Deal design lens
Risk
Risk is the possibility that the transaction produces a materially worse outcome than expected, together with the severity of that outcome and the question of who bears it.
As a design lens the question is narrower and more useful: which uncertainty or downside is preventing the other side from accepting — or would make the resulting deal unacceptable to you — and can that risk be reduced, evidenced, bounded, sequenced, insured, shared, or reallocated?
Three things that are constantly confused.
Risk and Credibility
Credibility asks whether they believe the Promise will hold in their circumstances. Risk asks what adverse outcomes remain possible, how bad they are, and who carries them. A pilot raises Credibility by producing evidence. A termination right does not make success more likely — it makes being wrong cheaper. A guarantee can reduce risk without making anyone more confident the thing will work. Different problems, different moves.
Risk and Burden
Burden is the effort and disruption expected even if the plan holds. Risk is what lies beyond the plan. A six-month migration is burden. The possibility that it takes eighteen months, or never completes, is risk.
Risk and Commitment
A long term magnifies exposure because exit is harder, but the length itself is Commitment, not risk. A short cancellable pilot lowers exposure without necessarily improving expected performance. The two interact on almost every deal and still respond to different design moves.
Actual exposure and their estimate of it.
Actual risk
The underlying exposure as best anyone can understand it. Changed by structure: scope, sequencing, controls, redundancy, allocation.
Perceived risk
Their estimate of that exposure, formed with incomplete information and their own history of things that went badly. Changed by evidence.
You usually have to work on both, and neither can be measured with any precision. A deal can stall because the exposure is genuinely too large, or because they cannot yet tell that it is not. Those call for different responses, and treating an information problem as a structural one is expensive.
Where risk tends to live.
A prompt list, not a taxonomy. The productive question is never "have we enumerated all risks" — it is which one is blocking this transaction.
Likelihood and consequence are not the same input.
Risk depends on how likely something is and on how bad it would be. A rare but catastrophic outcome can dominate a decision that also contains a dozen frequent minor ones. A regulated data exposure, a public failure, an outage in a revenue-critical process — any of these can decide the deal on their own, regardless of how improbable everyone agrees they are.
Different people also care about different failure modes, which is why "the risk is low" so often fails to persuade. It answers a question nobody in the room was asking.
Three different things you can do about it.
Reduce the risk
Change likelihood or severity: narrower initial scope, staged rollout, better implementation resourcing, redundancy, controls, data validation, a technical change that removes the failure mode entirely.
Generate information
Pilots, prototypes, diligence, testing, references, acceptance criteria, milestones. The underlying probability may not move at all — but the decision is made with far less uncertainty attached to it.
Reallocate consequences
Warranties, guarantees, indemnities, service credits, caps, insurance, escrow, termination rights, refunds, contingent payments. These change who pays when things go wrong, not whether they do.
Principle
Do not confuse making failure less likely with making failure less expensive.
Both are legitimate design responses. But if the real obstacle is a stakeholder who cannot survive the failure, a stronger warranty may help where more reassurance will not — and if the obstacle is genuine doubt about whether it works, an indemnity is not an answer to it.
Risk tends to belong with control.
A useful heuristic: a risk sits best with the party best able to prevent, control, monitor, insure, diversify, or price it. If one side controls uptime, data quality, approvals, or the implementation steps, allocation can follow that control — and the allocation itself creates the incentive to manage it well.
This is a design heuristic, not a rule of fairness or law. Sometimes the party best able to control a risk cannot afford to carry it. Sometimes accepting a risk you do not control is the cheapest way to unlock a transaction worth far more than the exposure. The heuristic tells you where to look, not what to sign.
Information is a purchasable good.
A pilot, phased deployment, milestone, test, audit, prototype, sample, diligence exercise, or field-limited initial licence all do the same job: they produce knowledge before the large bet is placed. Frequently the cheapest available move in a stalled deal is not a concession but a smaller first step that resolves the question everyone is arguing about.
Distinguish elapsed time from informative time. Six months of waiting for internal consensus produces nothing. Six weeks of a scoped pilot against agreed criteria produces the one fact that decides the transaction. Slow is not the same as careful.
The design question is usually: can we learn cheaply before we bet heavily?
Downside is distributed. Upside often is not.
Individuals inside the other side carry specific failure modes personally while sharing little of the gain. That asymmetry blocks transactions that look plainly attractive at the enterprise level, and it is invisible unless you read their Decision Context.
Security
Owns the breach, shares none of the productivity gain.
IT
Owns implementation failure and the support load that follows.
Legal
Owns liability and precedent.
Procurement
Owns variance against benchmark and the awkward renewal.
Finance
Owns the budget overrun and the write-off.
Executive sponsor
Owns the business outcome and the visible reversal.
Designing for this usually means giving one specific person a defensible position: a completed security package, a contractual cap, a rollback path, a pilot they can point to. Not more enthusiasm about the outcome.
Risk allocation has a price.
Every risk you absorb has economic consequence — in expected cost, capital, insurance, capacity, or precedent for the next transaction. Taking more of it may reasonably come with higher price, tighter scope, more control over how the work is done, longer commitment, or a cap somewhere else.
That does not mean every risk concession needs direct compensation. Taking on a risk you can bear cheaply — because you have done it a hundred times, because you can diversify it, because you control the thing that would cause it — can be one of the highest-value deal moves available. It costs you little and removes the obstacle entirely.
Risk-oriented design moves.
Each does a specific job. Choose by which problem you actually have, not by which clause is easiest to add.
Paid pilot with success criteria
Buys information cheaply and defines in advance what counts as working. Being paid matters: it keeps the exercise commercially serious on both sides.
Phased rollout with an expansion gate
Caps the exposure of any single phase and creates a natural decision point where new evidence exists.
Milestone payments
Aligns cash with demonstrated progress. Reduces their financial exposure and signals that you expect to earn it.
Acceptance criteria
Turns an argument about whether it worked into a test agreed before anyone was invested in the answer.
Service credits and remedies
Bound the consequence of routine failures. Useful, but rarely large enough to compensate a real business loss — they are a signal as much as a remedy.
Liability caps and insurance
Price and bound tail exposure. The right question is who can carry or diversify the tail most cheaply, not who can be argued into it.
Escrow and reversion
Protects against counterparty failure or a stalled asset. Particularly relevant where value depends on continued access to something you do not control.
Rollback and portability
Reduces the cost of being wrong. Often more persuasive than an assurance that nothing will go wrong.
Named dependencies
Most implementation failures trace to an assumption nobody wrote down. Naming them converts hidden risk into scheduled work.
Field limits and staged exclusivity
Contains the strategic downside of a rights deal while leaving room to widen once the arrangement has proven itself.
Example · Enterprise software
The return is compelling. They still will not sign.
Everyone agrees the expected return is strong. What the other side actually fears is a failed implementation: money spent, a year lost, a visible reversal, and someone carrying the blame. The offer on the table asks for a three-year enterprise commitment before any of that uncertainty is resolved.
The alternative structure: a paid pilot with agreed success criteria, phased rollout, payments tied to milestones, and an expansion gate at the end of the first phase.
Credibility
Rises through information — the pilot produces evidence in their environment rather than arguments about someone else's.
Risk
Falls twice: less exposure per phase, and a real decision point before the large bet.
Commitment
Smaller and later. This is usually the change that makes the deal signable.
Timing
First value arrives sooner; full economics arrive later, which you have to be able to live with.
Burden
Front-loaded but narrower — one scoped exercise instead of an enterprise programme.
Value Capture
Lower and slower year-one revenue, higher probability of any revenue, and a defined path to expansion earned on evidence.
Worth separating the two kinds of move inside that package. The pilot, success criteria, and expansion gate generate information. The milestone payments mostly shift downside. Both are useful; only the first changes what anyone knows.
Example · Technology licence
Demanding a broad indemnity is not a risk design.
You want rights to technology and you are worried about several things at once: whether the IP is valid, whether it performs outside the owner's own environment, whether a third party will claim it, and whether it integrates with what you already run. The reflex is to demand a broad indemnity. The owner refuses, and the transaction stalls on a clause rather than on the underlying question.
A better set of responses addresses each worry with the instrument suited to it. Technical diligence and a validation milestone address performance. A field-limited initial licence contains the strategic downside while you learn. Milestone payments keep your cash exposure proportionate to what has been demonstrated. Escrow and reversion protect against the asset stalling or the owner failing. A capped indemnity on third-party claims addresses the one exposure the owner is genuinely better placed to know about, at a size they can actually accept.
None of this removes the uncertainty. It makes the remaining uncertainty small enough, known enough, and fairly enough distributed that both sides can proceed.
Architectural note
Credibility owns belief in the Promise. Risk owns the residual uncertainty, the consequences, and who carries them.
Risk is kept here as a stress-test lens rather than as a claim about the fundamental dimensions of a deal. Separating it from Credibility removes duplication that existed in the earlier six-dimensional treatment and makes the available moves clearer: evidence for belief, structure for exposure, allocation for consequence.
Risk audit
Run this against a live offer.
- 01What specific bad outcome is each side worried about?
- 02Is the problem disbelief in the Promise, actual downside exposure, or both?
- 03What here is expected burden, and what is genuinely uncertain risk?
- 04Which risks are low-probability but severe enough to dominate the decision?
- 05Which party controls, prevents, or monitors each material risk best?
- 06Can we reduce the risk itself?
- 07Can we generate information before a larger commitment is made?
- 08Can we bound or reallocate the consequences more cheaply than we can prevent them?
- 09What does taking more risk cost us in economics, control, insurance, precedent, or capacity?
- 10Does the revised structure make the offer easier to accept while remaining worth doing for us?
Next
Risk can be tolerable and the answer still no.
What often remains is that the other side is being asked to surrender too much future optionality, too early, on too little information. That is Commitment. You can also work a whole transaction end to end with the Deal Geometry Map.