Monthly attrition from about 14% to under 3%

The knot

Over the years I’ve hired, onboarded and trained 100+ people. Attrition on a 100+ contractor team costs more than it looks like on paper. Every departure means recruiting again, ramp-up time, a coverage gap while the replacement learns the work, and a schedule that’s a little less stable than it was.

When I checked when people were leaving, a large share had only done a few shifts. In 2024, half of everyone who left did so within their first 90 days on the project. Three things were behind it:

Recruiting didn’t have a clear picture of who we needed.
Training shared too much information at once.
Once they were live, new people felt lost on the project.

So the fix had to happen before and during the first 90 days.

What I did

Gave recruiting a clear list of what we needed in a candidate.
Sent every new person a welcome email with clear objectives, and rebuilt training as interactive slides with quizzes.
Split the team into newcomers and veterans. For their first 90 days, newcomers get more attention: more QA reviews and more 1:1s. The glide path below has the details.
Built a risk-assessment framework. At the end of shadowing, each new person is rated on five signals: timeliness, attendance, tools set up on time, engagement, and an end-of-training knowledge check. The higher the risk, the more support they get. The full method is in the risk-rating guide.
Redesigned onboarding, initial training and continued learning so the material is spread over a longer period instead of front-loaded. The core of it is five main topics that cover around 80% of contacts; everything else comes later, once people are steady.

The glide path

Later I turned the onboarding into a glide path with five stages: training, shadowing, newcomer, veteran review, and veteran.

Every stage has a deadline counted from the first day on the project. Training done by day 30, shadowing done by day 45, working live by day 50, and a veteran review around day 80 to 90. A slow start shows up in the first weeks instead of the third month.
KPI targets are the same from day one. What ramps up is autonomy, support and accountability. Newcomers see their real numbers from the start and get five QA reviews a week, a weekly 1:1 and coaching, but no performance warnings until they’re veterans.
Conduct and reliability rules apply in full from day one. The glide path only covers performance.
A weekly “falling behind” report flags anyone who misses a milestone, so their manager steps in with coaching rather than a warning.
Nobody becomes a veteran automatically. A review decides it, looking at QA trend, attendance, conduct and KPI direction, and can move someone up early or hold them at newcomer for longer.

What changed

Monthly attrition went from about 14% in early 2024 to under 3% in 2026.

People also stayed longer before leaving. In 2024, half of the people who left did so within their first 90 days, and 11 left within their first month. In 2026, it was one in five within 90 days, and nobody left within a month. The typical leaver had been on the project for 88 days in 2024, and for more than a year in 2026.

Ramp-up costs came down with it, and the schedule stopped needing weekly repairs. The onboarding program became the basis for company-wide policy.

What I’d tell someone doing the same

Look at when people leave before spending on retention. If the answer is “in the first 90 days”, the fix is in hiring and onboarding, and it’s much cheaper than a retention program.