Before You Call It "Measured" — How Marketing Tools Should Label Estimates Honestly

A Score Borrows Trust

The scores marketing tools produce work on the premise that the reader will not recompute the basis. That is why scores carry responsibility. While building diagnostic tools like AEGIS Insight and C³ Cube, we spent recent months checking ourselves on how well we keep that responsibility. A fair amount of what we found was embarrassing, which is why we are publishing it. The same problems are common both among those who build scores and among those who read them.

The Labeling Gaps We Found

What this list has in common is that these were problems born of inertia, not bad intent. Names and defaults that looked reasonable at first drifted away from actual behavior over time.

Three Labeling Principles

First, label the source type of a value. Directly collected values, model-inferred values, and pre-publication self-assessments are different kinds of information. We attached a reproducibility tier to each item, rating structural items checkable in code highest and items dependent on external signals lower. To be honest, the percentages in those tiers are approximate labels we assigned, not statistically guaranteed values, and that should be stated too.

Second, do not let a name overstate the method. The moment the name "measured value" is attached, people stop questioning it and the model reading the prompt judges that data to be enough. This is what actually happened to us. A name changes behavior, not just labeling.

Third, do not fill in a value you could not obtain. The moment you fill a gap with a plausible value because you dislike a report looking empty, the tool starts producing performance instead of information. Leaving it blank or writing "could not be measured" is better.

A Label Alone Is Not Enough

The most instructive case was Naver search volume. Our pipeline already had a tag and a screen badge for whether a value was measured. But the step that computed the score was not actually reading that measured data. The label said "measured" while the calculation was using an inferred value. Labeling becomes false not only when a label is wrong but also when the label is right and the data flow is broken. So when auditing labels, you have to follow the path the data takes to the score, not just look at the screen.

From the Reader's Side

These principles are not only for those who build tools. If you receive a marketing score or ranking, you can ask three things. Where did this value come from? If measured again, would it give the same value? How was it handled when the value could not be obtained? If a tool cannot answer, it is safer to double-check before copying the number into a report.

What Still Remains

We are not perfect either. The AEGIS Index gives the largest weight to the GEO axis to reflect priorities in the AI era, but GEO depends heavily on external signals and is relatively the least certain axis. We documented this trade-off with its rationale on screen and in docs, but the structure itself has not gone away. And for Google-side search volume in C³ Cube, we have not yet secured a reliable measured source, so some reliance on estimation remains. Our standard is to make visible where we are incomplete rather than claim perfection.

Self-Check -- Where to Start Right Now

Pick the one number you rely on most in your current marketing dashboard and ask it the three questions above. For site-structure readiness, AEGIS Insight lets you see on screen which items are measured and which are estimated. That a score does not substitute for citations themselves is covered in our AEO KPI design post.

Closing Thoughts

A number's value comes not only from its accuracy but from how visible it is how it was obtained. Calling an estimate an estimate does not diminish a tool's value; it protects it. If there is a number you are about to copy into a report today, start by checking whether its name overstates the method.