Analytics

The three metrics most dashboards get wrong

Analytics ยท Attribution

Most marketing dashboards are optimized to look busy, not to be useful. Impressions, reach, and engagement rate survive quarter after quarter not because they're the right numbers to watch, but because they're easy to report and almost always trend upward. That's a bad combination, and it's why so many reporting decks say a lot without actually saying anything.

Impressions without context

Impressions measure exposure, not interest. Reported alone, they reward channels that generate volume regardless of relevance. Pair them with a quality signal, like assisted conversions or a view-through window tied to actual pipeline, and impressions become genuinely diagnostic instead of decorative.

๐Ÿ’ก

Key insight: A metric reported in isolation is rarely dishonest on its own. It's the absence of a paired quality signal that lets it mislead. The number isn't lying to you; the report is just incomplete.

Engagement rate as a stand-in for intent

A like or a comment is a weak proxy for purchase intent, but it's often treated as the headline number because it's the easiest thing a platform surfaces natively. The metric worth tracking instead is engagement that correlates with a defined next step. An email signup. A demo request. A return visit. Not engagement in the abstract.

Last-click attribution

Last-click models systematically over-credit bottom-funnel channels like branded search, and under-credit the awareness and consideration touches that made the branded search happen in the first place.

MetricWhat it actually measuresPair it with
ImpressionsExposure, not interestAssisted conversions
Engagement rateInteraction volumeA defined next-step conversion
Last-click creditThe final touchpoint onlyA data-driven or position-based model

"A dashboard's job is to change a decision, not to fill a slide."

Common questions

Should I stop reporting impressions entirely?

No. Impressions are still useful as a measure of exposure. The problem is reporting them alone, without a quality signal alongside them.

What is a data-driven attribution model?

A model that distributes conversion credit across multiple touchpoints based on actual contribution, rather than giving all the credit to the last click before conversion.

The takeaway

A dashboard's job is to change a decision, not to fill a slide. Any metric that survives on a report purely because it's easy to pull and trends upward is worth questioning, even if nobody in the meeting wants to be the one who asks.

Key takeaways

  • Impressions measure exposure, not interest. Pair them with a quality signal.
  • Engagement rate is a weak intent proxy unless it's tied to a defined next step.
  • Last-click attribution systematically under-credits the touches that built awareness.
  • Any metric kept purely because it's easy to report and trends upward deserves scrutiny.

Related Articles

Read next.

Meta Ads

Why Meta Ads creative dies faster than your reporting shows

By the time CTR visibly drops, a Meta ad has usually been fatiguing for weeks.

Read Next โ†’
LinkedIn Ads

Why LinkedIn Ads look expensive until you do the math right

A high cost per click looks alarming next to Google or Meta. It's the wrong comparison.

Read Next โ†’
AI

Where AI automation actually saves marketing teams time

A grounded look past the hype, at the workflows worth automating first.

Read Next โ†’
Portrait of Karthikeyan Srinivasan

Karthikeyan Srinivasan

Founder of Anextera Technologies and Director of New India Social Welfare Foundation โ€” a technology entrepreneur working across AI, marketing, and analytics.

Newsletter

Founder Notes.

Monthly insights on AI, technology, marketing, business, and leadership. No spam.

Let's talk

Need help applying these ideas to your business?

I read every message myself.