How to Weight Scorecard Questions:Two Levers, Not Ten
How to weight scorecard questions: you have two levers, option values and pillar weights. Question count is not one of them. Here is how to set both properly.
You don't weight scorecard questions one by one. You have two levers: option values set the scale inside a pillar, and pillar weights set the balance between pillars. Question count isn't a lever at all. Use equal weights unless you can defend the difference out loud.
How to weight scorecard questions: the two levers that exist
You can't weight scorecard questions with a box on each one, so stop looking for it. Most tools either don't have one or save it and never read it. You'd set it, feel clever, and nothing would change.
What you actually have is simpler, and it fits on a card.
| Lever | What it controls | The trap |
|---|---|---|
| Option values | The scale inside a pillar, and how much one question can move it | A question's top score is whatever its best option is worth. Add a better option and everyone's old answers shrink. |
| Pillar weights | How much each pillar counts toward the overall result | They're ratios, so you can stop doing sums to reach a hundred. |
| Question count | Nothing. It is not a lever. | You cannot weight scorecard questions by adding more of them. It does not make the pillar count for more. |
That's the whole model. Everything else people worry about is one of these three in disguise.
Weights are ratios, so drop the arithmetic
Pillar weights get normalised. The tool divides each by whatever they add up to. So twenty, twenty, twenty, twenty, twenty and one, one, one, one, one produce the same scorecard.
Stop fiddling to make weights total a hundred. It buys you nothing and invites mistakes. Pick the ratios you believe and move on.
Then ask whether you believe any of them. Use unequal weights only when you can defend the inequality out loud, to a sceptical customer, in one sentence. "Cash flow counts double because a business that runs out of money doesn't get to fix anything else." That's conviction. "I wrote more questions about it" isn't.
Weight should express what you think matters. It shouldn't be a by-product of how many questions you happened to write. Conviction, not volume.
Free guide · 28 pagesBuild a scorecard that tells the truth.The Scorecard Playbook: an assessment that separates people and gives you a reason to talk.Get the free guide →Do not weight scorecard questions by counting them
This is the counter-intuitive one, and it catches almost everybody.
The average is taken inside the pillar first. So a pillar with three questions and a pillar with twelve count exactly the same if their weights match. More questions don't mean more influence. They mean more precision, or more noise, depending on whether the questions are any good.
Which gives you a cleaner way to decide how many questions a pillar deserves: as many as it takes to separate people, and no more. If you want to know whether a question is earning its place, how to find dead questions in a scorecard shows you how to read it.
If one question must count double, double its values
You'll sometimes want one question to matter more inside its pillar. Because per-question weights usually don't exist, the lever is the option values.
Make its options worth twice as much. That's it.
But understand the hazard before you do. A question's maximum possible score is normally computed as the highest value among its options. There's no field where you type it. The scale lives in the options themselves.
So pick one scale for the whole instrument and write it down, because nothing records the choice. Two conventions coexist in real data, zero to four and zero in steps of twenty-five up to a hundred, and nothing enforces either. Percentages still come out right because the maximum is derived. What breaks is everything that assumed a range.
The yes/no question that eats a pillar
The cleanest example of why this matters. Drop a yes/no question into a zero-to-four scorecard and Yes scores a hundred, sitting next to neighbours that top out at four. That one question silently owns almost the whole pillar.
Write it as a two-option question on your own scale instead.
The multiple choice with no ceiling
Pick-several questions have a maximum equal to the sum of every option. Five options worth twenty-five each give that one question a maximum of a hundred and twenty-five. On a zero-to-four scale that's more than a whole pillar.
Use them rarely, and set their values with that sum in mind. Free text is different: it contributes nothing to either side of the calculation, so it drops out rather than scoring zero.
Question type is a weighting decision in disguise
You might think the type of question is a design choice. It's a scoring choice. Each type has its own ceiling, and the ceiling is the weight you didn't mean to set.
- Single choice. The maximum is the best option. It moves on by itself. This is your spine.
- Pick several. The maximum is the sum of every option. It waits for a press.
- Slider. The maximum is always a hundred, and it waits. It also pre-answers itself at the midpoint, so it counts as answered and scores half before anyone touches it. In your data that looks identical to a deliberate middle answer.
- Free text. Scores nothing and drops out. It's the one question someone can legitimately skip.
A slider is the sneakiest. You add it for variety and it quietly hands every lazy respondent fifty per cent of a question. If you must use one, put it near the end and treat its midpoint as unreliable. The spine stays single choice.
Free assessmentHow trapped are you, really?Take the Corporate Suffocation Index, the free solopreneur readiness assessment.Take the assessment →The denominator rule: ask only what you score
Every question you show counts toward the denominator, answered or not. A skipped question is a zero, not an exemption. That's only fair if every answer is required, which is why serious implementations refuse partial submissions.
The deeper version bites harder. Any path that shows someone fewer questions must divide by fewer questions.
One implementation showed each person a subset and then divided by the whole instrument. Scores came out twenty-six to sixty-two points low. Seven out of seven people landed in the wrong band. Nothing errored.
If you branch, skip or personalise your questions, check the maths on the shortest path before you launch. A wrong score is a false statement about someone's business, in a document they may act on. That's worse than a clumsy layout, and it's invisible.
Freeze it before you launch
Add one better option to a live question and everyone's score on it drops, forever. A four-option question topping out at seventy-five becomes a five-option one topping out at a hundred. Every past answer of seventy-five falls from a full score to three quarters of one. Nothing recomputes. You now have two populations you can't compare, and no record it happened.
Worse, in most tools editing a question's options destroys and recreates them with new identifiers. Translations hang off those identifiers. Anyone part-way through is holding the old ones and scores nothing on that question. Fixing a typo can do real damage.
So: question wording is safe to edit. Option wording isn't. Changing an option after launch is a new version of the instrument, not an edit.
Scores are written once, at completion. There's no rescore. Every change to a value, a weight or a boundary forks the instrument. Tune before you drive traffic, then batch changes, date them in your own notes and never read a trend across an edit. The bands you cut the scores into are the other half of that fragility, and how to set scorecard band boundaries deals with them.
The pattern across all of this is the same. Weighting problems don't announce themselves. Nothing errors, the percentages look tidy, and a wrong score simply sits in someone's results page. You catch them by checking the mechanics before launch, not by reading the output afterwards.
Free guide · 28 pagesBuild a scorecard that tells the truth.The Scorecard Playbook: an assessment that separates people and gives you a reason to talk.Get the free guide →A weighting routine for a Monday evening
- Choose one scale. Write it on paper.
- Set equal pillar weights. Change one only if you can say why in a sentence.
- Check every question tops out where its neighbours do. Hunt for the yes/no and the pick-several.
- Check the shortest path. Does the denominator shrink with the questions shown?
- Freeze the options. Then translate, then launch.
Knowing how to weight scorecard questions is one step in the bigger job of how to build a scorecard. Get it right once and you won't think about it again.
Free assessmentHow trapped are you, really?Take the Corporate Suffocation Index, the free solopreneur readiness assessment.Take the assessment →Frequently asked questions
Do pillar weights have to add up to a hundred?
No. They're normalised, divided by whatever they total, so equal weights of any size give the same scorecard. Choose the ratios you believe and ignore the total.
Does adding questions to a pillar make it matter more?
No. The average is taken inside the pillar first, so a pillar with three questions and one with twelve count equally at equal weights.
How do I weight scorecard questions so one counts double?
Double its option values. A per-question weight field usually doesn't exist, or it is saved and never read.
Can I change an option after launch?
Treat it as a new version, not an edit. It can orphan translations, break anyone mid-assessment and change the scale for everyone before them, with nothing recalculating.
The full method, with the checklists, is in The Scorecard Playbook, a free 28-page playbook.
How trapped are you, really?
Take the Corporate Suffocation Index, the free solopreneur readiness assessment.