A baseline is a record of what was agreed, and its entire value is that it does not move quietly. The moment somebody edits the schedule so that it matches what is now happening, the difference between plan and reality vanishes, and every measurement that depended on that difference goes with it.
Why a baseline changes by decision rather than by edit
Variance is the gap between an agreed position and an observed one. Move the agreed position to meet the observation and the gap closes without anything having improved, which makes an unrecorded baseline edit the quiet destruction of the project's measurement rather than an administrative shortcut.
Integrated is the load bearing half of the name. A change that presents itself as scope always lands somewhere else too, so the analysis has to run across scope, schedule, cost, quality, resources and risk together. A request that turns out to cost nothing and take nothing extra usually means the analysis stopped at the first dimension.
The path a request takes
The five dimensions are assessed together, because a change that presents itself as scope lands on the other four as well. On adaptive work the lower lane decides which route a request takes, so reshaping the remaining work stays with the team while anything touching the outer envelope joins the path above.
A change request is submitted by anybody, in writing, describing what is wanted and why. It gets logged whether or not it looks sensible, because the log is the evidence that a decision was taken rather than a preference exercised.
Impact analysis is where the actual work sits. Somebody competent assesses what the change does to scope, to the schedule and the critical path, to cost and the reserves, to the quality characteristics that were agreed, and to the risk position. The output is not an opinion about whether the change is a good idea. It is a statement of what accepting it would cost.
A decision follows, and it is approve, reject or defer. Deferral is a real answer rather than an avoided one, since a change that is right for the next release and wrong for this one is a sequencing decision that happens to be recorded as a refusal today.
Approval then triggers two steps that are routinely forgotten. The affected baselines and plans are updated, and the people whose work has just changed are told. A change approved and never communicated has been paid for and not received.
Who is allowed to say yes
The change control board is the group holding authority over changes that cross agreed thresholds. Its composition follows the money and the risk, so it usually includes the sponsor, the customer or their representative, and whoever will own the operational consequences.
Below those thresholds the project manager decides alone, and the thresholds themselves are set deliberately when the change management plan is written. Small changes that sit inside contingency and inside agreed tolerance are settled there, because routing everything to a committee produces a board that rubber stamps instead of examining. Anything touching the cost baseline, the delivery date, contracted scope or the benefits case belongs to the board however small it looks.
An approved change resets the baseline, and that is the point
Once a change is approved the baseline is reissued to include it. A project that takes on three extra weeks of work and carries on measuring against the old plan reports a variance for the rest of its life that nobody can act on, because the gap being reported is the approved change rather than a delivery problem.
Reissuing restores the meaning of every measurement built on the baseline. It also leaves a record of how the plan grew, which is often a more interesting story than the variance figures. Four approved changes that added twenty per cent to the budget is a governance finding, and it stays invisible while each one is absorbed rather than baselined.
Where adaptive delivery puts the same problem
On adaptive work the boundary moves rather than disappearing. Inside an iteration the team is expected to learn, so reordering, reshaping and dropping items on the strength of what the last increment revealed is the method working as designed. Demanding a change request for that would defeat the reason for iterating at all.
What stays baselined is the outer envelope. Funding, the release commitment given to a customer, contracted scope, regulatory obligations and the architecture the organisation has to live with all still change by decision. Hybrid delivery runs both at once, with an iteration that absorbs change freely sitting inside an envelope that does not, and the common failure is a team treating the envelope as negotiable because the iteration is.