Your report is technically working, and that is the problem. It carries twenty numbers, every one accurate, and you still cannot answer the question your CFO asked last Thursday. There is a version of this where the fix is a better platform, and a version where the fix is deleting most of the slide, and only one of those costs money. You already know which three numbers you actually look at, and you have never written down why the other seventeen are still there. Which number, if it moved, would change what you do next month? That number is a KPI, and the rest is decoration with a bill that never appears on the slide.
Your report is technically working, and that is the problem. It carries twenty numbers, every one of them accurate, and you still cannot answer the question your CFO asked last Thursday. There is a version of this where the fix is a better platform, and a version where the fix is deleting most of the slide. Only one of those costs money. Here is the uncomfortable part: you already know which three numbers you actually look at, and you have never written down why the other seventeen are still there. If someone asked you today which single number, if it moved, would change what you do next month, what would you say out loud? That number is a KPI. The rest is decoration, and decoration comes with a bill that never appears on the slide.
About 23% of marketers say they are confident they track the right KPIs (Ruler Analytics, quoted by most of the pages ranking for this term), and that figure has barely moved in three years. If you sit in the other 77%, your report is probably longer than it was in 2024, not shorter. The extra length has a price that never shows up in a screenshot: engineering hours keeping a pipeline alive, a spreadsheet somebody rebuilds on the first of every month, an add-on nobody has renegotiated since it was bought. The pages you are reading for help make this worse by publishing a longer list every quarter, and none of them ever removes one. The nine below are the ones I would defend in a budget meeting.
What a marketing KPI has to earn before it gets a slot
If you ask what is a marketing KPI, the honest answer is a metric that passes three questions. Everything else is a measurement, and measurement is not free.
- Does it change a decision? Name the decision, not the theme. "Campaign performance" is not a decision. "We pause this channel at month end" is.
- Can the team that owns it move it? A number your team cannot influence is somebody else's KPI. Watching it is fine. Reporting it as yours is not.
- Does the tracking cost less than the decision is worth? No page in the top twenty asks this, and it is the question that cuts the list fastest.
Apply that test to a standard twenty-metric marketing dashboard and the arithmetic turns brutal. Gartner's CMO guidance, quoted through 1ClickReport in 2026, caps an executive view at 5 to 7 KPIs and a channel view at 8 to 12. Past that, clarity falls. The same research finds 83% of executive teams expect marketing to demonstrate ROI through data-driven dashboards, so the pressure runs toward more numbers, not fewer. Both facts cannot be satisfied by a longer list. They can be satisfied by a short one where each entry carries a named decision.
The third question is where the field goes quiet, so it is worth pricing properly. A metric is not one cost, it is three: the instrument, the join that marries it to revenue or an account record, and the maintenance that keeps it true next quarter. Fivetran's 2026 infrastructure benchmark puts average annual data budgets at $29.3M per organisation, with $2.2M of that spent on engineering labour keeping pipelines running. Digital Applied's 2026 figures add the softer side: $12.9M average annual cost of poor marketing data quality at enterprise scale, 47 TB of data produced per marketing stack each month, and only 23% of it actively used. Surveys collected by Coupler.io put manual reporting at 6 to 10 hours a week for the average marketing team. You are not short of data. You are paying to carry data you never act on.
The nine marketing KPIs that survive the test
Nine is not a round number picked for tidiness. It is what remains after the test above, and it sits at the ceiling Gartner's guidance implies for a view that still gets read. Search for marketing KPI examples and every result hands you a longer list; this one hands you a shorter one with a reason attached to every line. Costs below are labelled: verified where a vendor list price or published figure sits behind it, derived where the cost is a known mechanism plus a supporting benchmark rather than a quoted price.
# | KPI | Decision it changes | Owner | Tracking cost |
|---|---|---|---|---|
1 | Qualified pipeline created (marketing sourced, per period) | Where next quarter's budget goes. This is the number that ends the vanity argument. | Demand generation lead | Derived, high. Needs CRM opportunity data joined to campaign membership. The join is the expense, not the metric: $2.2M a year of engineering labour on pipeline upkeep at enterprise scale (Fivetran, 2026). |
2 | Cost per qualified opportunity | Whether to pause or scale a channel this month. | Paid media lead | Derived, medium to high. GA4 imports Google Ads cost natively; non-Google platforms need weekly CSV exports or a paid connector. |
3 | CAC payback period | How much the business can afford to spend to win a customer at all. It caps every other number on this list. | Growth lead with a finance partner | Derived, medium. Needs marketing spend plus gross margin per customer. Half the input lives outside marketing's systems, which is why teams skip it. |
4 | LTV to CAC ratio | Whether current acquisition spend is sustainable or buying growth at a loss. | Growth lead with finance | Derived, medium. Same finance dependency, plus retention or churn from the product side. Plan it on a review cadence, not as a live number. |
5 | Lead to customer conversion rate by source | Which source to stop paying for. | Demand generation ops | Derived. Nearly free if the lead source field is populated at capture; retro-fitting source onto historical leads is a data project. |
6 | Landing page conversion rate | Which page gets rewritten this sprint. | CRO or content lead | Verified, near zero. The GA4 free tier covers it. If you need unsampled reporting and warehouse export, GA4 360 starts around $50,000 a year (EmberTribe and Cardinal Path, 2026). |
7 | Marketing efficiency ratio (total revenue divided by total marketing spend) | The total spend level for the month, before any channel gets defended. | CMO with finance | Derived, near zero. Two numbers from two systems you already own. The cost here is political, not technical. |
8 | Branded search demand volume | Whether the upper funnel is doing anything at all. | Brand or SEO lead | Verified, free, conditional. Search Console covers it at no cost. The condition is the whole point: the query filter is what makes it non-vanity. |
9 | Target account engagement coverage (share of your named target list showing engagement before a demo request) | Whether demand is being created, or only captured. | ABM or demand generation lead | Derived, highest in the set. Needs account matching, meaning reverse IP or an account intelligence tool. Reference points: listening as an add-on at around $999/month on Sprout Social, rank tracking at $0.02 to $0.28 per keyword per month (2026). |
Three of these need a sentence more than a table cell.
Qualified pipeline created is first because it is the only one that survives a hostile meeting. Every other number on the default report can be up while revenue is flat. The pain is the join, and I would rather you paid for that join once than paid for seventeen numbers forever. Your CRM holds the opportunity, your ad platforms hold the membership, and the bridge between them is the most valuable data engineering hour in a marketing department.
CAC payback and LTV to CAC need a finance partner, and that is not an obstacle. It is why they work. The moment gross margin enters the conversation, marketing stops reporting on a parallel track and starts reporting on the business. Review them monthly or quarterly rather than live, and expect the first version to be wrong in an instructive way.
Target account engagement coverage is the leading indicator that gets skipped. It is the hardest thing on the list to instrument, because account matching needs reverse IP resolution or a paid intelligence layer. For a list under a few hundred named accounts, the honest answer is a quarterly measurement run by hand. Say out loud that it is quarterly, and it still beats a monthly report full of impressions nobody can act on.
The vanity metrics, named, with the cost of keeping each one
A metric earns a place on the default report for social reasons, not analytical ones. Somebody built it, or it came with the template, or a vendor put it in the first row. Here is the set I would cut, and what cutting actually costs.
Vanity metric | Why it is on the report | What keeping it costs |
|---|---|---|
Raw impressions | First line of Google Ads and every social dashboard, and the 2024 academic page still ranking at position two recommends it as KPI number one. | Collecting it is free; including it is not. Every impression on the slide occupies a slot a decision could use, and a 20-metric list already breaks the 5 to 7 cap. |
Search engine rankings without a query filter | Rank trackers surface an average position by default, so it lands in the export whether you asked or not. | $0.02 to $0.28 per tracked keyword per month (verified band, 2026). The bill scales with keyword count, and the unfiltered average is the one number a filter would fix for free. |
Follower count | It only moves upward, so it always looks like progress. | Audience tooling in the $49 to $999+ per month band if anyone is asked to explain it. Two competing pages ranking for this term already call it vanity, so cutting it costs you no credibility. |
Page views without conversion context | It is the default landing page report in GA4. | Free to collect, and misleading for the same reason: heavy traffic to irrelevant pages hides bad targeting rather than revealing it. |
Email list size and open rate | Both grow without effort, and privacy features inflate opens. | Free, but the apparent engagement suppresses the deliverability work that would move revenue. |
Cumulative registrations or downloads without an active user ratio | Cumulative counts grow on inertia even while engagement declines. | Free to display, expensive to defend the first time a churn number appears beside it. |
Average position across all queries, branded included | One number, and it looks strategic. | Free, and the most easily gamed number in the set, because branded queries lift the average for reasons marketing did not cause. |
Every row is a metric somebody will defend in the meeting. That defence is the real cost of keeping it, which is why the test needs a third question about tracking cost, not only two about decisions and ownership.
Lagging versus leading: why a monthly report arrives too late
A report built only from the month's outcome cannot be acted on inside that same month. That is a timing problem, not a quality problem, and it is the most common structural flaw in a marketing KPI report.
Lagging indicators are outcomes. Qualified pipeline, revenue, CAC payback, conversion rate by source. They are the numbers that matter, and they are the numbers you can only read after the money is spent. A report holding nothing else is a scoreboard for a game that finished two weeks ago.
Leading indicators are inputs you can still change. Landing page conversion rate is the easy one, because a rewrite takes a sprint. Target account engagement coverage is the hard one, and it is the reason the nine-metric set has a leading entry at all. Add one more that almost nobody instruments: the share of target accounts that saw three or more touches before a demo request. It separates demand generation from harvesting people already in market.
The minimum viable leading set is two numbers, one behavioural and one reach-based, reviewed weekly. Everything else can stay lagging. What those two buy you is the ability to change course inside the month, which is the only reason reporting exists.
Where AI-search metrics sit in a marketing KPI report
Visibility and citation numbers are the newest addition to the pile, and they belong in one place: beside branded search demand, as a check on whether the upper funnel is moving, never as a replacement for pipeline.
Two things worth stating before you put them in front of a board. Several of these measures are not yet stable enough to hold anyone to a target, because the surfaces change and the samples move. And they are cheap to instrument relative to a CRM join but expensive to interpret, which is the inverse of everything else on this list. If you are building that part out, the AI-search metric subset covers the visibility and citation set properly. This page answers a different question, which metrics earn a slot on the report at all, in any channel, and it does not restate that set.
The boundaries here matter, because this cluster already holds four published pages. If your open question is which analytics platform to buy, that is the tools job and this is the selection job. If the problem is attributing the numbers to a touchpoint or picking a vendor from the attribution tool roundup, those pages own it. Here attribution appears once, as a line on the tracking bill priced in setup time, and no model gets compared: last-click runs about a day, a data-driven model four to eight weeks, marketing mix modelling eight to sixteen, and unified MTA plus MMM twelve to twenty (Digital Applied, 2026). That range is the closest thing to a public price list for tracking difficulty, and it is why the third question carries so much weight.
The paid half of the same set has its own home in the paid-search view of the same set. And every number has to land somewhere, which is why the dashboard these numbers live on matters more than the export format. When the connections are live, a report becomes a view. When they are not, it becomes a monthly rebuild.
One more cost decides whether metrics like CAC payback are even available to you: serious reporting sits behind a tier. On HubSpot's Marketing Hub, analytics and reporting requires the Professional tier at roughly $800 to $900 per month, with Enterprise around $3,200 to $3,600. If your team is on Starter, the constraint on your KPI list is a licence line, not a data problem.
Turning four numbers into one headline the team can reproduce
Four of the nine ladder up into one, and the arithmetic has to be repeatable by a colleague without you in the room.
Take marketing efficiency ratio as the headline. Total revenue divided by total marketing spend makes it the cheapest honest summary on the list and the hardest to argue with. Keep three supporting numbers under it: qualified pipeline created, cost per qualified opportunity, and CAC payback period. Three of the four come from systems you already own.
What I would not build is a composite score. Weighted indexes feel rigorous and fail in exactly one situation, the meeting where a channel owner asks how their number rolled in. If you cannot reconstruct the calculation on a whiteboard in under a minute, the score gets rejected and you are back to twenty metrics.
Frequently Asked Questions
What is a marketing KPI?
A marketing KPI is a metric that changes a decision and has an owner. It is a subset of your measurements, not a synonym for them. If the number moving would not cause anyone on your team to do something different, it is a measurement: useful for diagnosis, not for a KPI report.
How many marketing KPIs should I track?
Five to seven for an executive view and eight to twelve for a channel view, following Gartner's guidance quoted through 1ClickReport in 2026. The practical test is not the count but the defence: what would you do differently if this number fell? If you cannot answer, it comes off the slide.
What is the difference between a KPI and a metric?
Every KPI is a metric, and almost no metric is a KPI. A metric is any measurement you collect. A KPI passes the three-question test: it changes a decision, its owner can move it, and tracking it costs less than the decision is worth. Impressions are a metric. Qualified pipeline created is a KPI.
Which marketing KPIs matter most?
Qualified pipeline created, cost per qualified opportunity, and CAC payback period, in that order. Pipeline tells you whether demand exists, cost per opportunity tells you whether the channel earns its price, and payback tells you whether the business can afford the programme. The HubSpot State of Marketing Report for 2026 lands close to that ranking from a different direction: lead quality and MQLs (39%), lead to customer conversion rate (34%), and ROI (31%) are what marketers themselves name.
How often should I review marketing KPIs?
Lagging KPIs monthly, leading indicators weekly. Pipeline, payback and efficiency ratio are review-cadence numbers, and daily readings of them produce noise rather than action. Landing page conversion and target account engagement should be weekly, because the point of a leading indicator is that you can still change something before the month closes.
What goes on a marketing KPI dashboard?
Only the numbers in the review cadence, in the same order every time, with a threshold against each one. A marketing dashboard is a reading surface, not a data dictionary. Keep diagnostic measurements in the warehouse and pull them when a KPI moves.
The bottom line
Cut the list to nine, attach a decision and an owner to each one, and price the tracking before you commit to it. Three of the nine cost near zero to run. What changes is what your team stops doing: no reconciliation on the seventeen numbers you removed, no meeting where somebody defends follower count, no report that takes a week to assemble because it was never wired to its sources.
That last point is the one I care about most. A marketing KPI report becomes a monthly rebuild instead of a view when campaigns and reporting live in different systems and somebody has to carry numbers between them. Running campaigns and reading their results from one workspace is the structural fix. Pick the nine first. The plumbing is the easy part once the list is short.
