Measuring the ramp curve of a new outsourced hire

Research findings

A new outsourced hire is often judged in two moments: the first week, when everything is new, and the first month, when the owner expects full output. The space between those moments is where the real learning happens, and it is usually unmeasured. A ramp curve fills that gap by showing how independence grows across a series of defined checkpoints. The goal is to see the shape of the learning, not to force it into a straight line.

This review describes a practical way to measure ramp for an administrative or support role. The population is work items assigned to one new hire during the first ninety days, and the unit is a completed item with its review result. The purpose is to decide when to reduce review, not to rank the person against a hypothetical ideal. The curve is a management tool, not a certification.

Start with the tasks the role will own and the acceptance rule for each. A task is not ready for a ramp measure until a reviewer can say what good looks like without arguing. Use the same examples that were used in training, and keep them available. If the acceptance rule changes mid-ramp, the curve restarts because the target changed.

Define three or four checkpoints. A useful set is shadowing, assisted work, supervised work, and independent work. Each checkpoint should have an exit condition: a sample size, an accuracy threshold, and a response-time expectation. The conditions should be observable in the queue, not based on a manager's impression of confidence.

Measure two things at each checkpoint: the share of items that pass the acceptance rule and the share that required a hint from the reviewer. A hint is any intervention that changes the work without taking it over. Tracking hints separately from errors shows whether the person is learning the task or still depending on someone else to interpret it. Record the hint text in a short list so the pattern can be reviewed later.

Do not use raw throughput as the main ramp measure. A new hire can complete many easy items and avoid the hard ones, which makes the volume look strong. Pair volume with the mix of task types and the pass rate on the harder type. If the hard type is missing from the sample, the ramp is incomplete.

Review cadence matters as much as the measures. Weekly for the first month, then every two weeks, is enough for most roles. The review should look at a small sample of real items and one adverse case. It should produce one change to the training, the tools, or the review rule. A review with no change is a status meeting.

Expect the curve to be uneven. A new hire may improve quickly on routine work and plateau on exceptions, or the reverse. A flat week is not automatically a failure, especially after a process change or a new task type. Look at the trend across checkpoints and the reason for the flat spot before drawing a conclusion.

Discussion

Use the curve to plan the review load. If the pass rate is high and hints are low, reduce the sample rather than removing review entirely. If the pass rate is unstable, keep the sample and add a second reviewer for the hardest task. The goal is a review level that matches the evidence, not a permanent training mode.

The first checkpoint should test orientation, not speed. Can the new hire find the queue, the source, the examples, and the person to ask? A person who cannot locate the inputs will look slow even when the task is understood. Fix access and navigation before measuring output.

Use the same reviewer for the first few checkpoints when possible. A changing reviewer introduces a new interpretation of the acceptance rule and makes the curve harder to read. If a handoff is necessary, have the reviewers compare a small sample before the new one takes over. Write the handoff date in the ramp record so a later reader knows who judged what.

Write the ramp plan before the start date and share it with the new hire. The plan should name the checkpoints, the sample sizes, and the review dates. A plan created after a weak week looks like a response to the person rather than a design for the role. Review the plan at each checkpoint rather than waiting for the final month.

Keep a short note on what the new hire found confusing. The answers are training material for the next hire and evidence for the next process review. A ramp that produces no notes has probably not been examined closely enough. Share the notes with the training owner before the next hire starts.

Implications and limits

Review the ramp with the provider as well as the internal owner. The provider sees attendance, training, and replacement patterns that the client may not, while the client sees the work and the customer impact. Compare both views at the checkpoint and record any disagreement. A shared review is more useful than two separate verdicts.

A ramp plan should also say what happens if a checkpoint is missed. The answer is not automatically to extend the review or replace the hire. Check the input, the training, and the task mix first, then choose a remedy with a date. A written response keeps a difficult week from becoming an improvised decision.

Finally, record what the ramp does not show. It does not measure judgment on cases outside the sample, and it does not predict retention. It also cannot compare two hires who started with different training or different task mixes. Report those limits beside the curve so the reader does not turn a useful management tool into an unsupported verdict.

Choosing checkpoints and exit rules

Write each checkpoint as a condition, not a date. Shadowing might end when the new hire can explain the task and identify the stop rules. Assisted work might end after ten reviewed items with no missed stop rule. A date-based checkpoint tells you how long someone has been present, not what they can do.

Keep the exit sample small enough to review carefully. Ten to twenty items per checkpoint is workable for a small role. Include at least one adverse case so the check tests the hard path. Record the sample date and the reviewer so the decision can be reconstructed later.

Separating hints from errors

Define a hint as a reviewer prompt that changes the work, such as a reminder to check a field or a pointer to the right policy. An error is a completed item that fails the acceptance rule. Track both, because a low error rate with a high hint rate may mean the reviewer is quietly doing the hard part. Use the same definition in every checkpoint so the hint count remains comparable.

Record the hint reason in a short list. A repeated hint about the same field points to a training gap, while scattered hints point to confidence or task mix. Use the pattern to change the training material before changing the review level. The new hire should see the same list so the feedback is consistent.

Using the curve to change review

Reduce review in steps. Move from every item to a sample, then from a daily sample to a weekly one, only when the evidence supports it. Keep a full review for high-risk tasks such as money, access, or regulated records. The curve should change the sample size, not the standard.

If the curve stalls, check the inputs before the person. A missing example, a slow approval, or a tool that changed can flatten the line. Give the new hire a chance to work with the repaired input before deciding that the fit is wrong. Record the repair date so the next review compares like with like.

Limits of ramp measurement

A ramp curve is local. It reflects one role, one training set, and one review group. It should not be used to compare providers or to set a universal hiring bar. The same task in a different queue may produce a different curve for reasons that have nothing to do with the person.

The curve also says little about long-term retention or judgment on novel cases. Use it to decide when to reduce review and where to improve training. Pair it with regular conversation and a broader performance review for the decisions it cannot support. State that boundary in the ramp summary so the reader does not overread the line.

Questions