Morten A. Giraffe

Chapter III / Make it responsible

Measure the Outcome—not the Decoration

Tie design evaluation to the user task, business outcome, evidence, guardrail, and decision that should follow.

All 18 rules4 min read
01Scenario

Start with the moment where the interface asks too much.

The redesign dashboard looks victorious. Page views are up. Time on page is up. Clicks are up. Then qualified inquiries fall. Visitors needed more pages and more clicks to understand the offer. Activity rose because the experience became harder. The numbers were real. The interpretation was not.

Design should be evaluated against the task and business outcome it was meant to improve, using behavior and observation together. More clicks, motion, or time on page are not automatically success.

02Interface Lab

Show the failure and the correction.

The lab uses controlled markup instead of raster text so the lesson stays legible, accessible, and easy to revise.

Annotated screen

Rule 18

123
  1. 1Vanity dashboardThe report celebrates more activity without connecting it to the page job.
  2. 2Retrospective hypothesisThe team decides what the redesign was supposed to prove only after seeing the numbers.
  3. 3No guardrailsA conversion increase is accepted even if lead quality, accessibility, or trust declines.
03Principle

Make the useful path easier.

Outcome, diagnostic, guardrail

Outcome metrics answer whether the task improved. Diagnostic metrics help explain why. Guardrails prevent improvements that harm trust, accessibility, performance, or downstream quality.

Analytics and observation need each other

Analytics can show where behavior changed. Observation, support notes, and qualitative feedback help explain why it changed.

Small data requires honesty

Low-volume businesses may not have statistical certainty. That does not block improvement; it requires cautious language and direct evidence.

04Rebuild

Turn critique into a usable rule.

Remove

  • Vanity dashboard

    The report celebrates more activity without connecting it to the page job.

  • Retrospective hypothesis

    The team decides what the redesign was supposed to prove only after seeing the numbers.

  • No guardrails

    A conversion increase is accepted even if lead quality, accessibility, or trust declines.

Build

  • Return to the page job

    Purpose defines the outcome worth measuring.

  • Pair behavior with observation

    Use analytics for where and qualitative review for why.

  • Decide before launch

    Write what result would make the team keep, revise, or reverse the change.

05Field Test

Use it on a real page.

Before launch, write the measurement plan in one table.

  1. Step 1

    Name the expected behavior change.

  2. Step 2

    Choose outcome, diagnostic, and guardrail evidence.

  3. Step 3

    Set a review window.

  4. Step 4

    Define what would cause keep, revise, or reverse.

  5. Step 5

    Compare results with observation after release.

Evidence
Use analytics, support notes, direct observation, and qualified inquiry review where appropriate.
Pass
The team can make a decision from the evidence without pretending certainty.
Fail
The report celebrates activity with no link to user task, business outcome, or guardrail.
06Use By Role

The rule should help the person making the next decision.

Business owner

Ask whether the metric represents a useful business and visitor outcome, not just activity.

Designer

Write the design hypothesis before changing the interface.

Developer

Instrument useful events without sending personal form values or noisy vanity data.

Reviewer

Watch for displacement: one number improving while a more important outcome worsens.

Not every valuable outcome is immediately measurable. Use qualitative evidence and honest judgment rather than inventing certainty or refusing to improve.

07Checklist And FAQ

Check the page, then answer the doubt.

  • The page job defines success.
  • Outcome metrics are separated from diagnostics.
  • Guardrails protect trust and access.
  • The hypothesis is written before release.
  • Analytics avoids personal data.
  • Observation helps explain behavior.
  • Low-volume uncertainty is stated honestly.
Why can more clicks be a bad sign?
More clicks may mean the visitor had to work harder to understand, compare, or complete the task.
Which conversion metrics matter?
The ones tied to the page job: qualified inquiry, completion, comprehension, recovery, activation, retention, or another real outcome.
How do analytics and usability testing work together?
Analytics shows behavior patterns; usability review and observation help explain the reasons behind them.