At $1M to $3M, four ecommerce KPIs are enough: contribution margin per order, blended acquisition cost, 60-day repeat rate, and cash conversion cycle. At $3M to $10M, add marketing efficiency ratio, CAC payback period, contribution margin by SKU, cohort LTV at twelve months, and subscription retention by month. Past $10M, add channel incrementality, cohort payback by acquisition source, inventory turns against cash, contribution margin by customer segment, and forecast accuracy.
Formulas and benchmarks for most of those live in the nine ecommerce metrics that predict scale, which is the dictionary. This post covers which entries belong on your desk this quarter, how often you look, and who gets held to each one.
I’ve reviewed the numbers behind brands from $1M to $60M+, including Ancestral Supplements from $28M to $60M+ and Dr. Gabrielle Lyon’s $6M practice. The reporting failure I run into most often has nothing to do with missing data. Founders have too much of it and no ranking, so the number that gets acted on is whichever one moved most last week.
Why forty KPIs works out to zero
A founder tracking forty KPIs is not tracking forty KPIs. Attention is the scarce input, and a dashboard with forty tiles spends it on whatever is red today, which is usually a 9% dip in a metric with a 12% weekly standard deviation.
The test: a KPI earns a slot only if you can name the decision a move in it triggers and the person who makes that decision.
Sessions fail. No one has changed a media budget because sessions moved 6%. Contribution margin per order passes: a $4 drop tells you to reprice or renegotiate freight, and someone owns that call.
$1M to $3M: survival economics
At this size the business is deciding whether it can fund its own customer acquisition. Four KPIs, and I’d argue with you about adding a fifth.
- Contribution margin per order. Dollars left after every cost that varies with the order.
- Blended acquisition cost. All marketing spend divided by new customers, agency fees and creative included.
- 60-day repeat rate. By first-purchase cohort, not aggregate.
- Cash conversion cycle. Days between paying your supplier and getting paid by customers.
The number that outranks the rest: contribution margin per order. At $2M in revenue with $16 of contribution per order, you have roughly $50K a month to run everything on, and that ceiling decides your ad budget, your first hire, and how much inventory you can commit to.
What to deliberately ignore until $3M:
- Channel-level ROAS. Order volume is too low for channel splits to mean anything, and platform attribution overstates every channel anyway.
- Attribution modeling. A multi-touch tool at $2M buys a more expensive version of a guess.
- LTV projections. A model built on eight weeks of history is an assumption wearing a decimal point. Use 60-day repeat rate until you have twelve months of cohorts.
Brands stall here more often from a repeat rate problem than an acquisition problem. That diagnosis got its own piece: why repeat rate is falling while acquisition looks fine.
$3M to $10M: efficiency and mix
Spend is big enough that where it goes matters, and the catalog is wide enough that averages hide losers. Add five:
- Marketing efficiency ratio. Total revenue divided by total marketing spend, reviewed weekly.
- CAC payback period. Months until an acquired customer has returned what they cost.
- Contribution margin by SKU. Quarterly, with shipping weight and return rate loaded in.
- Cohort LTV at twelve months. One fixed horizon, held constant so cohorts are comparable.
- Subscription retention by month. Months 1, 3, 6, and 12, tracked apart from each other. Benchmarks for supplements are in subscription retention benchmarks.
The number that outranks the rest: CAC payback period, which sets how fast you’re allowed to grow. Two months of payback means you can double spend on operating cash. Nine months means a financing conversation, and brands that skip the check look profitable right up until the inventory PO clears.
This is also the stage where reporting has to stop being a file. When one person rebuilds the numbers in a spreadsheet on the first Tuesday of every month, the definitions drift and the analysis lands three weeks late. What you need is a fixed set of queries or exports with written definitions, refreshed on a schedule, producing the same page every week.
$10M to $25M+: allocation and forecasting
The job changes here. Unit economics are settled, and now you’re deciding where the next $500K goes while committing inventory against a number you predicted a quarter ago. Add:
- Channel incrementality. What sales actually stop when a channel is paused, measured with holdouts or geo tests.
- Cohort payback by acquisition source. Same math, split by where the customer came from. Meta and affiliate cohorts diverge by month three, often by a factor of two.
- Inventory turns against cash. Turns read next to the cash conversion cycle, which turns an ops metric into a growth constraint.
- Contribution margin by customer segment. Subscribers, one-time buyers, and bundle buyers priced separately.
- Forecast accuracy. Formula: 1 − (|forecast − actual| ÷ actual), run monthly on revenue and on units. Inside 10% at the month level is the bar I hold teams to. Miss it two months running and every downstream plan is fiction.
The organizational shift: each KPI gets a named owner or it belongs to nobody, and owners are individuals, since a KPI owned by a team is owned by no one. The head of growth takes MER and blended CAC, the retention lead takes month 3 and month 6 retention, ops takes turns and the cash cycle, finance takes forecast accuracy. When a number moves, one person explains it and one person decides what changes. Brands that skip this end up with a dozen beautiful KPIs and a meeting where everyone agrees the trend is concerning.
The stage table
| Stage | KPI | Cadence | Owner | Decision it triggers |
|---|---|---|---|---|
| $1M-$3M | Contribution margin per order | Weekly | Founder | Repricing, discounts, freight |
| $1M-$3M | Blended acquisition cost | Weekly | Founder | Ad budget up or down |
| $1M-$3M | 60-day repeat rate | Monthly | Founder | Product and post-purchase flow |
| $1M-$3M | Cash conversion cycle | Monthly | Founder | Order size and supplier terms |
| $3M-$10M | Marketing efficiency ratio | Weekly | Head of growth | Budget shift between channels |
| $3M-$10M | CAC payback period | Monthly | Head of growth | How fast spend can scale |
| $3M-$10M | Cohort LTV at 12 months | Monthly | Head of growth | CAC ceiling for next quarter |
| $3M-$10M | Subscription retention by month | Monthly | Retention lead | Onboarding and win-back work |
| $3M-$10M | Contribution margin by SKU | Quarterly | Founder or ops | Kill, reprice, or promote a SKU |
| $10M+ | Channel incrementality | Quarterly | Head of growth | Reallocation across channels |
| $10M+ | Cohort payback by source | Monthly | Head of growth | Per-channel spend caps |
| $10M+ | Inventory turns vs cash | Monthly | Ops | PO timing and size |
| $10M+ | CM by customer segment | Quarterly | Finance | Offer and pricing structure |
| $10M+ | Forecast accuracy | Monthly | Finance | Whether the plan gets rebuilt |
Read the last column first. Any row you can’t fill in is a row to delete, and most brands have four or five.
Designing the reporting cadence
Weekly metrics review, 30 minutes. Four to six numbers, each shown next to last week and a four-week trend line. The owner speaks to their number for two minutes: what moved, whether it’s signal, what they’re doing about it. Anything needing analysis becomes an assignment with a name and a date. Decisions get written down. Past 30 minutes it has become a status update and should be rebuilt.
Monthly business review, 90 minutes. Cohort tables, CAC payback, LTV by cohort, retention curves, contribution margin. You’re checking whether cohorts improve month over month, the question that predicts next year.
Quarterly reallocation, half a day. SKU margin, channel incrementality, segment margin, and forecast accuracy against the last plan. Money and headcount move here and nowhere else, which keeps the weekly meeting from turning into a budget negotiation.
A good weekly review is boring. Most weeks nothing moves enough to act on, and the discipline is in saying so and leaving.
Vanity KPIs that survive too long
- Sessions. Replace with new customers acquired and blended acquisition cost.
- Platform ROAS. Replace with weekly MER and monthly blended CAC.
- Gross margin. Replace with contribution margin per order. Gross margin flatters supplement brands, since COGS is the small part.
- AOV in isolation. Replace with AOV paired to 60-day repeat rate. Bundles that lift AOV while pushing out the reorder date read as a win for two months.
- Email open rate. Replace with revenue per recipient.
- Follower count. Replace with new customers who name your content in a post-purchase survey.
How to build the reporting without a data team
Three moves, in this order.
Get the cost inputs into one place. Landed COGS by SKU, freight, fulfillment per order, payment processing, and total marketing spend including fees and creative. A spreadsheet and a week of annoying emails to your 3PL and bookkeeper. Most brands find their contribution margin was wrong by $3 an order.
Fix the definitions in writing. One page: what counts as a new customer, whether returns are netted, whether influencer product cost counts as marketing, which date field a cohort is keyed to. Everyone who touches the numbers signs off. Skip it and two people quote CAC in the same meeting 30% apart.
Then automate. Scheduled exports, a warehouse, or a tool like Triple Whale or Northbeam. A brand that automates a wrong definition gets wrong numbers faster and trusts them more. The same ordering applies to what to automate first.
When more measurement is the wrong move
If your 60-day repeat rate is under 15% or your contribution margin per order is under $8, stop building the dashboard. Better instrumentation will tell you the same thing with more decimal places. Go fix the offer: the product, the price, the size of the first order, or the expectation the ad sets before the box arrives.
The other case is volume. Under roughly 400 new customers a month, cohort analysis is mostly noise. A five-point move in a 180-customer cohort is nine people, which could be one influencer post or one bad shipping week. Track the four survival numbers, keep the raw data clean, and spend the analysis hours talking to customers.
Common questions
How many KPIs should an ecommerce brand track? Four to six under $3M, eight to ten between $3M and $10M, twelve to fifteen past $10M. The limit is how many numbers a leadership team can hold at once, which in my experience is about six on a weekly cadence.
What’s the single most important ecommerce KPI? It changes by stage. Under $3M it’s contribution margin per order. From $3M to $10M it’s CAC payback period. Past $10M it’s forecast accuracy, since a wrong plan costs more than a mediocre month.
How often should I review metrics? Weekly for MER, blended acquisition cost, and contribution margin per order. Monthly for cohorts, payback, and subscription retention. Quarterly for SKU margin, segment margin, and channel incrementality. Below a few hundred orders a day, daily review is noise-chasing.
Do I need a BI tool? Not under $5M. Spreadsheets and platform exports produce every number on this list, and building them yourself teaches you which ones you use. Past $5M a tool pays for itself in analyst hours, once the definitions are written down.
Why doesn’t my dashboard match my P&L? Usually one of four things: the dashboard counts gross revenue while the P&L nets returns and discounts, marketing spend excludes agency and creative fees, COGS is standard rather than landed cost, or the two use different month boundaries. Reconcile once, write down the answer, and the question stops coming back.
If you’re running a $3M to $15M brand and the monthly numbers arrive too late to act on, or two people in the same meeting quote different CACs, the problem is the reporting system. Growth OS is the version of this I install under every engagement, so the founder and I argue from one set of definitions from week one.