How to Choose the Right Marketing Metrics and KPIs

For marketers and analysts who open GA4 or an ads dashboard, see 50+ metrics staring back, and pick the first few that sound important. This is the filtering logic that turns "here's a lot of data" into "here's what we should actually track."

Not every metric deserves to be tracked, and almost no metric deserves to be a KPI. Before adding a number to a dashboard, know three things about it: what type it is (a platform default, something you built, or something you calculated), what job it's doing (exposure, engagement, perception, experience, cost-efficiency, or conversion), and whether it's tied to a real business goal at all. Most reporting problems come from skipping straight to the numbers without doing this filtering first.

---

The real problem isn't a lack of data

Open GA4's default reports, or any major ad platform, and the problem was never going to be finding metrics. Facebook alone surfaces reach, impressions, engagements, video completion rate, cost per result, and dozens more — and that's before you add a single custom event. The problem is that most of what's on offer isn't relevant to what your business is actually trying to prove, and dashboards tend to fill up with whatever the tool shows by default rather than what someone deliberately chose.

The fix isn't a bigger dashboard. It's a filtering process — a way of looking at any given metric and deciding, quickly and defensibly, whether it belongs anywhere near a report your team makes decisions from.

---

Start with a real business objective, not a metric

Everything below is a filter, and a filter needs something to filter against. That something has to be a real business objective — and it's easy to grab something that looks like one but isn't.

"Increase Facebook followers," "improve engagement rate on social content," and "achieve a $1.50 cost per acquisition" are not business objectives. They're a means to an end. Each one is already shaped like a metric — it's specific, it's measurable, it even sounds like a goal — but none of them says why the business would be better off. A business objective has to tie to something that creates material value for the business, not to a number on one platform.

In marketing, almost every real business objective is a version of one of five things:

• Awareness — increase brand or product awareness
• Positioning — reposition the brand, or improve how it's understood
• Perception — improve how people feel about the brand or product
• Consideration — increase purchase or usage consideration
• Leads/Sales — drive sales, revenue, or conversion

Run the three fake objectives back through this list and they stop being objectives and start being candidate metrics: more Facebook followers might be evidence of progress on an awareness objective. A better engagement rate might be evidence of progress on a consideration objective. A lower cost per acquisition might be evidence of progress on a leads/sales objective. The objective is the "why"; the metric is how you'll know. Collapse the two and you end up optimizing a number that was never tied to anything the business actually needed — which is exactly how vanity metrics end up on dashboards in the first place.

Once you know which of the five categories an objective falls under, unpack it with four more questions before you go anywhere near a metric:

• Problem — What problem or challenge is the business actually trying to overcome?
• Location — What country, region, city, or market does this apply to?
• Target audience — Who exactly is this for, and what defines them?
• Timeframe — When does the campaign start and end, and by when does the objective need to be hit?

"Increase brand awareness" answers none of these on its own. "Increase awareness of [product] among new parents in [city] over the next quarter" does — and it's that unpacked version, not the one-liner, that should be driving which metrics get chosen.

Before applying any of the filtering below, be able to state the real objective in one sentence and name which of the five it falls under. Everything downstream — metric type, metric category, hard vs. soft, KPI or not — is a test of whether a given number actually serves that sentence.

For the full path from business objective through goals, targets, and segments, see our Digital Marketing and Measurement Model post. This post picks up from there, once the objective is set, and focuses specifically on filtering the metrics and KPIs that prove it.

---

Start with metric vs. KPI

A metric is anything you can quantify. A KPI is the small number of metrics your team has explicitly chosen because they're tied to a business goal someone is accountable for. Every KPI is a metric, but the reverse almost never holds — "average session duration" is a real, measurable metric, but it's rarely anyone's job to move it, which is exactly why it shouldn't sit on a KPI dashboard next to numbers that are.

The practical test: for any metric currently on your dashboard, can you name the specific goal it's supposed to be evidence for, and who's accountable for moving it? If the answer is no, it's clutter — not wrong to look at, just misplaced. Move it to a deeper exploration report you check when something looks off, and keep the dashboard itself to a handful of numbers someone actually owns.

---

Know what type of metric you're looking at

Before deciding whether a metric matters, it helps to know how it was produced — because that tells you how reliable and how fragile it is.

Metric type	What it is	Example
Platform (off-the-shelf)	Comes native from the tool, zero setup required	Sessions, impressions, likes — built into GA4 or an ad platform by default
Custom (engineered)	You built it yourself, usually through event tracking	A specific CTA click tracked via Google Tag Manager, a scroll-depth trigger
Calculated (derived)	Combines two or more existing metrics into a new one	eCommerce transactions ÷ GA4 users, to build a blended conversion rate

This distinction isn't academic — it directly predicts how a metric can fail you. Platform metrics are on loan from the tool: when GA4 replaced Universal Analytics, page load time simply stopped being a native metric. It hadn't been deprecated by mistake; it was just no longer part of the default platform set, and anyone who relied on it had to go rebuild it as a custom event through Tag Manager. A calculated metric, by contrast, only breaks if one of its inputs breaks — which is usually easier to diagnose. Knowing the type of every metric on your dashboard tells you, in advance, which ones are at risk the next time a platform ships a redesign.

---

Know what job the metric is doing

Every metric, regardless of channel, is really trying to answer one of a small number of questions. Sorting metrics by the question they answer — rather than by which platform they come from — is what makes a dashboard readable across channels instead of becoming a wall of platform-specific jargon.

Category	Question it answers	Typical metrics
Exposure	Did anyone see this?	Impressions, reach, sessions
Engagement	Did they do anything once they saw it?	Clicks, time on page, engagement rate
Perception	Did they like what they saw?	Sentiment, likes vs. dislikes, comment tone
Experience	Was the experience itself any good?	Page load time, error rate, crash rate
Acquisition	What did it cost to get this outcome?	CPC, CPM, cost per lead
Conversion	Did they actually do the thing that matters?	Leads, purchases, conversion rate

A useful sanity check: look at your current dashboard and ask which category is overrepresented. A dashboard that's 80% exposure and engagement metrics with nothing in conversion is measuring attention, not outcomes — which is fine if attention genuinely is this quarter's objective, and a real problem if it isn't.

---

Split conversion metrics into hard and soft

Conversion metrics deserve one more layer of care, because "conversion" gets used loosely. Split every conversion metric into hard (tied directly to revenue) or soft (a real outcome, but not revenue itself).

	Hard conversion	Soft conversion
Ties to	Revenue, directly	A meaningful step before revenue
Examples	Purchases, signed contracts, paid subscriptions	Leads, free trial signups, content downloads, newsletter signups
What it tells you	Did the business make money	Is the pipeline that eventually makes money healthy

Both are legitimate KPIs — the mistake is treating a soft conversion as if it were proof of revenue impact. "We generated 200 leads" and "we generated $40,000 in revenue" are different claims, and a campaign that's great at the first can still be mediocre at the second if lead quality or follow-up is weak. Keep them visually and conceptually separate on any report that goes to a decision-maker.

---

Run the survivors through the DEM test

A metric can pass every filter above — it's tied to a real objective, its type and category are known, it's correctly labeled hard or soft — and still not deserve a spot on the dashboard. Before it becomes a KPI, run it through one last check: is it Defendable, Explainable, and Meaningful?

• Defendable — If someone asks how the number was calculated or where it came from, is there a clear, repeatable answer? "The platform just shows it that way" doesn't count.
• Explainable — Can you say, in one sentence, why the number moved? If it takes a paragraph of caveats to interpret, it isn't ready for a dashboard yet, even if it's accurate.
• Meaningful — Would the number changing actually change a decision? If it could double or halve and nobody on the team would act any differently, it isn't meaningful — no matter how good it looks going up.

"Average session duration" is a useful example of a metric that fails this test while passing everything else: it's a real platform metric, it sits cleanly in the Engagement category, and it's usually both defendable and explainable. For most businesses it still fails Meaningful, because nobody changes what they do based on whether the number was 2:14 or 2:31. DEM is the gate that catches what survives type and category filtering but still shouldn't make the cut.

---

Putting it together: a worked example

A local service business runs a paid social campaign with the business objective of generating qualified consultation requests — a real objective, since it's a Leads/Sales objective tied to revenue, not a stand-in like "increase reach." Walking the filtering logic:

1. Channel type: Paid (see our customer journey mapping post for the full Paid/Owned/Earned/3rd-party breakdown) — its job in the journey is awareness and initial interest, not the close.
2. Metric category by stage: Reach and impressions (exposure) show whether the ad is getting seen at all. Click-through rate (engagement) shows whether the creative and targeting are working together. Cost per lead (acquisition) shows efficiency. Consultation requests (conversion — soft, since no money has changed hands yet) is the actual thing the business cares about.
3. Metric type: Reach and CTR are platform metrics, free out of the box. "Consultation request" is a custom event someone had to set up in GA4 or the CRM — it doesn't exist until it's engineered.
4. KPI or not: Of everything measured, maybe two make the KPI dashboard — cost per lead and consultation requests. Reach and CTR still get checked, but in a diagnostic view, not the top-line report, because nobody's actual job is "increase reach."

That's four short steps, and it's the difference between a report with 15 metrics nobody reads past line three, and one with two numbers everyone on the team can recite from memory.

---

Where this goes wrong

• Vanity metrics get promoted to KPIs. Followers and pageviews feel good to report, but they're almost never tied to a specific accountable goal — see our Views Aren't Growth post for the content-specific version of this exact mistake.
• A soft conversion gets reported as if it were a hard one. "Leads are up" quietly becomes "revenue is up" in a meeting, and nobody notices the gap until pipeline numbers don't match.
• A platform metric quietly disappears and nobody notices for weeks. This is exactly what happened across the industry during the GA4 migration — metric definitions and availability changed, and reports that weren't rebuilt kept showing stale or missing numbers. Our GA4 attribution fix for AI traffic post covers a related version of this problem.
• The dashboard has no stated business objective behind most of its rows. If you haven't done the Digital Marketing and Measurement Model work first — objective, goals, targets, segments — metric type and category filtering has nothing to filter against.

---

The short version

Before a metric earns a spot on a dashboard, know its type (platform, custom, or calculated), know its job (exposure through conversion), know whether it's hard or soft, and know the specific goal it's supposed to prove progress on. Most of what a tool shows you by default fails at least one of those checks. The metrics that survive all four are the ones worth calling KPIs — and there should be a lot fewer of them than the dashboard currently has.

Want a real metrics framework built for your business?

JYNLAB helps small and mid-sized businesses go from "a dashboard full of numbers" to a small set of KPIs tied to real business goals — and the GA4 setup to track them correctly. If you want a partner to help build that filtering process for your marketing, book a demo call and we'll walk through where you stand.

Book a Demo Call