Eight concierges at North Energy, a residential solar company, tracked their own performance in a shared spreadsheet. Calls answered, texts sent, projects moved, installs closed, days from sale to install. Every week, for every person.
Nobody saw it rendered back. A concierge could post a record month and hear nothing about it for a month, or never. Leadership had the opposite problem. They had the whole grid and no fast way to read it, so the team got judged on a monthly scroll and a gut call. Nobody had checked whether the numbers in the grid were even right.
What we built
A scoreboard on the office wall that reads the same spreadsheet the concierges already fill in, updates every five minutes, and asks nothing new of anyone.
The floor sees a live competition: rankings, a shared monthly goal the whole team moves together, badges with a printed guide for earning each one, and records that carry across months. Leadership sees the same figures without the game layered on top. Each concierge also gets a card with their own personal best, so someone who isn't winning the leaderboard still has something of their own to beat.
The board checks the source numbers before it trusts them, and says so when something looks wrong. Nothing about how the concierges enter data changed. The spreadsheet is still the spreadsheet. It just got a mirror.
What changed
For the floor, performance used to be invisible to the people producing it. Now it's on the wall within five minutes of anyone touching a cell, with the scoring formula printed underneath so it can be argued with.
For leadership, a monthly scroll through a grid became a glance. Same numbers, same source, no separate report to assemble by hand.
The board also caught something nobody had been checking for. On its first run, three of one month's metrics (texts sent, texts received, and answered calls) turned out to be the previous month's numbers, copied row for row and never overwritten. A formula elsewhere in the sheet was returning a -83.33% answer rate. Those were the figures leadership had been using to judge the team.
A missing entry never gets quietly turned into a zero either. On a leaderboard with real names attached, that would put a real person in last place for something they didn't do.
The math, shown so you can check it
Feedback went from roughly 30 days to 5 minutes. That's a structural change, not an estimate. A monthly review cycle became a display that re-reads the source every five minutes, so there's no cycle left to wait on.
Assembling a performance picture from the raw grid used to cost 45 to 60 minutes a week, more if anyone wanted it fresher than the start of the week. The board replaced that with a glance. Multiply by however many people on your team were doing that assembly by hand.
We won't claim the scoreboard raised install volume. Nobody measured that, and we'd rather say so than model a number we don't have. What we can say is that the team's totals now get verified against the spreadsheet's own independent sums, with 62 automated checks running on every change. That's how three copied months and a broken formula got caught on the first run.
Why self-reported numbers rot
Nobody at North Energy was lying about their numbers. They were busy, and a spreadsheet that nobody reads back doesn't get proofread. That's true of any team reporting its own metrics into a system nobody else looks at. The numbers drift. A copy-paste error survives for months. A broken formula keeps returning a wrong answer until someone happens to check the arithmetic by hand.
The fix isn't a stricter policy on data entry. North Energy already built an AI agent for a different version of this problem: staff hunting through 73 files on a shared drive for an answer that should have taken seconds. The scoreboard is the same move pointed at a different symptom. Instead of asking people to be more careful with a spreadsheet, put a system between the raw numbers and the people reading them, one that checks its own work before it shows anybody anything.
That's the pattern behind most Level 99 builds. Catch what a busy team doesn't have time to check, and show it before anyone has to ask. Plans start at LVL 10.
The full case study has the complete build and the rest of the math.
FAQ
Is this a performance review tool or a reporting fix?
Both, though it started as the second one. The leaderboard and badges are the version the floor sees. Underneath that, the actual fix is a system that checks the numbers before anyone acts on them, instead of trusting whatever a spreadsheet says.
Did the team's workflow change?
No. The concierges still enter data the same way they always did. The scoreboard reads the existing spreadsheet. Nobody was asked to use a new tool or fill in new fields.
How did it catch the copied metrics?
The board runs automated checks against the spreadsheet's own independent sums every time the data changes. When a number doesn't reconcile, or a formula produces something impossible like a negative answer rate, the board surfaces it instead of displaying it as fact.
Could this work without the leaderboard and badges?
Yes. The verification layer and the leadership view don't depend on the game elements. The badges and rankings are what makes the floor want to look at it. A team that doesn't want a leaderboard can still get the same checked numbers in a plain dashboard.