Granting somebody the power to refuse finished work looks like a control. Granting it without a condition attached is a decision to pay for late change.
Every mature governance framework contains an approval register: the parties who must sign before a deliverable is accepted. Building one is routine, usually delegated, usually done in an afternoon. It looks like the safest document in the pack.
It is also where most enterprises quietly commit to a cost they never price. A register naming who may refuse finished work, silent on what those people were required to examine while it was being made, has arranged for the initiative's most expensive interventions to be made by the people least informed about it.
Michael Greer states the corrective in one sentence, in a chapter of the Handbook of Human Performance Technology (Jossey-Bass, 1999), excerpted on a time-tracking software vendor's site: anyone with the power to reject or demand revision of deliverables after they are complete must be required to examine and approve them as they are being built. It has been available for twenty-five years and is almost never implemented — because doing so requires somebody senior to accept a constraint on their own authority [SOURCE DETAILS REQUIRED].
The Strategic Context
The economics are not in dispute. A change requested while a deliverable is still forming is cheap; the same change after completion is expensive, and the gap widens the further the work has travelled. That relationship is argued elsewhere and treated here as established [Related article: The Front End Owns the Outcome].
What follows is less often stated. If late change is expensive, a right exercisable only at the end is an expensive right — and the enterprise, not the holder, pays for it. The person exercising it bears no cost: they read the finished thing, form a view, and object. The rework, the schedule, the renegotiated contract and the demoralised team sit elsewhere on the balance sheet.
An authority whose cost falls entirely on someone else is exercised more freely than one whose cost is felt. That is not a claim about character but about how a control behaves when its price is invisible to the person operating it — and it holds for careful people as reliably as careless ones.
What an Approval Register Does Not Record
Look at a typical register and notice the absence. It names parties, names the artefact each must approve, and sometimes names a date. It does not record what each party is obliged to have seen before that date, when, or what happens if they did not. The obligation half of the arrangement is not a field.
Greer supplies the missing half specifically. Sponsors and stakeholders, he argues, are active participants rather than passive customers, and the responsibility accompanying the authority to approve has three parts: taking part in the early stages, helping define what will be built; completing reviews of interim deliverables in a timely fashion, which it frames as keeping the project moving; and helping expedite the team's access to subject-matter experts, target users and essential documentation [SOURCE DETAILS REQUIRED].
Read those three as governance rather than advice and they become testable. Each is an event that happened or did not; each has a date; each can be recorded beside the right-holder's name — which is precisely what an approval register omits.
He also argues that deliverables should evolve in successive approximations, with incremental reviews and approvals along the way. That is usually read as a delivery method. It is also what makes the obligation enforceable: without staged artefacts there is nothing for an approver to have examined, and the requirement becomes an exhortation.
Reframing the Issue
Stop treating an approval right as a permission and start treating it as a position with terms. A permission is granted once and held indefinitely; a position has duties, and the interesting question about a position is what happens when they are not discharged. Nearly every enterprise has answered that for its delivery roles and left it unanswered for its approving ones.
That asymmetry has a documented other half. The delivery leader typically carries full accountability for an outcome with almost no structural power to secure it — a mismatch argued elsewhere as a design choice rather than an accident [Related article: Accountability Without Authority: How Organisations Design Delivery Leadership to Fail]. This article is the same mismatch reversed: authority held without an obligation attached. An enterprise that has read one and not the other has half the picture.
The Three Settlements
If an approval right is a position with terms, the only question giving those terms force is what follows when they go unmet. Three settlements are available, and the value lies in choosing one deliberately rather than discovering later that the default chose for you.
The right narrows. The holder may still object, but only within a stated class — safety, regulatory compliance, a defect against a written specification — and not on preference, scope or approach that an interim review would have surfaced. Most enterprises would accept this if asked in advance.
The right transfers. It passes to a named party who did engage: a delegate, a technical authority, a peer who attended. This suits arrangements where the holder is senior, busy and formally required, and somebody else can carry the interest competently.
The right stands unchanged, and the enterprise records that it has chosen to accept late change from an unengaged approver, with an estimate of what that is worth. Some approvals are held by parties an enterprise cannot condition — a regulator, a joint-venture partner, a majority customer — and pretending otherwise is worse than pricing it.
Which is chosen matters less than making the choice before the deliverable is finished, by people who do not yet know whose right will be tested.
What This Instrument Is Not
Three boundaries, because each neighbours a different argument in this collection and the distinctions are fine.
It does not decide what a review may change or stop. Which class of defect a screen can catch, and who must be present for that entitlement to be real, is a question about the review event, answered elsewhere [Related article: Quality Assurance Cannot Tell You the Specification Was Wrong]. Nothing here concerns what happens inside a review; the subject is a right held by someone who may never attend one.
It does not specify what an approver should examine. Naming the events you will attend, the artefacts you will see and who attends on your behalf — and recognising that depth of visibility is worthless without capability to use it — is a separate decision with its own instrument [Related article: Visibility You Cannot Use]. This instrument uses that one rather than restating it, and supplies what visibility design leaves out: the consequence of declining the visibility you were offered.
It does not map influence. Reading a stakeholder set by exposure and capacity to withhold is a different exercise [Related article: Your Stakeholder Map Measures Attention, Not Exposure]. What is enumerated here is narrower: the holders of a formal right of refusal, some with very little influence. That is why the register misses the problem.
One adjacency: who may redefine what the initiative is for is a different population from who may refuse a deliverable, and the two lists repay comparison [Related article: Drivers, Supporters and Observers: Who Is Allowed to Change What Your Program Is For].
Two Handovers
Consider, hypothetically, a tollway concession approaching handover of a completed section. Three parties hold a right to refuse: the concessionaire, a state road authority and an independent certifier. The certifier has been present throughout. The road authority, short-staffed for a year, attended two of eleven design reviews and neither commissioning trial. At handover it objects to the gantry signage standard — an objection that would have cost a design revision eighteen months earlier and now costs a re-fabrication and three months. Nobody behaved badly, and the authority is right on the merits. The arrangement simply had no term saying what an unattended approver may object to.
The same shape appears in an agribusiness grain terminal operator's receival-and-outloading upgrade, hypothetically, where the right of refusal sits with an operations manager also running a harvest. He signs nothing during the build because he cannot spare the days, then rejects the commissioning handover on shift ergonomics — a real problem, invisible from a drawing, that an hour in a workshop would have surfaced. Here the useful settlement is the second: the right should have transferred months earlier to a shift supervisor who could attend. Nobody proposed it, because nobody had written down that the right had a condition.
Decision Framework
The veto conditions register, written alongside the approval register and by the same authority. For each party who may reject or demand revision at completion, three fields:
- the settlement that applies if they have not engaged — narrows, transfers or stands;
- who determines whether the condition was met, which must not be the right-holder;
- when that determination is made — early enough for a transfer to be possible, meaning at the first interim review the holder misses, not at handover.
The veto census. Who can reject or demand rework at completion? Build the list from what happens rather than from the register, and expect it to be longer; informal vetoes — a party whose objection nobody will overrule — belong on it.
The conditioning test. How many of those rights carry any condition at all, and who wrote it? In most enterprises the honest answer is none, and discovering that is the point.
The lapse test. What has happened when an approver who missed every interim review rejected a finished deliverable — and what would we have wanted to happen? The second half is the design question, far easier to answer about a past case than a live one.
From Strategy to Execution
Immediately. Take one initiative in delivery, complete the veto census, then ask for each name what they have actually seen. A morning's work, producing a list of unconditioned rights that is usually uncomfortable and always concrete.
Over the next two quarters. Add the three fields to the approval register for new initiatives, and make the determination point an event rather than a policy: the first missed interim review triggers a written decision about the settlement, by someone other than the holder. Pair it with the delivery-side discipline Greer insists on — work evolving in reviewable increments — because a condition on an approver is unenforceable if nothing was staged for them to review.
Over years. The durable change is that approval rights stop being distributed as recognition — granted for seniority, for politeness, to signal that a function matters. Once each carries a term the register becomes a statement about who is genuinely accountable, and it will shorten. Expect resistance, and read it as evidence the instrument is working.
One honest complication belongs here. Greer states fourteen principles, flatly, as a route to project success. An unattributed course handout in the same body of material is incompatible with that confidence: it says classic project management theory ignores the fact that some projects will not succeed however well they are defined, planned and implemented — and, of its own recommended sequence, that its author does not maintain the actions will ensure success [FACT CHECK REQUIRED] [SOURCE DETAILS REQUIRED]. Both can be held. A reciprocity rule does not make an initiative succeed; it makes the cost of a late refusal visible to the person holding the right, in advance. That is a narrower claim than the source's own.
Signals to Monitor
- The length of the veto census against the approval register. A large gap means the formal document does not describe the real control structure.
- Attendance at interim reviews by parties holding a right of refusal, tracked by name rather than function.
- The share of late change requests originating from approvers rather than the delivery team or the environment.
- Whether any right has ever narrowed or transferred because a condition went unmet. If none has, the terms are decorative.
- The interval between an approver's first missed review and anyone noticing — the practical measure of whether the determination point exists.
- Whether interim deliverables exist to be reviewed at all, since an enterprise producing nothing until completion cannot condition anything, and its sign-off is a pledge made without evidence [Related article: Most of Your Plan Is Somebody Else's Promise].
- Whether the reasons for a late refusal are ever examined. A veto that lands, causes rework and is never diagnosed is a repeatable event [Related article: The Review That Cannot Ask Why].
Questions for the Leadership Team
- Who can actually refuse a finished deliverable on our largest initiative, and is that list longer than the approval register?
- For each of them, what were they required to have examined first, and where is it written?
- What happens today when someone who saw nothing rejects something at the end?
- Which approval rights were granted for seniority or courtesy rather than accountability?
- Do we produce interim deliverables an approver could meaningfully review, or does everything arrive finished?
- If we conditioned every right of refusal next quarter, who would object — and what would that tell us?
Closing Perspective
The asymmetry is easy to miss because both halves look reasonable alone. Requiring senior sign-off on finished work is prudent. Not burdening senior people with interim reviews is considerate. Together they produce an arrangement in which the enterprise's most costly interventions are made by the people with the least context, when context is most expensive to acquire.
The correction is not to remove anybody's right. It is to notice that a right and a duty were separated at some point, probably by nobody in particular, and to put them back — or, where that is impossible, to say what the separation costs and accept it as a decision rather than inherit it as a default.
One question tests whether an enterprise has done this, and it can be asked of any initiative in flight: for each person who can stop this at the end, what did we require them to look at, and what happens if they didn't? Most organisations cannot answer. Those that can have usually shortened the list [Related article: Sponsorship Is an Office, Not an Endorsement], and found the shortening cost them nothing they wanted to keep. What then travels to the rest of the enterprise is a separate question [Related article: Written Down Is Not Passed On].
About EraNorth Insights
EraNorth Insights publishes practical analysis on strategy, projects, operations, transformation and decision intelligence for professional and organisational use. About EraNorth.