How much is certainty worth?

The £4,000 question

Consider a hypothetical manufacturer buying components for a scheduled production run. One supplier quotes £50,000. Another quotes £54,000 and offers reserved production capacity, progress reporting and backup transport.

Purchasing sees a £4,000 premium. Operations sees a possible interruption to the production schedule. Finance needs to know whether the more expensive offer reduces a real exposure or merely describes it more reassuringly.

All three are asking a reasonable question. None can answer it from the quotations alone.

The apparent question is why the business should pay more. The underlying question is what the less predictable option requires the business to do around it, and what the more expensive option actually changes.

Readiness services for artificial intelligence make this purchase increasingly visible. The current PwC US Enterprise AI readiness assessment combines diagnosis, recommendations and a roadmap; its page also describes a route into subsequent advisory work. That is evidence of a commercial offer, not buyer value, uptake or outcomes. The purchasing question is what it is worth to know more before committing. ¹

The question is older and broader than AI. It applies to suppliers, projects, insurance, service agreements and research. Complete certainty is rarely available. More often, a business is deciding which uncertainty to reduce, which consequence to protect itself against and which uncertainty it can afford to retain.

Uncertainty has a bill, even when nothing goes wrong

Return to the manufacturer. An uncertain delivery might require extra stock, a provisional labour arrangement or time spent checking progress. It might also prevent the sales team from making a delivery commitment to its own customer.

Some of that cost is visible. Some sits in another department. Some takes the form of an opportunity the organisation cannot safely accept.

These arrangements are not automatically waste. A buffer may be doing an essential job. Removing it without changing the underlying exposure is not a saving in risk management; it is a decision to carry more risk.

The commercial value of predictability can therefore be larger than a reduction in expected losses. It may allow the business to make a useful commitment it could not otherwise make.

A well-established corporate-finance model describes one mechanism: when external finance is more costly than internal funds, risk management can help preserve the ability to fund worthwhile investment. The mechanism is conditional; it does not establish a universal premium for assurance. ²

For a manager, that suggests a practical starting point: identify the cash, time, capacity and commitments currently held back because of the particular uncertainty. Then distinguish what could genuinely be released from what would still be needed.

Four purchases hiding inside one promise

Before pricing certainty, specify what it concerns: cost, timing, performance, the rules under which an arrangement operates, or recovery if it fails. A firm price can coexist with an uncertain completion date. A dependable service can still create an expensive dependency.

It then helps to separate four things commonly sold under the broad promise of reassurance.

On a narrow screen, each row is shown as a separate card.

The mechanisms can be combined, but none guarantees the whole outcome. The important question is which mechanism justifies which part of the price. A report, an insurance policy and a backup system are not interchangeable simply because all three make the board feel more comfortable. HMRC’s explanatory material likewise distinguishes avoiding, retaining, reducing, sharing and transferring risk. ³

Price the change, not the reassurance

Suppose the manufacturer’s team uses the following assumptions for that single order. The figures are entirely illustrative, not supplier performance data. Both quotations cover the same scope. A late delivery is assumed to create £30,000 of additional economic loss, including disruption and lost contribution after avoided costs.

On a narrow screen, each row is shown as a separate card.

Under those assumptions, paying £4,000 more reduces expected delay losses by £4,500. The higher-priced offer has a £500 lower expected total cost.

That is not yet a pure premium for certainty. It is an expected-cost advantage. The distinction matters: the buyer has not demonstrated a willingness to pay more for the same average outcome. It has identified an option with a better average outcome under the model.

The fragile part is the assumption about Supplier B. If its delay probability is 10%, rather than 5%, its expected total cost becomes £57,000. The £4,000 premium is no longer justified by expected delay losses alone. On these assumptions, Supplier B stops being cheaper on expected total cost once its delay probability exceeds about 6.7%.

That is a better supplier conversation than asking whether the offer is reassuring. What supports the lower failure probability? Does the evidence cover comparable orders, current capacity and the conditions of this contract?

The 2026 Green Book recommends testing how conclusions change when uncertain assumptions move. Although written for public-sector appraisal, that sensitivity discipline is useful here: find the assumption that changes the decision rather than polishing a single forecast. ⁵

In this risk-neutral, matched-scope illustration, £4,500 is the most supported by avoided expected delay losses alone. It is not a universal maximum price for assurance. A business might justify a higher premium because a disruption would threaten liquidity or an important commitment. Those consequences need to be made explicit, with care not to count losses already included in the £30,000.

Expected cost is also not a cash requirement. The manufacturer will not receive a £6,000 delay invoice because the probability was 20%. It faces the actual outcomes. Affordability of the bad outcome remains a separate question.

Do not manufacture savings from the buffers

Once the decision has been framed, several tempting calculations should be resisted.

Releasing a £100,000 cash buffer does not automatically create £100,000 of profit. The benefit may be lower financing cost or the value of deploying that money elsewhere. A contingency allowance is not necessarily expenditure that would otherwise have occurred.

Likewise, freed management hours are not automatically a payroll saving. They become economically useful through a real reduction in cost, a credible alternative use or improved performance. Those benefits can matter without pretending that every recovered hour appears in the accounts as cash.

Do not count the same event twice. If the delay loss already includes emergency freight and lost contribution, those amounts cannot reappear as separate benefits of the more reliable supplier. If a benefit depends on winning additional business, count its incremental contribution and associated costs, not the full sales value.

The comparison should cover the same period and scope on both sides. It should also subtract the cost of obtaining the assurance: fees, implementation, monitoring, minimum commitments and any flexibility surrendered. This is a disciplined comparison, not an invitation to add every reassuring possibility to a benefits column.

An assessment is worth what it can change

Information needs a different pricing question from reliability or insurance.

In standard expected-value analysis, the value of perfect information is the improvement from being able to choose with that information rather than choosing with what is known now. It provides a theoretical upper bound for information about the same decision. A real assessment is imperfect, may arrive late and has its own cost. The established concept does not make an assessment worth a fixed percentage of the project budget. ⁶

Before commissioning one, specify which finding could change the choice, scope, sequence, safeguards or negotiated terms. An assessment that reveals an attractive project to be unworkable may be more valuable than one that reassures the sponsor.

If every plausible finding leads to exactly the same action, its direct decision value is limited. There may still be a legitimate need for documentation, compliance or independent assurance. Name that job separately instead of claiming that more information must improve the investment decision.

This is an important test for AI-readiness products. Does the output identify a constraint the organisation can act on, or only locate it on a maturity scale? Does it support a defensible decision not to proceed? Is any recommendation commercially separable from the provider’s next implementation sale?

Useful research can make a manager less confident in the preferred plan while making the eventual decision better. Confidence and decision quality are not the same outcome.

A workable exit can be worth more than a firmer forecast

Sometimes the cheapest way to proceed responsibly is not to resolve more uncertainty. It is to reduce the commitment exposed to it.

A bounded trial may leave performance uncertain while establishing spending limits, review dates, stop conditions and a tested route back to normal operations. Its value comes partly from the ability to stop or change course. A spending cap does not itself cap liability, data loss or service disruption; those exposures need separate treatment. The Green Book describes real-options analysis as one way to value this flexibility under uncertainty. ⁵

Waiting can also be rational when useful information is likely to arrive and a decision is expensive to reverse. Bloom’s structural model describes a temporary wait-and-see mechanism following uncertainty shocks. It does not mean every delay is wise, or that removing every uncertainty is worth the price. ⁷

There is a corresponding danger in buying certainty too aggressively. A fixed-price contract can shift some cost exposure, but it cannot make an unclear scope clear or a weak supplier financially resilient. Current UK government commercial guidance warns that weak information and inappropriate risk allocation can produce risk premiums, poor pricing and contracts that suppliers struggle to sustain. ⁴

A credible recovery plan therefore needs more than an exit clause. Can the business retrieve its data, transfer the work, use an alternative and afford the transition? Who has authority to stop the arrangement, and what happens while the replacement takes over?

The aim is not to predict failure perfectly. It is to know enough about the response to failure that the remaining uncertainty is acceptable.

The Certainty Premium Test

Use these five questions before paying extra for an assessment, assurance package or more predictable arrangement. This is an Andrew’s Intel management instrument, not a validated scoring scale.

1. Which decision are we trying to improve? Name the outcome, time horizon and current alternative, including proceeding without the purchase, waiting or doing less.
2. What changes if we buy this? Separate better information, more reliable performance, financial protection and flexibility. Specify the uncertainty or consequence each element addresses.
3. What evidence supports that change? Examine relevant performance, assumptions, exclusions and the supplier’s capacity to deliver. Ask what remains uncertain and who still bears it.
4. What is the incremental value to this business? Compare expected losses, necessary buffers, usable capacity and enabled commitments without double-counting. Subtract the premium, ongoing burden and lost options. Keep unpriced benefits visible rather than giving them invented precision.
5. What would reverse the decision? Identify the assumption, result or event that would make the purchase unattractive. Record the tolerable downside, decision owner and review or stop trigger.

Consequential financial, contractual and insurance decisions still need appropriate specialist review.

Enough certainty to make the next commitment

Predictability has no single market price. The same improvement in delivery reliability may be marginal to a business with slack capacity and decisive to one facing a time-critical commitment. The same report may change one buyer’s decision and merely decorate another’s.

The relevant comparison is not what other businesses pay for reassurance. It is what this particular uncertainty costs this business, what the proposed purchase genuinely changes and whether that change makes a better decision possible.

Do not pay for the feeling of certainty. Pay for a specific improvement in the decision you can make.