Shopify Analytics: The 15 KPIs Every Store Owner Must Track

A store doing $24,000 a month at 65% gross margin looks like it clears $15,600 of margin. Subtract discounts, the shipping you absorbed, payment fees, packaging, pick and pack, and a returns provision, and the real figure is closer to $8,000. Subtract acquisition and the owner is running a business that only works because repeat customers carry it.

None of that shows up on the Shopify home screen. Here are the fifteen numbers that do the work, which six of them actually change decisions, and the specific way each one lies to you.

Start with the metric almost no list includes

1. Contribution margin per order. What is left after every cost that scales with the order. Not gross margin. Gross margin is the number that flatters you.

Work a $60 order at a store with a 35% cost of goods:

• Gross revenue, $60.00
• Average discount across all orders at 8%, minus $4.80
• Cost of goods, minus $21.00
• Shipping you paid because you offer free shipping, minus $9.20
• Payment processing at 2.9% plus 30 cents on the $55.20 collected, minus $1.90
• Packaging and pick and pack, minus $2.20
• Returns provision, 8% of orders costing about $12 each in net loss, minus $0.96

Total cost that scales with the order: $40.06. Contribution margin: $19.94, or 33% of gross revenue.

Gross margin said 65%. The number you can actually spend is half that. Now put acquisition next to it. If blended customer acquisition cost is $22, the first order from a new customer loses $2.06. At 400 orders a month with 60% from new customers, that is 240 orders losing $494 in total, and 160 repeat orders contributing $3,190. The whole business is the repeat orders. That single calculation reorders most stores' priorities, and it takes an hour in a spreadsheet.

How it lies: if you compute it on a single order rather than a blended average, you will pick a flattering one. Use the last 90 days of every order, including the discounted ones and the returned ones.

The decision it drives: what you can pay for a customer, and whether free shipping survives.

The five other metrics that change what you do

2. LTV to CAC, and the payback period underneath it. The ratio everyone quotes is 3:1 or better. The ratio is also the easiest number in commerce to fake.

Two ways it goes wrong. First, most stores compute lifetime value on revenue rather than contribution margin. Our customer places 1.6 orders in year one, so revenue LTV is $96 against a $22 CAC, a 4.4:1 ratio that looks excellent. On contribution margin it is 1.6 times $19.94, so $31.90 against $22, a 1.45:1 ratio. Same customer, same store, one number says invest and the other says be careful.

Second, lifetime value on a cohort that has not aged is a forecast, not a measurement. A cohort acquired in March cannot have a twelve-month value in May. What gets reported as LTV is usually a projection with an optimistic curve fitted to it.

The metric that actually governs how fast you can spend is CAC payback period: how many months until a cohort's cumulative contribution margin repays what you paid to acquire it. Our cohort recovers $19.94 on the first order and clears $22 when the second order lands, which for a median second-order gap of 74 days is about 2.5 months. Now the cash consequence: spending $10,000 a month on acquisition with a 2.5 month payback means carrying roughly $25,000 of unrecovered acquisition cost at steady state. Stretch payback to six months and you carry $60,000. That is the number that decides whether growth is fundable, and no dashboard shows it by default.

3. Repeat purchase rate. The share of customers who order again. Define it as a share of a cohort at a fixed age, such as 90 days after the first order.

How it lies: computed over all time, it rises automatically as your store gets older, because old customers have had more chances to come back. An all-time repeat rate that climbs every quarter tells you nothing except that time passed. Fix the age and compare cohorts.

The decision it drives: acquisition versus retention. Move the 90-day repeat rate from 18% to 24% on 240 new customers a month and you add 15 orders a month, about $300 in contribution, and it compounds because every future cohort inherits it. Compare that to what 15 more orders would cost you to buy at $22 each.

4. Conversion rate by source. The blended rate is close to useless because it moves when your traffic mix moves, even when your site is identical.

Take 20,000 sessions split as 2,000 email at 4%, 10,000 organic at 2%, 4,000 paid social at 0.6%, and 4,000 direct at 1.5%. That is 364 orders, a 1.82% blended rate. Now double paid social to 8,000 and take organic down to 6,000, changing nothing else. Email still converts at 4%, organic at 2%, paid social at 0.6%. Orders fall to 308 and the blended rate reads 1.54%, a 15% drop. Every source performed identically. Someone will spend a month redesigning the product page.

The decision it drives: where the next dollar of traffic spend goes, and which channel's landing experience is genuinely broken rather than just cheap.

5. Average order value. Revenue divided by orders. Simple, and worth real money: at 400 orders a month, four extra dollars of AOV is $1,600 in revenue, and if it arrives as an added unit at 35% cost of goods it carries $2.60 of contribution each, so $1,040 a month of actual margin.

How it lies: it is a mean, so it follows outliers. One $900 order in a month of 400 moves AOV by more than two dollars on its own. Track the median next to it. If the two diverge, your AOV is being set by a handful of buyers you cannot reproduce.

The decision it drives: the free shipping threshold, bundle design, and whether a quantity break is worth building.

6. Cash conversion. How many days your money is tied up: days of inventory on hand, plus days until payouts land, minus the days of credit your suppliers give you.

A store holding 90 days of inventory, paid out in 2 days by its processor, on 30-day supplier terms, runs a 62-day cycle. At $24,000 of monthly revenue and 35% cost of goods, that is $8,400 a month of goods, so the cycle ties up about $17,360 of working capital at all times. Double revenue and you need to find another $17,000 before you sell a single extra unit. This is how profitable stores run out of money, and it is invisible on every revenue chart ever drawn.

The nine that support them

Keep these, check them less often, and do not build a meeting around any of them.

• 7. Revenue. The scoreboard. It tells you nothing about why, and is only useful next to a comparison period.
• 8. Gross margin. Revenue minus cost of goods. Useful for pricing and product mix, misleading as a health check, per the first section.
• 9. Customer acquisition cost. Acquisition spend divided by new customers, including agency fees, creative costs, and platform fees. Excluding those is the most common way this metric flatters a store.
• 10. Revenue per session. Conversion rate and AOV in one number. Good for comparing channels, bad for diagnosing anything.
• 11. Sessions. Volume. It matters as a denominator and as a check that a traffic change explains a revenue change.
• 12. Traffic by source. Necessary to make number 4 possible. Watch mix shifts, not absolute counts.
• 13. Cart abandonment rate. Abandoned carts over carts created. High is normal. Only act when it moves.
• 14. Checkout abandonment rate. The one to actually watch, because the causes are mechanical: a surprise shipping cost, a failing payment method, a required field. Baymard's checkout research puts documented abandonment at roughly 70% across the industry, so treat your own rate as a trend line, not a verdict.
• 15. Sell-through rate. Units sold over units received, per SKU, per season. It drives reorder decisions and warns earliest that cash is about to be trapped in stock.

Take bounce rate off the list

Bounce rate in GA4 is not what it was in Universal Analytics. It is defined as the inverse of engagement rate, and a session counts as engaged if it lasts longer than ten seconds, or fires a conversion event, or has two or more page views. That means bounce rate is a threshold artefact. A visitor who reads your product page for nine seconds and leaves, and one who bounces instantly, are the same number. A change in your consent banner or your page speed can move it without a single visitor behaving differently.

Use something with a decision attached instead: product page view rate per session, scroll depth on your top landing pages, or add-to-cart rate by landing page. Those tell you what to change.

A metric you never act on is a cost

Every number on a daily dashboard costs attention, and attention is the scarcest thing in a small commerce team. The test is simple: can you name the decision this metric would change, and the threshold that would trigger it? If not, it does not belong on your daily view.

A cadence that survives contact with a real week:

• Daily, three or four numbers only. Ad spend, orders, site availability, payment failures. Things where a one-day move needs a same-day decision.
• Weekly. Conversion rate by source, AOV and median order value, contribution margin per order, checkout abandonment. Enough sample to be real, short enough to steer.
• Monthly. CAC payback period, repeat rate by cohort at a fixed age, cash conversion cycle, sell-through by SKU. These move slowly, and checking them weekly produces noise you will act on.

Anything you look at more often than you could act on it, move one cadence slower.

Where Synton fits

Synton's Analytics app opens on a question rather than a dashboard. You pick or type a business question, read the inline answer with the key number and a one-line takeaway, then open the full view behind it. Being precise about that question bar, because it is easy to overpromise: it matches what you type against a built-in list of questions. It does not send your question to a model and it will not invent an answer. If nothing matches, it says so.

The rail carries eleven views: Command Cockpit for the headline operating numbers, Pulse, Anomalies, Revenue, Traffic, Attribution Lab, Funnel, Cohort Retention, Customer Insights, and Forecast Room. Command Cockpit writes a plain English line under every acronym, so CAC payback and MER are explained where they appear. More views exist than the rail shows, including profit studio, unit economics, and demand planning, reachable through Cmd+K.

One habit worth knowing: a dash means the number could not be worked out, never that it is zero. Those need opposite responses, because a dash is a connection problem and a zero is a result.

The Today app is the daily layer: a ranked list of exceptions, each with a Why column and, where one exists, a comparison against the median for your vertical. You can tune each anomaly detector's sensitivity, or pause a noisy one.

Three honest limits. Scheduled report delivery is not available yet and says so if you try. The paid-channel table in Analytics, Revenue estimates Meta, Google and TikTok revenue at a flat figure per conversion rather than reading real revenue, so treat that table as a rough shape. And TikTok return on ad spend reads as zero throughout, because conversion value is not requested when campaigns sync.

There is no agent called an "Analytics Agent". Synton ships 18 autonomous agents, and the ones touching this territory are Churn Prediction, Cart Recovery, CRO Optimizer, and Performance. Enabling one does not give it a clock: first enable puts it in suggest mode, where it drafts and waits for you, and scheduling is a separate switch.

What to do Monday morning

1. Build the contribution margin calculation from the top of this article using your own last 90 days. One afternoon. Nothing else here matters as much.
2. Split conversion rate by source and check whether last quarter's decline was a mix shift rather than a site problem.
3. Compute CAC payback and the working capital it implies at your current spend. That is your real growth ceiling.
4. Pick your four daily numbers and remove everything else from the daily view. Move the rest to weekly or monthly.
5. Add median order value next to AOV and see whether they agree.

Creating a Synton account is free, needs an email and a password and no card, and you can connect a store and read your dashboard and store reports before spending anything. AI work runs on a paid plan.