Making a Build/Buy/Partner Recommendation the Board Can Trust

A persuasive memo wins the meeting. A defensible one survives what comes after it: the director who re-reads it on the flight home, the auditor who asks for the underlying file two quarters later, the LP who wants to know why the firm chose as it did. Persuasion is a property of the presentation; defensibility is a property of the record. The difference shows up the first time someone pulls a thread, because a persuasive memo is optimised for the room while a defensible memo is optimised for scrutiny that does not yet know it is coming.

In practice, a board tests a build/buy/partner recommendation against three properties:

• Traceability of claims: every material finding, from market size to integration risk, can be followed back to a named document, dataset or call, so a challenge in the meeting ends with evidence rather than with reassurance.
• Comparability of options: build, buy and partner are evaluated on the same baseline of assumptions, so the board is choosing between alternatives rather than between a polished case and two caricatures.
• Reproducibility of the record: the analysis, the dissent and the sign-offs are captured in a form that a second team could reconstruct months later without relying on anyone's memory.

A polished argument tends to fail under challenge precisely because it is polished. McKinsey's global survey of 2,207 executives, which examined a recent strategic decision at each respondent's company, found that satisfactory outcomes were associated with less bias, thanks to robust debate, an objective assessment of facts and a realistic assessment of corporate capabilities.^[1] The same survey found that the best practice cited least often, regardless of outcome, was actively seeking evidence contrary to the initial plan. A memo built to persuade has no slot for that evidence, which is exactly where an experienced director probes.

The stakes are higher for this class of decision than for most. In the same survey, merger or acquisition decisions, which by nature tend to be less frequent, were the most likely to be missing elements of good decision making when the outcomes were poor.^[1] Infrequent, high-consequence choices get less practice, so the discipline has to come from the structure of the memo rather than from repetition. That structure is what the rest of this article builds, step by step, and it connects naturally to a broader decision-making framework for moving from analysis to choice.

Board-ready options are not labels. Each of build, buy and partner is a distinct strategic logic with an executable path behind it: a different answer to where the capability comes from, who owns it, how fast it arrives and what it costs to reverse. The option-integrity discipline is simple to state and often violated in practice: if an option would not survive being chosen, it does not belong in the memo. Strawmen and cosmetic variations, included only to make the preferred option look considered, are the fastest way to lose a skeptical room, because directors pattern-match them within minutes. The underlying craft is the same one decision-record practice prescribes for any significant choice: explicitly list all the serious alternatives that were considered, together with their pros and cons.^[2]

Concretely, each option in the memo should carry:

• Its own strategic logic: why this route to the capability is coherent on its own terms, not merely why it is worse than the recommendation.
• An executable path: the sequence of moves, the owners and the realistic timeline from decision to operating capability.
• Resource needs: capital, people and management attention, stated honestly enough that the board can see what the option would displace.
• Risks and failure modes: what breaks if this option is chosen and the assumptions behind it prove wrong.
• Traced evidence: each material claim linked to the document, dataset or expert call it comes from.

Alongside the two to four credible options, the memo needs a baseline of inaction: what happens if the firm does nothing. Without it, the board cannot see the cost of deferral, and deferral is the option every board implicitly holds. Comparability then does the rest. All options should be costed and risked on one shared baseline of assumptions, covering the demand outlook, the discount rate applied, the resourcing costs and the horizon. Where an option genuinely depends on a different assumption, that deviation is made explicit as a named scenario rather than buried in a spreadsheet tab. When each option quietly carries its own assumptions, the debate stops being about evidence and becomes political, because directors argue about the inputs instead of the choice. The same shared-baseline logic underpins how deal teams structure an investment committee memo for partner scrutiny.

The comparison layer is where a recommendation earns or loses its defensibility. It starts with explicit trade-offs. Choosing build means something else stops or slows; choosing buy means accepting integration risk and a valuation that is someone else's mark; choosing partner means accepting dependency and a shared road map the firm does not fully control. A defensible memo states what stops, which resources shift and which risks are accepted, in the body of the argument rather than in an appendix. The test is practical: a memo that avoids these statements produces an agreement that dissolves after the meeting, because each director walks out with a different understanding of what was just approved.

Sensitivity comes next. Most build/buy/partner choices turn on two to four drivers that actually differentiate the options, for example the pace of demand, the scarcity of the talent that build depends on, or the integration cost that buy depends on. The memo should run sensitivity on those drivers and present ranges rather than false precision. A range with a stated source is defensible; a point estimate with three decimal places is a liability, because the first director who asks how it was derived will get an answer that undermines everything else in the deck.

The third element is the skeptic's read: pre-answering the first three objections a skeptical director will raise, and actively surfacing the evidence against the recommended option. This is the practice McKinsey's survey found least often applied, even at companies whose decisions turned out well.^[1] It is also the discipline that separates a memo that invites challenge from one that resents it, a distinction explored further in presenting AI-assisted analysis to a skeptical board. A recommendation that arrives with its own counter-evidence attached changes the tone of the meeting: the board stops hunting for the hidden weakness and starts deciding.

The decision record is the artifact the memo becomes at the moment of choice: a single, structured document that captures not just what was decided but what was known, considered and signed when the decision was made. The form is borrowed from engineering, where it is mature. The UK government's architectural decision record framework prescribes an agreed template in which each record contains a defined set of fields, established to ensure visibility, traceability and strategic alignment.^[3] The template transfers directly to a strategic options decision:

• Title and date: what was decided and when, stated so the record can be filed and retrieved.
• Status: proposed, accepted or superseded, so no one mistakes a draft for a commitment.
• Context: the capability gap, the forces at play and the situation as it was understood at the time.
• Decision: the option chosen, in one unambiguous sentence.
• Consequences: what the choice commits the firm to, including what stops and what is accepted.
• Stakeholders consulted: whose input the decision drew on, and whose dissent it carries.
• Links to supporting documents: the evidence base, attached rather than summarised away.

Broader decision-record practice adds the analytical layer: the assumptions in play, the constraints the decision operated under, the candidate positions considered, the opinions expressed along the way and the rationale for the selection. As with ADRs, it is valuable to explicitly list the serious alternatives together with their pros and cons, because that list is what later readers will want to reconstruct.^[2]

The final required component is the claim-to-source map: every material finding in the memo tied to a named document, dataset or call. This is the norm in professional investing already. A Stanford research brief on investment memos, co-authored with Addepar, found that 97% of the professional investors surveyed have a formal investment memo template and process, and 78% of their firms require memo approval before making any investment.^[4] The same discipline applies to any board-facing options memo, and a working protocol for it is set out in how to fact-check an AI-generated strategy deck.

Freezing means capturing the options, the evidence, the assumptions, the dissent and the sign-offs exactly as they stood when the board committed, and then making the record immutable. The word matters. A living document that gets tidied up after the meeting is not a record; it is a narrative written with hindsight. Decision-record practice is explicit on this point: once a record is accepted it should never be reopened or changed, and documents should not be modified if the decision changes.^[2] What freezing buys is a clean answer, months later, to the question of what the board actually knew at the time.

• The options as presented, including the ones not chosen, with their evidence and economics intact.
• The assumption baseline and the named scenarios that deviated from it.
• The dissent log: who objected, on what grounds, and what would have changed their mind.
• The sign-offs: named, dated human approval at each stage of the review.

When a decision is revisited, the discipline is supersession, not editing. A new record is written that references the old one, states what has changed and why, and takes over as the current decision, while the original stays untouched in the archive. This keeps the history intact and makes the evolution of the strategy legible instead of overwritten. The status field exists for exactly this: a superseded record remains retrievable, linked to its successor.^[2]

Finally, sign-off is named and dated at each stage, rather than attributed at firm level. A record approved by "the deal team" or "the strategy function" gives no one standing to defend the analysis in front of the board, an LP or an auditor. Named sign-off does. It also concentrates care at the moment it matters: an analyst who knows their name will sit next to a market-size claim for the life of the record checks that claim differently.

The first payoff is a reduction in re-litigation. Boards revisit decisions, and should, but a frozen record separates the two questions that otherwise blur together: whether the choice was right, and whether it was made well. When a stakeholder asks why the firm chose as it did, the answer is retrieved, not reconstructed. That protects both the decision and the people who made it, because the debate moves from "what did you know" to "given what was known, does the choice still hold".

The second payoff is learning. Memos are time-stamped artifacts whose assumptions can be tracked as they change, which is precisely what makes them diagnostic; the Stanford research brief notes that the assumptions in a memo are valuable and should be tracked over time in case they change.^[4] The same brief observes that many investors do not track the performance of opportunities they passed on, which creates a loss of information on relative performance and on how decision processes could improve.^[4] A frozen record of a build/buy/partner choice preserves the rejected options in the same way: the firm can later ask whether the buy option it declined would in fact have outperformed, and calibrate the next memo accordingly.

The third payoff is provenance, and it has become a present-tense question rather than a future one. Deloitte's 2025 GenAI in M&A survey of 1,000 corporate and private equity leaders found that 86% of responding organisations have integrated generative AI into their M&A workflows, with 65% of them doing so within the past year.^[5] Boards increasingly know that machine assistance is in the room, so they ask which findings were machine-assisted and what source each traces to. A record with a claim-to-source map answers that question by construction, which is the core of meeting the IC and LP evidence bar for AI-assisted analysis.

To make the structure concrete, consider a clearly illustrative case: a mid-market industrial firm deciding how to acquire a data-analytics capability for its service business. There is no single correct decision for any specific deal, and nothing here is advice on what the firm should choose; the example is about the shape of the record, not the answer. The team writes a memo with three options, each carrying its own logic, resource needs, risks and traced evidence, all compared on one shared assumption baseline:

The board challenge in the meeting is where the structure pays off. The chair asks for the source behind a market-size figure the recommendation leans on. Because every material claim in the memo is mapped to a named document, the team produces the underlying market study on the spot, shows the figure in context and notes the scenario in which the assumption deviates. The challenge ends in thirty seconds and, more importantly, ends in evidence. In a memo without the claim-to-source map, the same question produces a promise to follow up, and the doubt outlives the meeting.

Six months later, the record does its quiet work. A new board member asks why the firm chose to build rather than buy. The archive yields the frozen record: the three options with their economics, the assumption baseline, the dissent log showing the CFO's reservation about hiring timelines and what would have changed her mind, and three named, dated sign-offs. The answer is retrieved in minutes, not reconstructed from memory. This is the workflow Decisity is built for: an AI-native strategy consulting platform that turns a brief and uploaded documents into a board-ready, source-cited deliverable, so that the three-step workflow from ingestion to a traceable deck produces the decision record as a by-product of the analysis rather than as an afterthought.

Related reading in this series: How to Score and Shortlist Acquisition Targets You Can Defend and Making an Acquisition Valuation Defensible: The Price Bridge.

• McKinsey & Company: Flaws in strategic decision making — Global Survey
• Martin Fowler: Architecture Decision Records
• UK Government Digital Service: Architectural Decision Record framework
• Addepar / Stanford Long-Term Investing Initiative: Investment memos and decision-making
• Deloitte: Generative AI in M&A survey