Change control run exactly as prescribed destroys the record of every commitment the enterprise has broken, at a rate proportional to how much change it approves, and the only quantitative input the approving body receives is a forecast computed on a baseline its own decision will invalidate.
Change control asks an enterprise to do two things that cannot both be done. The first is stated as the most frequently skipped step in the whole process: once a change is approved the baselines must be formally reissued, the new schedule superseding the old and the new cost baseline replacing the previous one, because otherwise every variance report compares today against an obsolete plan and means nothing. The second is stated with equal force: reissuing a baseline erases accumulated variance history and makes performance measurement meaningless, and a baseline reissued monthly is not a baseline but a moving average.
Both statements are correct. The mandated action and the forbidden action are the same action. The discipline that holds both positions offers exactly one line between them, and the line is motive. Reissuing the baseline is legitimate when it follows a genuine scope change and illegitimate when it is done to cover an overrun. Motive is not an observable property of a change record. No control system reads intent, no auditor can separate the two from the artefacts, and the artefacts are identical in either case.
Then there is the rate. A large engineering programme may log several hundred change requests in a quarter. Reissue after every approved one and the reference is replaced continuously, so the enterprise's memory of what it promised is destroyed fastest on the programmes where the promise mattered most. This is not a failure of compliance. It is what compliance produces.
The decision itself carries the same defect. The only quantitative input the approving body receives is an impact assessment whose cost line is a revised forecast at completion, and that forecast is computed from performance measured against the baseline the approval is about to supersede. The board decides on a number that its own decision voids, and once the new baseline issues the number cannot be reproduced, because the performance history it was derived from no longer exists.
The Strategic Context
Consider a hypothetical seed and crop science business developing a new trait through to commercial launch. Breeding cycles cannot be compressed, dossier work runs on external timetables, and the launch is anchored to a planting window that comes once a year. Change requests arrive continuously from field results, from regulators and from the commercial team, each properly raised, assessed across scope, schedule, cost, quality, resources and risk, and approved by a board doing its job carefully.
After three years the programme reports on plan. It is on plan, against the plan it is executing. The launch window has moved through two planting seasons, and no document in the reporting pack states that, because each move was absorbed into a reissued baseline and the reissued baseline became the reference for the next report. The variance was real every time and survived nowhere.
The same shape appears in a hypothetical government land titles registry replacing its core system, where the divergence is visible from outside because a commitment date was stated publicly and public statements are not reissued. The internal baseline moves with each approved change; the external commitment does not. The enterprise holds two records, one maintained meticulously and one not held at all, and it will be judged against the one it did not keep.
What Leaders Misread as Discipline
The first misreading is that the change log preserves the history. The reason it does not is structural rather than administrative. A change log records requests, dispositions and impact assessed at the time. It records deltas. It does not record outcomes and it does not accumulate. After two hundred approved changes an enterprise holds two hundred authorised deltas, each measured against whatever baseline was current when it was raised, and no document stating what was originally promised or how far the programme has departed from it. Reconstructing that means replaying every delta against superseded baselines that were not retained. Nobody does it, and on most programmes nobody could.
The second misreading is that delegated authority is a minor convenience. Changes below a stated cost and duration threshold are approved without convening the board, which is sensible and necessary. Erosion staying below the threshold therefore never reaches the approving body, while erosion above it reaches the body one item at a time. The aggregate is invisible in both directions, and the aggregate is the only number describing the enterprise's actual position.
The third is that the performance indices measure performance. After a reissue they measure performance since the reissue. Every re-baseline resets them to par, which means a programme's reported efficiency is a statement about the period since it last admitted it was wrong. At portfolio level this produces a result that should trouble any board: a portfolio approving a high volume of change will report consistently strong performance and deliver consistently poor outcomes, and both readings are accurate.
Reframing the Issue
The useful reframing is to stop treating the baseline as one artefact serving two purposes. It is being asked to be the measurement reference for control and the evidentiary record of commitment at the same time, and those two roles have opposite requirements. Control needs a reference that tracks reality closely enough to generate actionable variance, which means it must move. Accountability needs a reference that never moves, because a promise that is edited after the fact is not a promise. One document cannot do both, and the discipline's insistence that it must is the source of the contradiction.
Separate the roles and the contradiction dissolves. Keep reissuing the measurement baseline exactly as prescribed, because that instruction is sound. Stop asking it to carry the commitment record, and put the commitment somewhere it cannot be overwritten. That costs discipline rather than money, and few enterprises do it, because the commitment record has no owner and no home.
The levels behave differently here. On a project, a reissued baseline is a local loss of memory. On a programme it becomes a systematic one, because component projects re-baseline at different times and the programme reference is a composite of freshly reset parts. At portfolio level it becomes an institutional incapacity: the enterprise loses the only dataset from which it could ever learn how far its own estimates habitually miss. How a range estimate becomes a single committed number in the first place is a separate mechanism, treated elsewhere in this collection under the estimate-to-commitment conversion; this article begins at the moment the commitment exists.
What the Reissue Actually Deletes
The variance history is the estimating record
The accumulated variance against an original plan is not merely a scorecard. It is the only empirical record an enterprise possesses of how wrong it usually is, by category of work, by supplier, by estimating method and by team. Re-baselining deletes it as a matter of routine, which severs the one feedback loop that could improve future commitments. This is why enterprises with decades of delivery history estimate no better than they did at the start. The information was generated every time and destroyed every time, in accordance with the procedure.
The forecast that justified the decision cannot be reproduced
The impact assessment is the only quantitative basis the approving body ever has. Its forecast at completion is derived from performance indices measured against a baseline that the approval supersedes. Once the reissue occurs, the indices reset and the derivation is unreproducible, so the decision cannot be audited on its own terms afterwards. Whether the approval was sound becomes a matter of recollection. Whether the plan being amended contained all the work in the first place is a prior question this article assumes away, and it belongs to [Related article: Nothing Is Authorised to Test Whether the Plan Is Complete].
The cumulative departure is recorded nowhere
Each change is assessed against the current baseline, so each looks proportionate. Nothing in the process computes the running total against origin, and no artefact is designed to hold it. This is a different failure from four baselines coexisting at one moment and disagreeing about the same project, which is the subject of [Related article: Four Baselines, One Project, No Referee]. Here there is only one baseline at any moment, it is internally consistent, and it has quietly become a description of a different undertaking from the one that was funded.
Decision Framework
The original-commitment archive is a record created once, at the moment a commitment becomes external, and never superseded. It is not the measurement baseline and must never be used as one. It is held outside the planning system by someone who does not report to the programme.
| Contents | Written when | Altered by |
|---|---|---|
| Commitment set: deliverables, completion date, authorised cost, named performance obligations | Contract signature, board approval or funding release | Never |
| Basis: decomposition version, estimating method, contingency and its rationale | Same moment | Never |
| The individuals who made and who accepted the commitment | Same moment | Never |
| Running departure from origin, in scope, date and cost, after each approved change | At each approval | Appended only |
| The forecast shown to the approving body at each decision, retained beside what followed | At each approval | Appended only |
Every reporting pack then carries two frames rather than one. Current-baseline performance answers how delivery is running against the plan being executed. Origin variance answers how far the enterprise has now departed from what it promised. Both are published on the same page, and neither is permitted to stand alone.
The thresholds convert this from a record into a control. Set a cumulative origin variance limit in advance, expressed as a percentage of the original authorised cost and a number of weeks against the original date. Beneath the limit, changes are approved as changes. Above it, the next request is not a change request at all: it triggers re-authorisation of the whole undertaking by the body that authorised it originally, on a fresh business case, with the option of cancellation genuinely on the table. That single rule is what stops a hundred proportionate decisions from delivering an outcome nobody would have approved as a proposal.
The one-day diagnostic: ask for a single page stating the original commitment date, the original authorised cost, and the cumulative departure from each as at today. If it cannot be produced within a day, the archive does not exist, whatever the change log holds.
From Strategy to Execution
Immediate. For the three largest programmes in delivery, reconstruct the origin figures now, from the contract or the approving minute rather than from the planning system, and publish the cumulative departure alongside the current-baseline report at the next meeting. Reconstruction will be painful once and impossible later, so the cost of doing it falls every month it is not deferred.
Medium term. Make archive creation a condition of funding release, so no commitment becomes external without one. Set the origin variance limits at the same moment, in the authorising minute, before anyone knows which project will breach them. Require every change approval to state the cumulative position, not just the increment, and require the approving body to record it.
Long term. Feed the archive into two places that currently receive nothing. The closure review should compare origin, final baseline and outcome, and the resulting departure figures should become an input to the next estimate for work of that type, held as institutional data rather than in the memory of whoever was there. The archive is denominated in the same accrual terms as the report it accompanies, and the distinct problem that the constraint which actually ends an enterprise is denominated in cash belongs to [Related article: Accrual Governs the Report; Cash Governs the Outcome].
Signals to Monitor
Nobody can state the original commitment date without opening a contract. Performance indices sit near par on a programme widely understood to be late. The proportion of approved changes rises in the quarter before a reporting milestone. Superseded baselines are not retained, or are retained without the date they were replaced. The change log records assessed impact but no outcome. Programmes report on plan for years while their delivery dates move. No approval record in the last year states a cumulative position. Delegated approvals are not aggregated anywhere. The closure review compares the outcome only with the final baseline.
Questions for the Leadership Team
- For our largest programme, what was the original committed date and authorised cost, and how far have we now departed from each in total?
- How many baselines has that programme had, and can we produce every superseded one with the date it was replaced?
- When the board last approved a major change, what forecast was it shown, and can that forecast be reproduced today from surviving data?
- What proportion of changes approved in the last year were approved under delegated authority, and what do they total?
- At what cumulative departure from origin would this enterprise re-authorise a programme rather than approve another change, and where is that threshold written?
- What does our estimating data say about how far our commitments of this type have historically missed, and if we have none, what happened to it?
Closing Perspective
An enterprise that follows this procedure faithfully will arrive at a position where it is performing well against a plan it wrote recently, in pursuit of an undertaking materially different from the one it funded, with no surviving record of the difference and no individual who can be said to have concealed anything. Every step was authorised. Every reissue was correct. The record was destroyed by the control, not in spite of it.
What a board carries here is narrow and unavoidable. Either it commissions a record of the original promise that no approved change can overwrite, and sets in advance the departure at which the whole undertaking returns for re-authorisation, or it accepts that its accountability horizon is the length of one baseline. That horizon is currently measured in weeks on the programmes that matter most, and nobody has ever been asked to agree to it.
About EraNorth Insights
EraNorth Insights publishes practical analysis on strategy, projects, operations, transformation and decision intelligence for professional and organisational use. About EraNorth.