{{img:hero}}Analytics can feel like a pile of numbers that all want to be “the one.” This guide is a myths-vs-reality decision framework: if you’re trying to answer X, then look at Y, and ignore Z (with examples you can adapt).

Less guessing, more “what decision does this number support?”

Assumption: you’re viewing reports in Chrome on a Mac, but the logic works in any analytics tool.

Myth #1: “More data means more clarity”

Reality: More data usually means more chances to confuse a signal with noise.

If you can’t say what action you’d take if the number goes up or down, then it’s not a decision metric yet—it’s just a chart.

Example: You add 12 widgets to a dashboard. Everyone stares at it. Nobody changes anything.

If/then fix:

  • If this metric changes, then what will we do this week?
  • If we wouldn’t do anything, then remove it from the main view (keep it in a drill-down).
  • If two metrics tell the same story, then keep the one that’s easiest to explain.

{{img:funnel}}A helpful rule: one “decision dashboard” and one “diagnostic drawer.”

Myth #2: “Pageviews (or sessions) are the best north-star metric”

Reality: Volume metrics are only “best” when volume is actually your goal.

If your business model depends on ads or awareness, then pageviews/sessions can be primary.

If your business model depends on signups, trials, leads, or purchases, then pageviews/sessions are context, not success.

Example scenarios:

  • If you run a newsletter, then measure visitor → email signup rate (plus the volume feeding it).
  • If you sell a product, then measure purchase conversion rate and revenue per visitor (volume helps, but it’s not the win by itself).
  • If you’re B2B, then measure qualified lead rate (and define “qualified” in plain language).

Quick gut-check: if traffic doubled but conversion halved, then would you call that success? Your answer tells you whether volume is the right “north star.”

Myth #3: “A spike or drop means something important happened”

Reality: A spike is often tracking changes in tracking, not changes in behavior.

If a metric changes suddenly, then run this order of operations before you narrate a story:

  • If multiple metrics moved together (sessions, conversions, revenue), then it’s more likely real.
  • If only one metric moved (especially an event count), then suspect instrumentation.
  • If the change lines up with a release, tag change, consent change, or campaign launch, then start there.
  • If the change is isolated to one browser/device/region, then check segmentation artifacts (not just “the site is broken”).

{{img:magnifier}}Example: “Purchases” drop 30% but payment provider revenue is flat. Then it’s probably a tracking issue (or attribution/reporting delay), not an actual sales collapse.

Myth #4: “Conversion rate is always the best way to compare performance”

Reality: Conversion rate can lie when the mix of traffic changes.

If you change acquisition (new channel, new campaign, new geography), then compare both volume and rate.

If you only look at conversion rate, then you might miss that you actually improved total outcomes.

Example:

  • Week A: 1,000 visitors, 5% convert → 50 conversions
  • Week B: 2,000 visitors, 3% convert → 60 conversions

If your goal is total conversions, then Week B is better even though the rate fell.

If your goal is efficiency (e.g., paid spend is fixed), then the rate drop might be a real problem.

Decision framing that helps teams stay calm:

  • If we’re capacity-limited (support, onboarding, inventory), then efficiency metrics matter more.
  • If we’re growth-limited (need more customers), then total outcomes matter more.

Myth #5: “Attribution tells you what caused the conversion”

Reality: Attribution usually tells you how credit was assigned, not what truly caused the decision.

If you need to decide where to spend money next, then attribution is a starting point, not a verdict.

If you change the attribution model (or the reporting window), then numbers will “move” even if reality didn’t.

Practical if/then use:

  • If you’re comparing two paid channels, then look at on-platform results and your site conversions for the same period (expect mismatch).
  • If you’re launching something new, then track a simple before/after for the target action (signups, demos, purchases) alongside attribution.
  • If you need true causality, then consider experiments (holdouts, geo tests) rather than arguing over last-click vs first-click.

{{img:compass}}A good mental model: attribution is a compass, not a GPS route replay.

Myth #6: “Averages are the most honest summary”

Reality: Averages hide the shape of what’s happening.

If you’re summarizing performance, then pair an average with one distribution-friendly check.

  • If it’s time-based (load time, time to purchase), then use median and a p90/p95 style tail metric.
  • If it’s money-based (order value), then check whether a few large orders are skewing the mean.
  • If you’re seeing “average time on page” jump, then confirm it’s not a tracking artifact (like double-counted sessions or missing exit events).

Example: median checkout time improves, but p95 gets worse. Then most people are fine, but a minority is hitting a real friction point worth investigating.

Takeaway: a small, repeatable decision script

When a metric shows up in a meeting, run this in order:

  • If we can’t name the decision it supports, then it doesn’t belong on the main dashboard.
  • If it moved suddenly, then check instrumentation/segmentation before storytelling.
  • If it’s a rate, then also look at volume and traffic mix.
  • If it’s attribution, then treat it as directional—not causal.
  • If it’s an average, then add one “shape” check (median or tail).

That’s the whole framework: fewer charts, clearer decisions, and less panic when the line wiggles.