Fig. 1 — Cycle time by station over 90 days. Sample data.
Measure where
work actually waits.
Because every hand-off happens in Kilter, every hand-off is measured. Find the station where time disappears and fix that first.
Cycle time
Medians and tails, not flattering averages.
Track the median and the 85th percentile by station, team and hub. The long tail is where customers lose patience.
p50 and p85 by station
Wait share versus work time
Annotations for process changes
Dashboards
One view for the weekly review.
Build the page your operations review needs — throughput against plan, SLA health and oldest open items — and share it with a link.
Drag-and-drop dashboards
Scheduled email and chat digests
CSV export and warehouse sync
Six numbers,
defined in plain sight.
Every metric in Kilter has a formula anyone can read. Teams argue about what to do next — not about what the number means.
Cycle time
2.6d
completed_at − started_at
Elapsed time from first work to done. Reported as the median and the 85th percentile, never the average.
Wait share
38%
idle_time ÷ cycle_time
How much of the cycle a request spends waiting for someone rather than being worked on.
SLA margin
+41%
budget − time_used
How much time was left when the work finished. Near-misses show up here weeks before breaches.
Throughput
412/wk
items_done ÷ week
Completed items per period, shown against plan so a busy week isn’t mistaken for a good one.
Load
86%
committed_hrs ÷ capacity
Committed work per person or team over the next eight weeks. Above 85% quality starts to slip.
First response
14m
accepted_at − created_at
How quickly a hand-off is acknowledged by its new owner. The earliest signal of a stuck process.
The details
that make it work.
Cycle time
By station, team, hub and period.
SLA margin
How close you came, not just whether you breached.
Throughput
Volume against plan, week over week.
Load
Committed work by person and team.
Filters & segments
Slice by any field in your workflows.
Warehouse export
Snowflake, BigQuery and S3 on a schedule.
A rollout measured
in weeks, not quarters.
The plan our customer engineers use with most teams. Yours will be adapted to your processes.
01
Pick one process
Choose one with a clear start and end.
Week 1
02
Instrument the stations
Kilter measures automatically once work runs through it.
Week 1
03
Review the distribution
Look at the p85, not just the median.
Weeks 2–4
04
Fix the slowest station
Usually Assign. Then measure again.
Month 2
We finally plan from what people are actually doing, not what we remember them doing.
Amara Nwosu
Head of Delivery, Teller & Hart (fictional)
Read customer stories
