On this page
Learn to read any e-commerce dashboard and answer four questions: what the number means, where it comes from, what to suspect when it moves, and which decision it drives.
For operators new to e-commerce, inside an active eCom organization. Formulas match the platform's metric catalog; money is stored as integer VND.
Lesson 1: the pipeline#
Every report reads one pipeline from a different angle: viewer, click, order, delivered, payout, decision.
| Team | Main question | Watches | Typical decision |
|---|---|---|---|
| Commercial | Are sales good? | GMV, NMV, orders, AOV, channel mix | Which channel, product or price to push |
| Growth | Is the campaign worth the money? | ROAS, CAC, conversion rate, new customers | Increase budget, stop, change the offer |
| Creator | Which creators sell? | Creator GMV, ROI, cancel and return rate | Rebook, renegotiate, stop |
| Operations | Enough stock to sell and deliver? | Inventory, stockouts, days of supply | Restock, transfer, reduce push |
| Finance | Is the money arriving correctly? | Payout, fees, variance, margin | Reconcile, dispute, close the books |
| Data | Can this number be trusted? | Source, grain, sync status, mapping | Fix the pipeline, mapping or backfill |
Checkpoint: name the pipeline stage each of your own team's numbers is read at.
Lesson 2: core metrics#
| Term | Formula | Watch for |
|---|---|---|
| GMV | Value of goods ordered in the window | Demand only; says nothing about returns, fees, discounts or COGS |
| NMV | GMV minus cancelled orders, and minus returns when the organization chooses that definition | GMV growing without NMV: cancellations are rising |
| AOV | NMV / confirmed orders (not cancelled or returned) | NMV up, orders up, AOV down: more or smaller orders, or discounting |
| Orders | Count of orders | Always ask the grain: paid or all, order date or payment date |
| Units sold | Quantity sold | Not the order count; inventory decisions need units |
| Payout | Money the marketplace settles to the seller | Never compare today's payout with today's GMV; settlement lags |
| Variance | Actual payout minus expected payout | Negative is short-paid; often timing, a late return or a fee mismatch |
| ROAS | Attributed revenue / ad spend | Not profit; nothing subtracted for COGS, fees or returns |
| CAC | Acquisition spend / new customers | Compare with AOV, repeat rate and margin |
| Days of supply | Inventory / average units sold per day | Compare with restock lead time |
The organization picks the NMV definition once, under Settings > Organization; the NMV tile's info icon names the one in force. Under the default, a refunded or returned order still counts in NMV, so read returns separately before calling a month good.
Checkpoint: for a headline that grew, say whether it was GMV or NMV and what the gap between the two could hide.
Lesson 3: margin and payout#
Gross margin (NMV minus COGS) has not yet paid for ads or creators; contribution margin subtracts them and answers "is this campaign worth continuing". Settlement is the other leg: expected payout comes from orders, actual from marketplace statements; classify each variance line as timing, fee, return, mapping or a real dispute before closing the books (eCom reports).
Checkpoint: a campaign shows ROAS 4 and loses money; name the three costs ROAS never saw.
Lesson 4: creators and live commerce#
| Term | Meaning | Used for |
|---|---|---|
| KOL | Large-scale influence | Awareness, launches |
| KOC | Small consumer-style influence | Reviews, conversion |
| Affiliate | A tracked link paid by commission | Performance sales |
Read live commerce through GMV per viewer: many viewers with low GMV per viewer is attention without conversion. Creator ROI is (creator GMV minus creator cost) over creator cost; recheck it with NMV, since the top-GMV creator may be the most expensive or drive the most cancellations.
Checkpoint: for last week's best live session, state GMV per viewer and how much of its GMV was cancelled.
Lesson 5: stock and SKU#
A SKU is a product code; each marketplace can carry a different code for the same product, hence SKU mapping. A wrong mapping splits SKU revenue and misranks top SKUs. When GMV drops, check whether the hero SKU ran out first. Sell-through is units sold over units sold plus on hand; low means slow stock.
Checkpoint: list the hero SKUs and their days of supply against restock lead time.
Lesson 6: the data model#
| Layer | Holds | Read it when |
|---|---|---|
| Bronze | The original payload, sync-job records, rejected rows, the last synced checkpoint | A number is missing: is the connector running |
| Silver | Orders, order lines, products, creators, inventory snapshots, settlement lines | You need a specific root cause |
| Gold | Pre-aggregated by day, platform, SKU or creator | You need a trend, a top contributor, a card |
| Term | Meaning |
|---|---|
| Dimension | A column to group or filter by |
| Fact | An event table carrying measures (order, order item, payout line) |
| Grain | What one row represents |
Orders are one row per order, order items one row per product line: sum GMV after joining the two and it multiplies by the item count.
Checkpoint: for the table behind your most-used card, say what one row represents.
Lesson 7: the daily routine#
| Chart | Reads | Watch out |
|---|---|---|
| KPI card | Current value, period, change vs prior or target | Tells you something changed, never why |
| Line | Steady growth vs a one-day spike; a drop to zero | A sync failure looks identical to a real drop |
| Bar | Top contributor, who beat target | Add the "Other" bucket back when reconciling the total |
| Scatter | Two axes at once, such as CAC against GMV | High GMV plus low cost is the only quadrant to scale blindly |
| Heatmap | A colored matrix, such as cohort return rate | Does the color fade quickly; is a newer cohort doing better |
| Table | Where a decision gets made | Sort by the metric the decision depends on first |
| Symptom | Suspect | Action |
|---|---|---|
| GMV moves | Orders vs AOV, platform, SKU, a sale day, discounting, stockout, sync | Split NMV into confirmed orders x AOV, then by platform and SKU |
| GMV up, NMV flat | Rising cancellations, or a status or date mismatch | Break down by platform, SKU, creator, status; check settlement |
| High ROAS, losing money | ROAS uses GMV; COGS, fees or returns are high; attribution too broad | Read contribution margin: NMV minus COGS, creator and media cost |
| A creator looks good, finance disagrees | High cost, high returns, uncounted retainer, low-margin SKU | Compare creator GMV with NMV, cost, ROI, cancel rate |
| Orders cancel despite stock | Marketplace and warehouse stock out of sync, wrong SKU mapping | Check the latest snapshot, reserved quantity, mapping, sync job |
Daily check, fifteen minutes: set the range and filters, read the KPI cards and trend, break down by platform, SKU and creator, check margin, returns, stock and settlement exceptions, then write three actions.
Checkpoint: today's three actions, each with an owner and the condition under which you reverse it.