Marketing KPIs: how to pick the few numbers that run your funnel

Marketing KPIs are the few numbers a marketing team commits to moving, each with a target, an owner and a deadline. A metric is anything you can measure; a KPI is the metric you chose because it tells you whether a specific business goal is on track. Most teams work best with one primary KPI and two to four supporting KPIs per funnel stage, which is far fewer than their dashboards contain.

Key performance indicator (KPI) is a metric chosen to show progress toward a specific business goal, with a target value, a named owner and a time frame. Every KPI is a metric, but most metrics are not KPIs.

What is the difference between a KPI and a metric?

Your analytics tool can report hundreds of metrics: sessions, page views, scroll depth, time on page, clicks per button. Each of them is a fact about what happened. None of them is a KPI until someone decides it matters for a goal.

A useful test is to ask three questions about any number:

• Does it connect to a business goal? “Revenue from new customers this quarter” does. “Average time on page” usually does not, at least not directly.
• Would a change in it change a decision? If conversion rate drops by a third, you pause a campaign or fix a checkout bug. If page views drop by a third because you removed a slideshow, you probably do nothing.
• Can a team influence it? A KPI has to be something a person or team can move through their own work.

Analytics tools draw a similar line. Google Analytics 4, for example, lets you mark the events that measure “an action that’s particularly important to the success of your business” as key events, and only those feed its key event reports. Marking an event as a key event is a good first filter, but it is not yet a KPI.

If a number passes all three questions, it is a candidate KPI. It becomes an actual KPI once it has a target (“raise conversion rate from 1.5% to 1.8% by end of quarter”), an owner (one person, not “marketing”) and a review rhythm. Metrics that fail the test are still useful. They become the diagnostic layer you open when a KPI moves and you need to know why.

The difference is also one of quantity. A team might watch 200 marketing metrics and hold itself accountable for 10 KPIs. That ratio is healthy.

How do you pick marketing KPIs for each funnel stage?

The startup investor Dave McClure’s widely used “pirate metrics” framework splits a customer’s life into five stages: acquisition, activation, retention, referral and revenue (AARRR). For an online business that also pays for reach, it helps to add an awareness stage in front and to fold referral into retention. That gives five stages: awareness, acquisition, activation, revenue and retention.

The table below lists KPI examples for each stage. Treat them as a menu: pick one primary KPI per stage and add supporting ones only when they explain the primary one.

To go from the menu to a short list, work through these steps:

1. Start from the goal, not the dashboard. Write down the quarter’s business goal in one sentence, such as “grow revenue from new customers by 20% without raising ad spend.”
2. Find the stage that limits the goal. If plenty of people arrive but few add to cart, activation is your constraint, and that is where the primary KPI should sit. A sales funnel analysis shows which step loses the most users.
3. List candidate metrics for that stage and the stages next to it. Use the table above as a starting point.
4. Run each candidate through the three-question test. Drop anything that does not connect to the goal, would not change a decision or cannot be influenced.
5. Choose one primary KPI per stage and at most four supporting ones. Supporting KPIs should explain the primary one, not compete with it.
6. Give every KPI a target, an owner and a review date. Without these, you have a list of metrics with a nicer name.
7. Add a guardrail. For each primary KPI, name one number that must not get worse while you push it. If you push conversion rate, watch AOV so you do not buy conversions with discounts.

Paid acquisition KPIs need their own arithmetic

Awareness and acquisition KPIs on paid channels are linked by three formulas. They are worth knowing by heart, because a change in one always shows up in the others:

• CPM = ad spend ÷ impressions × 1,000
• CTR = clicks ÷ impressions
• CPC = ad spend ÷ clicks, which also equals CPM ÷ (CTR × 1,000)

Say you spend $40,000 in a month and get 5,000,000 impressions. Your CPM is $40,000 ÷ 5,000,000 × 1,000 = $8. If 80,000 people click, your CTR is 80,000 ÷ 5,000,000 = 1.6% and your CPC is $40,000 ÷ 80,000 = $0.50. If CPM rises to $10 and CTR stays at 1.6%, CPC rises to $0.625. That is a cost increase you can see before revenue moves, which is exactly what a leading KPI is for.

These numbers only make sense if your paid traffic is labeled consistently. Tag every ad link with UTM parameters so that sessions, orders and revenue can be traced back to the campaign that paid for them.

What does a KPI tree look like for an online store?

A KPI tree breaks one top-level KPI into the drivers that multiply or add up to it. It shows each team which branch it owns, and it turns “revenue is down” into “conversion rate on mobile is down”, which is a problem someone can work on.

For an online store, the classic decomposition is:

Revenue = sessions × conversion rate × average order value

Here is a hypothetical store for one month. Every number in it is invented for the example.

Check the arithmetic: 200,000 sessions × 1.5% = 3,000 orders, and 3,000 orders × $60 = $180,000.

Now the tree helps you set targets. Suppose the team believes it can lift the cart-to-order rate from 25% to 30% by simplifying checkout. Orders become 12,000 × 30% = 3,600, conversion rate becomes 3,600 ÷ 200,000 = 1.8%, and revenue becomes 3,600 × $60 = $216,000. That is $36,000 more per month without spending more on ads. The same tree tells you that matching that gain through traffic alone would take 240,000 sessions at the old conversion rate, or roughly 40,000 more paid sessions at a $0.50 CPC, which is $20,000 of extra spend.

For the paid branch, compare the extra revenue with the extra spend using the ROAS formula, and for the retention branch, connect repeat purchase rate to customer lifetime value so you know how much a new customer is really worth.

What are leading and lagging indicators?

Kaplan and Norton, who introduced the balanced scorecard, argue that a scorecard should contain “outcome measures and the performance drivers of those outcomes, linked together in cause-and-effect relationships.” In marketing language, the outcome measures are lagging indicators and the performance drivers are leading ones.

Lagging indicators measure outcomes after they have happened: monthly revenue, orders, repeat purchase rate. They are the numbers the business ultimately cares about, and they are usually easy to measure but hard to influence directly. Leading indicators move earlier and predict those outcomes: add-to-cart rate, checkout start rate, CTR, new users from a campaign. They are easier to act on, but their link to the outcome has to be checked rather than assumed.

A good KPI set pairs the two. In the tree above, revenue and repeat purchase rate are lagging. Add-to-cart rate, cart-to-order rate and paid sessions are leading. If the add-to-cart rate falls in the first week of the month, you know revenue will be soft long before the month-end report confirms it, and you can investigate a broken page or a weak campaign while there is still time to recover.

Two practical rules help:

• Review leading KPIs weekly and lagging KPIs monthly. Weekly revenue is often too noisy to act on; weekly add-to-cart rate usually is not.
• Test the link once in a while. If add-to-cart rate has risen for three months and revenue has not followed, the leading KPI is not leading anything. When you test a change to a leading KPI with an experiment, size it properly first, as described in our guide to A/B test sample size.

What are the most common KPI mistakes?

Most KPI sets fail quietly. Nobody argues with them; people just stop looking. These are the usual causes.

Vanity metrics. Follower counts, total page views and raw impressions look good on a slide and rise easily, but they rarely change a decision. A follower count can double while revenue stays flat. Keep these as diagnostic metrics, not KPIs, unless you can show that they lead to an outcome you care about.

Too many KPIs. When everything is a KPI, nothing is. A team with 25 KPIs spends its review meeting reading numbers aloud instead of deciding anything. If your set does not fit on one screen, cut it.

No target and no owner. A KPI without a target cannot be on or off track; it can only go up or down. A KPI owned by “the team” is owned by nobody. In George Doran’s original 1981 SMART article, the A stood for “assignable”: specify who will do it. Later versions often swapped it for “achievable”, which quietly drops the owner. Doran also wrote that not every objective will meet all five criteria, so treat SMART as a checklist, not a rule.

Optimizing a KPI into meaninglessness. The anthropologist Marilyn Strathern summed up what is now called Goodhart’s law, after an observation by the economist Charles Goodhart: “When a measure becomes a target, it ceases to be a good measure.” If your KPI is cost per lead, the cheapest way to hit it is to accept low-quality leads. Guardrails exist for this: pair cost per lead with lead-to-customer rate, and conversion rate with AOV.

Inconsistent definitions. If one report counts orders including cancellations and another excludes them, the KPI is not one number. Write down the exact definition and data source for every KPI, and keep campaign tagging consistent so channel-level KPIs do not drift. When credit for a sale is split across channels, decide which attribution model your KPIs use and keep it fixed for the quarter.

How do you track marketing KPIs in AdPix?

Once the AdPix tag is on your site and purchases are tracked, AdPix reports supply most of the numbers in the tree above:

• Acquisition reports show sessions and users by channel, source / medium and UTM-tagged campaign. That covers the sessions branch and the acquisition KPIs.
• Funnels follow each step from first visit to purchase and show where users drop off, which gives you add-to-cart rate and cart-to-order rate.
• Order analysis, available on every plan, gives you order counts. From the Starter plan up, the online sales and revenue reports add revenue by product and campaign; divide revenue by orders to get AOV.
• One view of every user connects a person’s visits across devices and channels into one journey, which keeps retention KPIs from counting the same customer twice.
• Segments and cohorts let you follow customers acquired in one month and see how many order again, which is the cleanest way to track repeat purchase rate. Segments are available from the Starter plan up.

AdPix reports are not sampled, so a KPI you read in week one is computed on the same full data as the one you read at month end. Realtime data helps when you launch a campaign and want to check that its leading KPIs are moving. You can also export data at any time, which is useful if you keep targets and owners in a shared spreadsheet next to the numbers.

If you already use Google Analytics, AdPix can run alongside it while you compare the two; the Google Analytics alternative guide shows how to run that comparison without breaking your KPI history. The Free plan covers 100,000 events a month, and the pricing page lists the limits of each plan.

Your next step is small: write down this quarter’s business goal, find the funnel stage that limits it, and pick one primary KPI for that stage with a target, an owner and a guardrail. Build the rest of the tree from there.

CPM · CPC · CTR calculator

Cost and click metrics for any campaign. The free plan covers 100,000 events a month, no credit card. Setup takes under five minutes.

Start freeTry it free