How to Set Scorecard Band Boundaries Without Gaps
How to set scorecard band boundaries so every score lands in exactly one band: write them a hundredth apart, check for gaps, and move the cut points to fit.
Set scorecard band boundaries so that every score from zero to a hundred lands in exactly one band. Matching is usually inclusive at both ends, so write the next band to start one hundredth above the last one ends. Then move the cut points until every band has people in it.
How to set scorecard band boundaries that never lie
This is the dullest part of the whole project. It's also where a scorecard most quietly tells people the wrong thing.
No error appears. No warning. A person finishes, gets a band, and believes it. If the boundary was wrong, they believe a mistake.
There are three separate ways boundaries go bad. Learn all three, because you'll meet each of them within a month of launching.
Failure one: boundaries that touch
Here's the natural way to write a ladder. Zero to thirty. Thirty to fifty. Fifty to seventy. It looks tidy and it's wrong.
Matching is normally inclusive at both ends. So a score of exactly thirty fits in two bands at once. Which one wins? Whichever the system happens to sort first. In other words, by accident.
| How you write it | What happens at the edge | Verdict |
|---|---|---|
| Zero to thirty, then thirty to fifty | A score of thirty matches both. Sort order decides. | Wrong |
| Zero to thirty, then thirty point zero one to fifty | A score of thirty matches one band only. | Right |
| Zero to thirty, then thirty point zero one to fifty, then fifty point zero one to seventy | Every edge lands in exactly one band. | Right, and the habit to keep for scorecard band boundaries |
The best teaching story in the live data makes the point. Someone who scored exactly forty was told their level was "Good" for months. A per-pillar row spanning thirty to eighty-five began one hundredth of a point before the real band at thirty point zero one. It simply matched first. Nothing errored. Nobody noticed.
Write all your scorecard band boundaries a hundredth apart. Every time. It costs you nothing.
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 →Failure two: gaps in scorecard band boundaries
The second failure is the one that frightens me. It happens when a score matches nothing at all.
Three behaviours exist in the wild. The page renders nothing. Or it snaps to the nearest end. Or it hands the person the top result.
The last is genuinely dangerous, and it's more common than you'd expect. A ladder with a gap in it can show no result in the headline and the best result in the recommendations, on the same page. Imagine receiving that. Your own report disagrees with itself, and the generous half is the one you'll remember.
So make it a rule. After every edit, check that your bands tile from zero to a hundred with no gap and no overlap. Not once. Every time. It takes two minutes and it has no downside.
Failure three: cut points that fit your guess, not your people
The first two are mechanical. This one is about judgement.
On paper you pick round numbers. Zero to twenty, twenty to forty, and so on. It feels fair. It's also a guess, made before a single stranger has taken the thing.
Then the first fifty or so completions arrive and eighty per cent of them sit in one level. That is the most common finding in the whole exercise. And the cause is not your respondents. If you designed five levels and most people land in one, the cut points are wrong, not the people.
That's good news. Fixing cut points is arithmetic, not writing. It's the cheapest repair you'll ever make on a scorecard.
The calibration loop for bands
- Read the distribution of results. How many people are in each level?
- Move the boundaries so every level is populated.
- Check the levels still tile cleanly. No gap, no overlap, point zero one apart.
- Look at the distribution again.
Repeat until each band has real people in it. Treat your first edges as a draft. After the first batch of completions, make them final. The full routine for reading all the signals is in how to calibrate a scorecard.
What recalibrating does not do
Here's a caveat that will otherwise confuse you at eleven at night.
Recalibrating usually does not go back and change past respondents. Scores and results are written once, at completion. Nothing recomputes. People who finished before you moved the boundary keep the result they were given.
So the count in your editor can legitimately disagree with a chart that recalculates live. Neither is broken. They're answering different questions.
It also means every boundary change forks the instrument, and nothing records which version somebody took. Your analytics will happily reinterpret old history against the new cut points. The discipline is simple: tune before you drive traffic, batch your changes, date them in your own notes, and never read a trend across an edit.
Boundaries aren't the only thing that forks things. Option values and weights do the same, which is why how to weight scorecard questions belongs in the same sitting.
How many bands, and what each one is for
Every band you add is another row of writing in the grid, for every pillar. Five levels across five pillars is twenty-five cells. Seven levels across eight pillars is fifty-six, and you will abandon it.
So set the number of bands by the number of different things you'd say or do next. Here's the test from the build: write the follow-up for every result. If two bands would get the same email, you have one fewer band than you thought. Merge them and widen the boundary.
A band that sends someone to the same next step as the band above it isn't a band. It's decoration, and you're paying for it in evenings.
Once the edges are right, the label on each band matters just as much, because it travels into the headline, the contact record and your inbox. That's a different skill, covered in how to name scorecard result bands.
Free assessmentHow trapped are you, really?Take the Corporate Suffocation Index, the free solopreneur readiness assessment.Take the assessment →What a good edit looks like
Say you run a productised audit and your five bands came out lopsided: most people stacked in the second level, almost nobody in the fifth. You don't rewrite the questions. You don't blame your audience for being all the same. You slide the cut points until the middle bands are populated and the top band holds the few who earned it.
Then you do the boring check. Does the first band start at zero? Does each next one start a hundredth above the last? Does the top one reach a hundred? Then you look at the distribution once more.
That's a ten-minute repair, and it can transform how credible the scorecard feels. A ladder where everyone is "Developing" tells people nothing. A ladder where they spread out tells them where they stand, which is the only reason they came.
One last constraint. The internal identifier for a level is usually derived from its name once and then frozen. So fix the edges freely while you calibrate, but be careful what you do to the names. Renaming a level later orphans everything keyed to it, silently.
One more trap: per-pillar rows
The "Good" story came from a per-pillar row, not the overall ladder. That's a warning. A scorecard has two verdicts. The overall result says who you are. The per-pillar result says what to do next. Each can have its own ladder, and each ladder needs the same treatment.
Builders get the headline ladder perfect and leave the per-pillar rows sloppy. But the per-pillar result is the deliverable. It's the one people act on. A misplaced edge there sends someone to the wrong advice.
Check every ladder. The overall one, and each pillar's.
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 →Your ten-minute check of scorecard band boundaries
- Do the bands start at zero and finish at a hundred?
- Does each next band begin one hundredth above the last one's end?
- Is there any score that matches nothing?
- Do per-pillar ladders pass the same test?
- Is every band populated after the first fifty or so completions?
- Did I note the date of every change?
Frequently asked questions
Why leave a hundredth between scorecard band boundaries?
Matching is normally inclusive at both ends, so a shared boundary matches two bands and the winner is decided by sort order. Starting the next band one hundredth higher puts each score in exactly one band.
What happens if a score matches no band?
It depends on the tool. It might show nothing, snap to the nearest end, or give the top result. The last is the dangerous one, so check your bands tile from zero to a hundred.
Eighty per cent of my people are in one band. What's wrong?
The cut points, not the respondents. Move the boundaries until every level is populated, then check they still tile cleanly.
Will changing boundaries update old results?
Usually not. Results are written once at completion, so earlier respondents keep what they were given. Date your changes and avoid reading trends across them.
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.