packing cost). Not sticker price minus cost — bank credit, after commission, shipping, return fees, and the TDS/TCS lines. Why: it is the only margin that pays rent. Healthy band: positive and stable is the bar; many sellers discover whole SKUs living below zero once returns are allocated. Compute it monthly, per SKU, from the settlement statement — or let profit tracking do it per order.
2. Settlement variance. Formula: (expected payout − actual payout) ÷ expected payout, where “expected” comes from the marketplace's own fee structure applied to your orders. Why: this is your error-detection line; a persistent variance is either a fee you misunderstand or a mistake nobody has disputed. Healthy band: under ~1% explained; anything above 2–3% unexplained deserves a line-by-line look.
3. Claim recovery rate. Formula: claims credited ÷ claims filed (by value). Why: returns and disputes are only half the story — the other half is whether you got the recoverable money back. A low rate means weak evidence or missed windows; a near-100% rate usually means you are filing too few. Healthy band: meaningfully above half, by value, on claims worth filing.
4. Outstanding aging. Formula: value of orders delivered >30 days ago with no matching settlement credit. Why: money that is neither paid nor disputed is money quietly evaporating; settlement cycles are predictable, so anything outside them is an exception by definition. Healthy band: near zero past 30 days — every rupee there should have a name and a next step.
A note on cadence for the money cluster: settled margin and variance are monthly numbers — they need a full settlement cycle to mean anything. Recovery rate is quarterly; claims resolve slowly and a month's sample is too small to read. Outstanding aging is the exception: check it weekly, because the whole point is catching the order that fell out of the cycle before it fades into history. Matching each number to its natural rhythm keeps the dashboard honest — a daily glance at a monthly number teaches you to ignore it.
The returns cluster: one number is hiding two
5. Return rate, split. Formula: returns ÷ delivered-or-attempted orders — but always reported as two lines: RTO (never accepted) and customer returns (accepted, then sent back). Why: the two have different causes — address and COD behaviour versus listing expectation gaps — and therefore different fixes. A blended “21%” tells you to worry; the split tells you what to do.
6. Return cost per order. Formula: total return-linked costs (reverse fees, lost forward shipping, repacking, write-offs) ÷ all orders. Why: spreading return costs across every order shows you what returns do to your true margin — and what a one-point improvement is actually worth in rupees. Healthy band: category-dependent; the discipline is knowing yours and its direction.
7. Top-SKU return concentration. Formula: share of customer returns from your top three returning SKUs. Why: returns are rarely uniform — a handful of listings usually carry a large share of the pain, and that concentration is good news: it makes the fix a listing-content project, not a business-model crisis. Healthy band: if three SKUs explain more than a third of returns, you have a to-do list, not a mystery.
The operations cluster: the penalty-prevention numbers
8. SLA headroom. Formula: hours between when you actually completed a deadline-bound step (manifest close, handover) and the deadline — tracked as the weekly minimum, not the average. Why: averages hide the near-miss, and SLA penalties are undisputable once the timestamp lands. Healthy band: minimum headroom of several hours; a trend toward zero is a scheduled penalty.
9. Same-day dispatch rate. Formula: orders dispatched on order date ÷ orders received. Why: it is the cleanest single proxy for operational health — when it slips, everything downstream (SLA, ratings, RTO from delayed delivery) follows. Healthy band: above 90% consistently; below that, find the bottleneck before adding volume.
10. Exception rate. Formula: orders requiring manual intervention ÷ total orders. Why: exceptions are where your time actually goes, and an exception rate that grows with volume is the difference between a business that scales and one that merely gets bigger. Healthy band: single digits, falling over time as causes get fixed.
The operations trio also forms your early-warning chain, in order: exception rate rises first, dispatch rate slips second, SLA headroom collapses third — and the penalty arrives fourth. Sellers who track all three get two warnings before the first rupee is lost; sellers who track none meet the problem on the settlement statement.
The concentration cluster: fragility you cannot feel
11. Stock cover on top SKUs. Formula: units on hand ÷ average daily sales, per top SKU, in days. Why: stockouts on your best sellers are the most expensive quiet event in marketplace selling — visibility decays while you restock. Healthy band: 2–4 weeks on proven movers, tighter for seasonal items.
12. Top-5-SKU profit share. Formula: settled profit from top five SKUs ÷ total settled profit. Why: most small catalogs are far more concentrated than their owners believe, and concentration is fine only when it is known — it changes how much protection (stock, photos, claims attention) those five listings deserve. Healthy band: awareness; above ~70% means those SKUs are the business, so run them like it.
Where the data comes from — and the estimate trap
Every formula above has exactly four source documents behind it: the order list, the settlement statements, the returns records, and your own cost sheet. That is the good news — no exotic tooling is required to start. The discipline that decides whether the numbers mean anything is simpler and harder: use actuals, not estimates. The most common corruption we see in seller spreadsheets is a “commission ≈ 20%” cell typed in months ago, quietly standing in for the real per-order deductions ever since. The whole point of settlement variance (KPI 2) is to catch reality drifting from expectation — which it cannot do if your expectation was itself a guess.
Practical sourcing notes. Settled margin and variance come only from the settlement statement — never from the order panel, which shows you sticker prices. The return split needs returns logged on arrival, with a reason, which is a habit before it is a report. SLA headroom needs timestamps recorded at the moment of manifest close — reconstructing them later from memory produces fiction with decimal points. And the tax lines (the income-tax TDS under section 194-O and GST TCS that appear inside marketplace settlements) belong in your reclaim workflow, not in your margin — treat them as recoverable flows and track them separately, or your margin will look worse than it is.
If a number costs more than ten minutes a week to maintain by hand, that is not a sign to drop the number — it is a sign the underlying record-keeping should be automated. The twelve KPIs are cheap once the daily loop produces clean data as a by-product; they are expensive only when each one is a weekly archaeology project.
The 20-minute weekly review
Numbers that are not reviewed are decoration. The ritual that makes the twelve worth keeping is short and fixed: five minutes on money (margin trend, variance, anything aging), five on returns (the split, the worst SKU this week), five on operations (headroom minimum, dispatch rate, the exception list), and five to decide one action — fix a listing, dispute a variance, reorder stock, or explicitly nothing. Write the action down with an owner and a date. One action a week, fifty-two a year, compounds into a different business.
- Minutes 0–5, money: settled margin trend, settlement variance, anything sitting past 30 days unpaid.
- Minutes 5–10, returns: RTO vs customer-return split, top returning SKU, direction versus last month.
- Minutes 10–15, operations: minimum SLA headroom, same-day dispatch rate, open exceptions.
- Minutes 15–20, decide: one action, one owner, one date. Then close the dashboard.
Where Robnu fits
Eight of the twelve numbers come straight from data an OMS already handles: orders, documents, returns, and settlement statements. Robnu computes the operations and money clusters for AJIO and Meesho sellers as a by-product of running the daily loop — settled per-order profit, variance, the return split, headroom, exceptions — and puts them on the operations dashboard and the Understand views. The stock and concentration numbers still need your cost inputs; no software knows your product cost better than you do.
Robnu is free for everyone right now — every feature, every order, no card, no trial timer — and forever free under 25 orders/day once paid pricing launches. Start with the weekly review this Friday, even on a spreadsheet. The habit matters more than the tooling; the tooling just gives you your evenings back.