Table of contents
Research context
This protocol asks how long a new offshore engineer takes to make a first production change after starting on a client codebase. The unit is one engineer on one client engagement, followed from the contract start date to the first change that is merged, deployed, and monitored. The study measures elapsed calendar time and the stages inside it. It does not judge the engineer's ability or the quality of the onboarding program.
A first production change is a useful milestone because it combines access, context, review, and release in one event. An engineer can have every account and still be unable to ship because the test environment is broken or the reviewer is unknown. The measure should therefore record the whole path rather than a single gate. That is why the protocol keeps the stages separate and reports them together.
Set the scope to one client, one engineering lane, and one deployment model. Record the repository, the review policy, the access process, and the release cadence at the start. If the client changes the review or access process during the window, keep the original definition for engineers already in progress and analyze the later cohort separately. A policy change can look like faster onboarding when it is only a different gate.
Measure the onboarding work that happens before day one as its own interval. A client that prepares the repository guide, the starter ticket, and the access request before the contract starts is doing work that the engineer should not have to repeat. Record the date each preparation item was completed. The comparison between prepared and unprepared cohorts is often more useful than the overall median.
Define the population as every engineer who starts on the lane during the window and is expected to contribute production code. Include engineers who leave before their first change, because their absence is part of the result. Exclude engineers hired only for incident support or pure analysis, and state that exclusion. Test the unit definition on two recent onboarding records before collection begins.
A first change can be deliberately small, such as a documentation fix or a test-only commit. Record the change type and whether it touched production behavior. A small first change may be the right onboarding choice even though it does not prove the engineer can handle a complex feature. The protocol should not reward a risky first change over a safe one.
Record the start date, the contract date, the first repository access, the first development environment, the first assigned ticket, the first pull request, the first approval, the first deployment, and the end of the first monitoring window. Add the engineer's time zone and the client's review hours, because a short task can still wait a day for overlap. Record whether the first change was a feature, a fix, or a configuration edit. A second reviewer should be able to rebuild the timeline from these fields.
Practical application
Use calendar days for the headline and working days for the stage detail. Calendar days show the customer's experience, while working days show the team's capacity. Report both when the two differ by more than a week. A long calendar gap over a holiday is not the same as a slow review.
Report the median and the range for each stage: access, environment, first ticket, first pull request, first approval, first deployment, and monitoring. Show the number of engineers and the number of calendar days. Present the share of the total time that was controlled by the client and the share controlled by the offshore team. Use the same denominator for every stage rate.
Record whether the engineer had a named onboarding buddy and how often they met. The buddy relationship is hard to measure but easy to describe, and it often explains a fast or slow start. Ask the engineer at the end of the first month which person or document helped most. That answer is qualitative evidence that the timeline alone cannot provide.
Inspect the slowest cases and describe which stage dominated without naming the engineer. A long access stage may point to a missing client owner, while a long first-ticket stage may point to an unclear backlog. Compare the result with the time zone overlap and the review policy. Report engineers who left before a first change separately so their timelines do not distort the stage medians.
Keep the first-change definition stable when the client changes the release cadence. If releases move from weekly to daily, the deployment stage will shrink without any change in onboarding quality. Note the cadence in the scope and analyze the cohorts separately. Otherwise the reader will attribute a process change to the engineer.
The record cannot show whether the engineer understood the code or merely made a small change. It can show only the sequence and duration of the gates. A fast first change is not proof of readiness, and a slow one is not proof of poor onboarding. The protocol also cannot capture informal help that moved the work forward without appearing in the ticket history.
A useful additional view is the number of attempts before the first merge. An engineer who needs three review rounds is having a different experience from one who needs one, even if both ship on the same day. Record the review rounds and the reasons. The count helps the lead decide whether the gap is context, standards, or access.
Operating controls
Give the result to the engineering lead and the client owner and ask them to choose one observable change. The change might be a pre-provisioned developer role, a starter ticket prepared before day one, a named onboarding buddy, or a review slot reserved for new engineers. Record the owner, the effective date, and the measure that will show whether the change worked. Repeat the protocol with the same definitions.
Ask the client reviewer how long the first review took and what they had to explain. The reviewer's time is part of the onboarding cost and rarely appears in the engineer's timeline. Record the minutes spent on explanation and the number of comments that repeated a written guideline. A high explanation cost may point to a missing guide rather than a weak engineer.
Do not compare engineers who joined different lanes without stating the lane. A data platform and a marketing site have different review and release paths. If the sample mixes lanes, report the result by lane and keep the overall figure as context only. The comparison should be like with like.
Close the review by choosing one change and naming the date of the next cohort. A protocol that produces a chart but no decision becomes an annual ritual. The owner should be able to say which stage will change and how the next measurement will show it. Keep the raw timeline so a later reviewer can test a different question.
Record the size and risk of the first change so a fast timeline is not confused with a hard one. A one-line configuration fix and a payment-path feature both count as a first production change, but they carry different review and monitoring needs. Put the change type and the risk tier beside the elapsed time. The owner should compare engineers who faced similar first tasks.
Check whether the engineer was blocked by a missing non-production tool, such as a design file, a test account, or a product contact. These dependencies rarely appear in an access matrix, yet they can add days before the first ticket moves. Record the missing item and the date it arrived. The pattern often reveals a reusable onboarding kit that the client can prepare once.
Ask the client reviewer to record the number of comments that could have been answered by the repository guide. A guide that does not cover local setup or test commands creates review comments that look like engineer mistakes. The count is a cheap measure of documentation quality. Use it to improve the guide before the next cohort starts.
Sources
Frequently asked questions
What should happen first?
Begin with a bounded ticket, approved access, explicit acceptance evidence, and a named reviewer.
Who approves production changes?
The client's technical owner keeps final merge and release authority.
Related research
What share of offshore pull requests are abandoned before merge?
How much of an offshore engineer's week goes to client meetings?
