Velocity Drift — catch cliffs in your team velocity before they become trends
Velocity is the metric every VP asks about and no team trusts. Avium Signals tells you when it's actually shifting — and whether the shift is structural or noise.
Why velocity charts mislead
A velocity chart in Jira looks like a row of bars. They jiggle up and down. Some weeks are higher, some lower. Most teams have been told to ignore the jiggle, watch the rolling average. Fine — until the rolling average drops 15% over four sprints and nobody noticed because each individual sprint felt explainable.
The inverse problem also bites: a velocity drop from sprint 14 to sprint 15 looks dramatic on the chart, panic ensues, and a month later you realize it was holiday-week noise. The data didn't actually shift.
How agile teams try to read velocity manually
Reading a velocity chart by eye is a guess wearing a confidence costume. The common mistakes:
- Looking at the most recent two sprints, declaring a trend. Two data points isn't a trend.
- Ignoring variance — a team that ships 30/40/30/40/30 is steady; a team that ships 35/35/35/35/35 is also steady. The mean is the same; one is volatile and one is predictable. Burndown charts treat them identically.
- Calling a cliff (sprint 14 = 38, sprint 15 = 12) a 'bad sprint' instead of investigating: was it scope, was it people-out, was it a critical bug that consumed the team?
- Confusing capacity changes with velocity changes — half the team went on PTO, of course velocity dropped, but Jira's chart doesn't account for that.
How Avium Signals computes velocity drift
Velocity Drift is a deterministic signal — pure math on your sprint history, no AI. Avium computes it from the last 6 closed sprints (configurable) and classifies the pattern into one of four buckets:
- Cliff: the most recent sprint completed less than 65% of the period's peak. A sharp, recent drop — the most demo-able 'something just broke' signal.
- Trend rising: the second-half mean of the window is more than 12% higher than the first-half mean. Capacity is growing or process is sharpening.
- Trend falling: the second-half mean is more than 12% below the first-half mean. Different from a cliff — a gradual decline, often because attrition or scope shift has been masked by sprint-to-sprint variance.
- Volatile: high coefficient of variation (standard deviation / mean) above 0.45. The team's output is unpredictable enough that planning around an average is a mistake.
- Stable: none of the above. A team that ships within ±12% sprint-to-sprint is doing the hard work right.
Who reads this Signal
See velocity drift on your own sprints
Avium reads your last 6 closed sprints from Jira (or any work management tool, on whatever cadence you've configured) and surfaces the drift classification on the Insights page within seconds of import. Free tier shows the Signal; the AI Briefing that contextualizes it into a plain-English narrative ('your team has been volatile for two months; here's the likely cause') is on Business.
Start free — no credit card →