A scorecard that cannot tell someone something uncomfortable
is a horoscope with arithmetic.
For founders and experts who want an assessment that separates people and tells them something true. This playbook is how to design and build an assessment that separates people, tells them something true, and turns the result into a next step.
// 39 PAGES · SIX PARTS · THREE CHECKLISTS · FREE
Eight things to know about building a scorecard.
Each is a mechanism you can plan around, and the guide is plain about the limit.
The content is the product, and the questions are the cheap half
The instrument produces a coordinate — this person is in band three on pillar four. What they *receive* is a narrative for that coordinate and a set of things to do about it.
Choose the grid you are willing to write, and let that decide your pillar and band counts — not the other way round
Every failed scorecard I have examined chose the structure first and discovered the writing bill afterwards.
There are two viable sizes and nothing in between
Across nine live instruments the sizes cluster hard: lead magnets at 20–29 questions, one to six pillars, five to ten minutes; paid diagnostics at 125–145 questions, five to seven pillars, thirty to thirty-five minutes.
Discrimination is the quality metric, and no tool computes it for you
I checked: across the entire analytics layer of a mature assessment product there is no item-total correlation, no reliability coefficient, no discrimination index.
Measurement without advice is a quiz
This is the 145-questions-zero-advice failure, and it is seductive because building the measurement feels rigorous and writing the advice feels like copywriting.
Every archetype needs a downstream consequence
A band that produces the same next step as the band above it is not a band, it is a decoration. The cleanest discipline I know: write the five follow-up emails before you write the questions.
And the shared result is permanent
In every implementation I have looked at, a results URL is unauthenticated, complete, and never expires. Someone forwards it and a stranger reads it cold.
And the honest limit
A scorecard cannot create demand for a problem nobody has. If the subject is not something people already worry about, no amount of scoring will make them worry, and you will get completions with no conversations.
Six parts, and a last page.
Parts one to six are the thinking and the doing, worth your time even if you never touch a piece of software. The last pages show how to keep the machinery in one product.
What you are actually building
The measurement model
The verdict
The payload
Calibrate it
The checklists and the mistakes
Do it all in one place
Nine mistakes, all of them avoidable.
None is about design talent. They are failures of budgeting and calibration, which is good news, because both can be done in advance.
Choosing the structure before computing the content budget
Eight pillars and seven bands is 56 narratives. You will abandon it, and you will abandon it after building all the questions.
Measuring without advising
A hundred and forty-five questions and zero action items is a real, live instrument. It can tell you exactly where you stand and nothing about what to do.
Flattering everyone
A result incapable of being uncomfortable produces no urgency and no credibility, and thoughtful people spot it in seconds.
Mixing option ladders
A yes/no question scoring 100 among neighbours that top out at 4 silently owns its entire pillar.
Editing options after launch
It breaks translations, breaks in-flight respondents, and changes the scale for everyone who came before — with nothing recomputing and nothing warning you.
Boundaries that touch
0–30 / 30–50 double-matches at exactly 30 and the winner is an accident of sort order. Author at .01.
Grades instead of identities
"Level 2 of 5" is unmemorable, unforwardable, and useless in a lead notification.
Never calibrating
Everyone in one band means the cut points are wrong, and it is arithmetic to fix — but only if you look.
Bands with nothing listening
The most common silent failure there is: the tagging works, the rules fire, and no sequence exists on the other end. Nothing errors. It just looks like a quiet week.
Thirty-nine pages.
Yours in one click.
Six parts, three checklists and the nine mistakes. One email with the link, no drip sequence, and we do not sell the list.