Four baselines are produced by four processes and no instrument in the control system is built to detect that they disagree, so an enterprise can hold four internally correct pictures of one project and not know when it finishes.
The surest way for an enterprise to lose track of when a project will finish is to measure it four times, carefully, using four properly constituted disciplines. A project with a single plan has one answer, and everyone knows how much to trust it. A project with a scope baseline, a schedule baseline, a cost baseline and a risk baseline has four answers, owned by four functions and reviewed by four sets of reviewers. Those answers will not agree, and none of the four reviews is designed to find that out.
This is visible with unusual clarity in the material the discipline uses to teach itself. Take one small product development project through the standard sequence and each step produces a duration. The critical path yields one figure. The summed activity list yields a longer one, because it silently includes work the network runs in parallel. The stated start and finish dates imply a third, being calendar days rather than working days. The three-point estimate yields a fourth, with a variance and a probability attached. The cost baseline is then spread over a number of weeks matching none of them. Two documents give different durations to the same activity identifiers, and one puts a marketing stream on the driving chain while another gives it float and calls it non-critical.
None of this is caught, and the reason is structural rather than careless. Each artefact is correct by its own method, and there is no fifth artefact whose correctness is defined by whether the other four agree. When one document is rebuilt to fix an earlier error, the probabilistic duration computed on the superseded network is carried across unchanged and presented as the corrected value. Recomputation stops at the edge of the artefact being corrected, because nothing owns the join.
For an enterprise the consequence is not untidiness. It is that the number sent to a customer, a lender, a regulator or a board is whichever of the four reached the reporting template first, and that no procedure exists capable of reporting that the other three disagreed with it.
The Strategic Context
Consider a hypothetical container terminal operator commissioning an automated stacking yard alongside a berth extension. Engineering holds scope as deliverables and interfaces, planning holds the schedule as a network with float, finance holds cost as a time-phased curve tied to a drawdown, and assurance holds risk as exposures with dates. Each is audited within its own discipline.
The commercially binding commitment is none of these. It is the berthing window sold to a shipping line, which sets vessel calls, labour rosters, slot bookings and liquidated damages. That window comes from a date, and the date comes from whichever baseline the commercial team consulted. If the network says one thing and the cost curve is drawn over a shorter period, the operator has sold a service it has funded on a different assumption, and every later report will be internally consistent with whichever assumption its author inherited.
The equivalent in an aged residential care provider building a new wing is sharper, because the dates are populated with people. Scope is written around accreditation-relevant deliverables, cost around a capital drawdown, schedule around a staffing ramp recruited months ahead, risk around clinical governance. The occupancy date given to families is a fifth number, taken from whichever of the four the executive last saw. Nobody lies. Everyone is working from an authorised document.
What Leaders Misread as an Integration Problem
The common reading is that this is a tooling problem, solved by putting all four baselines into one planning system with linked modules. That reading is wrong, and expensively so.
An integrated planning system enforces referential integrity: every work package identifier in the scope structure also exists in the schedule and the cost ledger, no orphans survive, and a change propagates to every table referencing it. That is a real benefit, and not what leaders assume they bought.
Referential integrity is not agreement. The system can confirm that a work package exists in all four baselines while the four hold four different durations for it. It reports a clean integration status, because integration status in these products means the joins resolve, not that the answers reconcile. The enterprise reads that indicator as evidence its four pictures match, when all it evidences is that they use the same vocabulary.
The second misreading is that assurance covers it. Assurance is commissioned by discipline. A reviewer who finds the schedule internally sound reports it sound, and is neither resourced nor mandated to open the cost baseline to see whether it was built over the same period. Four clean reviews are not one clean review.
Reframing the Issue
The useful reframing is to stop treating the four baselines as four documents and start treating the space between them as an asset that has no owner.
Every enterprise can name the owner of its scope baseline, its schedule, its budget and its risk register. Very few can name the owner of the relationship between the schedule and the budget. That relationship is where the real commitment lives, and it has no cadence, no tolerance, no authority and nowhere for a finding to land.
This is the difference between a rule and a control. The rule is stated everywhere: every element of scope must appear in the schedule, and every scheduled element in the budget. It is asserted as non-negotiable and never assigned. A rule with no owner, no cadence, no threshold and no consequence is not a control but a sentiment, honoured as often as someone feels like honouring it.
The distinction bites differently at each level. On a project, divergence is an error a capable manager may catch. On a programme it becomes a systematic bias, because the baselines are produced by functions with different incentives, reporting lines and definitions of a good month. At portfolio level it compounds: the aggregate schedule sums schedule baselines, the aggregate cost sums cost baselines, and the two describe different portfolios. Leaders reconcile them by moving whichever moves more easily, almost always the schedule, because moving it costs nothing until the day it does.
How Four Correct Answers Become One Wrong Commitment
The join has no owner, so nobody is wrong
When the four baselines disagree, no individual has failed. The scheduler produced a defensible network; the cost controller produced a defensible curve. The divergence belongs to the relationship, and relationships do not appear on position descriptions. This is why the problem survives capable people, strong process discipline and repeated assurance: no role is measured on whether the four answers reconcile, so no role loses anything when they do not.
Correction does not propagate
The more damaging property is that fixing one baseline does not fix the others, and the enterprise usually believes it has. When a network is rebuilt because it was found unrealistically linear, figures derived from the old network stay in circulation, often reproduced inside the corrected document itself, carrying the authority of a corrected artefact while describing a plan that no longer exists. Any enterprise that has re-planned and reissued a pack should assume this has happened to it. How fine the decomposition is set, and therefore how quickly divergence can surface at all, is a separate question this article does not take up; it belongs to [Related article: The Rule of Thumb Behind Your Detection Time].
The reconciled number is the one with a customer attached
The four baselines are not equal in force. One has become an external commitment, and that one is the reference the others are quietly adjusted toward. In the terminal case the berthing window will not move, so the schedule is compressed on paper to match it while cost and risk still describe the uncompressed project. The enterprise now holds a fifth picture, the only one it is contractually bound to, and no baseline supports it. Where a substitution made during that compression reopens conformance evidence issued against an assembly as tested, the reconciliation becomes a technical evidence question rather than a controls question, and this article leaves that to [Related article: The Certificate Describes an Assembly, Not a Product].
Decision Framework
The reconciliation owner test is a governance test, not an analysis. It runs in one session and returns a binary result per join. Four baselines create six joins, and each passes only if all five of the test's questions already have a written answer.
| Join | Unit of comparison | Suggested tolerance |
|---|---|---|
| Scope ↔ schedule | Elements present in one, absent in the other | Zero |
| Scope ↔ cost | Elements with no cost line; cost lines with no element | Zero |
| Schedule ↔ cost | Period covered by the network versus by the curve | One reporting period |
| Schedule ↔ risk | Dated exposures falling outside the current network | Zero |
| Scope ↔ risk | Scope changes since the last risk review | Zero |
| Cost ↔ risk | Contingency drawn against exposures not in the register | Zero |
The five questions are: who is the named individual accountable for this comparison, not the function; on what event or date does it run; what difference constitutes a breach, in the unit of the join; what is that person authorised to stop, delay or refuse when tolerance is breached; and who receives the result in writing, by name.
Any join without a named individual is unrefereed, whatever the planning system reports. An enterprise that cannot name an owner for the scope-to-schedule and schedule-to-cost joins has no integrated baseline and should not present a single completion date as though it did.
The field version takes an hour. Ask the four baseline owners separately, in writing, on the same morning and before any of them confer, for the forecast completion date and the figure each used to derive it. The spread is the finding. If the replies are identical, ask each where the number came from; if three cite the fourth, the enterprise has one baseline and three copies, which is a worse problem.
From Strategy to Execution
Immediate. Run the field version on the two largest projects in the portfolio this month and publish the spread, not the average, to the body that receives the monthly report. Add a provenance line to every schedule figure shown to the board, naming the baseline that produced it and the date it was computed. That costs nothing and makes the next divergence visible as it appears.
Medium term. Assign the six joins to named individuals. The reconciliation owner should not be a committee; on most programmes it is the controls lead, given authority to withhold a monthly report until the joins reconcile or the exceptions are stated on its face. Attach the cadence to authorisation events: no funding release, no gate approval, no external commitment without a current reconciliation. Require every reissued artefact to state which figures were recomputed and which carried forward.
Long term. Move reconciliation from a check to a precondition: no commitment date issued, no capital drawn and no availability undertaking signed unless the joins are inside tolerance that day. That is a decision about the authority chain rather than about tooling, and it will be resisted because it slows the issuing of dates, which is its value. This article does not examine why the obligation texts underpinning the risk baseline are so often unavailable to the people holding it; that is the subject of [Related article: Whose Product Are Your Obligations?].
Signals to Monitor
The four completion dates are never shown on one page and no one can say who would put them there. A figure in a board pack carries no provenance and no computation date. A reissued plan reuses a derived number from the version it replaced. The planning system reports full integration and no check has been run on what the identifiers actually agree about. The forecast finish moves when the reporting owner changes, with no change to the work. A work stream carries float in one document and sits on the driving chain in another. The external commitment date has held for nine months while the schedule moved three times.
Questions for the Leadership Team
- For our largest programme, what are the four completion dates implied by the scope, schedule, cost and risk baselines as they stand today, and who has ever seen all four together?
- Which named individual is accountable for the comparison between the schedule baseline and the cost baseline, and what were they authorised to stop the last time those two disagreed?
- When we last re-planned a major project, which derived figures in the reissued pack were recomputed on the new plan and which were carried forward from the old one?
- What period does our cost curve cover, what period does our network cover, and by how many weeks do they differ?
- Which of our external commitments were set from a baseline figure, and which were set from a date that no baseline currently supports?
- If the four baseline owners were asked separately for the forecast finish date tomorrow morning, what would the spread be, and what would we do with the answer?
Closing Perspective
An enterprise that runs four baselines without a referee has not bought four views of one project. It has bought four projects that share a name, at four times the cost of one. The disciplines are not the problem; they work. The absence is, and it is an absence the enterprise chose by never assigning the join to anybody.
The choice sits with the board and is narrower than it looks. Either someone is named, empowered and measured on whether the four answers reconcile, with authority to withhold a commitment until they do, or the enterprise accepts that it does not know when its projects finish and stops publishing a single date as though it does. Publishing one number while maintaining four is not a controls weakness. It is a representation, made by the people who signed the report, about a fact they had no instrument to establish.
About EraNorth Insights
EraNorth Insights publishes practical analysis on strategy, projects, operations, transformation and decision intelligence for professional and organisational use. About EraNorth.