Turn Client Research into Recommendations

Convert research into a decision the client can understand and act on.

This skill owns the reasoning layer between evidence and advice: synthesizing findings, defining credible options, evaluating tradeoffs, and producing a defensible recommendation.

It does not replace the research workflows responsible for collecting, extracting, validating, or structuring the underlying evidence.

When the user or organization already has an established consulting, strategy, or decision-making method, follow that method unless explicitly instructed otherwise.

1. Frame the decision

Before analyzing the evidence, establish the decision being supported.

Identify:

• the client or internal stakeholder;
• the decision or question;
• the intended audience;
• the decision owner;
• relevant stakeholders and reviewers;
• why the decision matters;
• the stakes and consequences;
• the relevant time horizon;
• constraints;
• accepted decision criteria;
• deadline;
• required or preferred output format.

Clarify the desired decision support.

Determine whether the user needs:

• a specific recommendation;
• help selecting among existing options;
• new strategic options;
• a neutral comparison of alternatives;
• a recommendation contingent on certain conditions;
• identification of the next evidence required before deciding.

Establish what is already known and what additional information could realistically change the decision.

If the request is primarily to collect, map, extract, classify, or compare raw information, route that work through the appropriate research workflow rather than forcing incomplete research into a recommendation.

2. Select the appropriate evidence path

Begin with the strongest existing engagement-specific evidence.

Relevant material may include:

• the accepted engagement brief;
• previous research;
• interview notes and transcripts;
• meeting records;
• client-provided data;
• quantitative datasets;
• source links;
• calculations;
• working files;
• visible browser material;
• authenticated sources;
• connected applications;
• approved client-specific context.

Maintain strict separation between clients.

Never mix one client's:

• source material;
• calculations;
• commercial information;
• confidential data;
• permissions;
• artifacts;
• assumptions;
• learned context

with another client's work.

Reuse specialist research instead of duplicating it

Route missing evidence to the appropriate research capability.

Examples include:

• market mapping for landscapes, categories, taxonomies, and relationship maps;
• structured web extraction when the same fields must be collected across many entities or rows;
• company research when a sourced brief about one organization is required;
• account research when the company must be understood as part of a commercial relationship;
• presentation production when accepted analysis must become a visual client readout.

When those capabilities exist as installed skills, use the available environment's corresponding skill rather than assuming a particular namespace.

Combine research outputs with permitted material from authenticated sessions, files, meetings, and connected sources when doing so materially improves the analysis.

Keep supporting evidence inspectable.

Control research depth

Before launching substantial new research, determine:

• what information is missing;
• whether it could change the decision;
• the most useful sources;
• required depth;
• expected effort, time, or cost when material;
• an appropriate first review point.

Do not launch a broad research exercise when the existing evidence is already sufficient to support the decision.

3. Build a trustworthy evidence base

Organize evidence around the decision—not around the sequence in which information was discovered.

Separate evidence into clear categories.

Observed evidence

Directly supported information such as:

• measured data;
• documented events;
• externally verifiable facts;
• source-backed market information;
• system records.

Client-provided statements

Information supplied by the client, stakeholders, employees, customers, or interview participants.

Keep attribution when it affects interpretation.

Calculations

Show:

• source inputs;
• definitions;
• formulas or methodology;
• assumptions;
• transformations;
• relevant units and time periods.

Important calculations should be reproducible.

Analytical interpretation

Clearly identify conclusions, patterns, hypotheses, or implications inferred from the evidence.

Never disguise inference as observation.

Evidence quality checks

Preserve:

• source links or references;
• relevant publication or observation dates;
• definitions;
• units;
• calculation provenance;
• important methodological notes.

Identify:

• stale evidence;
• contradictory evidence;
• inconsistent definitions;
• missing information;
• questionable source quality;
• incomplete samples;
• methodological limitations;
• material uncertainty.

Reconcile definitions before comparing values that may have been measured differently.

When classification, scoring, research depth, or output structure depends on subjective judgment, review a representative sample with the user or responsible reviewer before scaling the approach.

Ask only for missing information whose answer could materially affect the available options, recommendation, risk assessment, or confidence.

4. Translate evidence into decision-relevant findings

Do more than summarize the research.

Identify:

• recurring patterns;
• material differences;
• causal or operational relationships supported by evidence;
• constraints;
• opportunities;
• risks;
• decision-relevant anomalies;
• implications for the client.

Separate interesting findings from findings that actually affect the decision.

For each major finding, make the reasoning path inspectable:

Evidence → Interpretation → Implication → Decision relevance

Where the relationship is uncertain, say so.

Do not imply causation from correlation unless the evidence supports it.

5. Develop credible options

Construct realistic alternatives that the client could actually pursue.

Each material option should describe, where relevant:

• what the option involves;
• intended outcome;
• key actions;
• requirements;
• dependencies;
• expected benefits;
• costs or resource implications when supported;
• implementation difficulty;
• risks;
• time horizon;
• reversibility;
• important uncertainties.

Avoid artificial option sets created only to make one preferred answer appear superior.

Include the strongest credible alternative to the recommended path.

When maintaining the current approach is a realistic choice, consider including a status-quo or no-action option so its consequences can be evaluated explicitly.

6. Evaluate the options

Compare credible options against the accepted decision criteria.

Possible criteria may include:

• strategic fit;
• customer impact;
• financial impact;
• implementation effort;
• speed;
• feasibility;
• organizational capability;
• operational complexity;
• risk;
• reversibility;
• regulatory or legal exposure;
• evidence strength.

Use only criteria appropriate to the actual decision.

When weighting or scoring criteria, make the methodology visible. Do not create arbitrary precision.

If numerical scoring would falsely imply certainty, use qualitative comparison instead.

Document important tradeoffs rather than hiding them inside a final score.

7. Form the recommendation

Recommend a path when the evidence supports one.

Explain:

• the recommended option;
• why it best fits the decision criteria;
• the strongest supporting evidence;
• expected benefits;
• important tradeoffs;
• risks;
• dependencies;
• implementation conditions;
• remaining uncertainty;
• confidence level where useful.

Include the strongest alternative and explain:

• why it was not selected;
• when it could become preferable;
• what evidence, event, constraint, or change would cause the recommendation to be reconsidered.

Never:

• convert a client assertion into an established fact;
• hide a logical leap;
• invent evidence;
• fabricate calculations;
• manufacture precision;
• imply certainty beyond the evidence;
• guarantee an outcome that has not been established.

When the evidence is insufficient for a strong recommendation, provide the most useful bounded output instead.

This may be:

• a partial recommendation;
• conditional recommendations;
• a narrowed option set;
• explicit decision thresholds;
• the highest-value next research step;
• a recommendation to defer the decision until a specific uncertainty is resolved.

Uncertainty is preferable to manufactured confidence.

8. Create the decision-ready artifact

Propose the artifact structure before investing in a polished final version when the structure materially affects the work.

A useful default structure is:

1. Executive recommendation
2. Decision to be made
3. Decision criteria
4. Key evidence
5. Decision-relevant findings
6. Credible options
7. Option comparison
8. Recommended path
9. Rationale
10. Risks and tradeoffs
11. Assumptions and uncertainties
12. Conditions that would change the recommendation
13. Next steps
14. Sources and analytical notes

Adapt this structure to the decision rather than following it mechanically.

Select the simplest useful format

Use a memo when the client needs a detailed written recommendation or an artifact that can be reviewed asynchronously.

Use a presentation when the recommendation will be discussed in a client readout or benefits materially from visual storytelling. Route production through the available presentation capability.

Use a workshop artifact when stakeholders need to examine evidence, challenge assumptions, evaluate options, or reach alignment together.

Use a supporting spreadsheet or analytical model when comparisons, scenarios, financial analysis, scoring, or calculations require transparent inputs, formulas, assumptions, and source notes.

For large research exercises, use structured datasets or CSV files as the evidence layer. Do not treat the raw dataset as a substitute for the recommendation or decision artifact.

Visual polish must never conceal weak evidence, unresolved uncertainty, questionable methodology, or unsupported calculations.

9. Stress-test the recommendation

Before finalizing, challenge the analysis.

Check:

• every material claim against its source;
• whether calculations can be reproduced;
• whether evidence actually supports the stated conclusion;
• whether important counterevidence exists;
• whether alternative interpretations are credible;
• whether definitions are consistent;
• whether source material is current enough;
• whether implementation is feasible;
• whether dependencies are realistic;
• whether client language and context are represented accurately;
• whether important constraints were omitted.

Ask:

• What would have to be true for this recommendation to fail?
• What evidence most strongly contradicts it?
• Which assumption carries the most risk?
• What change would make the alternative preferable?
• Which conclusion has the weakest evidence?
• What information could materially reverse the recommendation?

Surface unresolved decisions rather than burying them.

Recommend resolutions when the evidence supports doing so.

10. Review sharing and action boundaries

Before recommending or performing any sharing action, establish:

• audience;
• access model;
• destination;
• confidentiality requirements;
• source permissions;
• treatment of personal data;
• suitability of screenshots, examples, or visual evidence.

Treat the following as distinct approval states:

1. analysis drafted;
2. analysis internally reviewed;
3. recommendation approved internally;
4. approved for client delivery;
5. delivered to the client;
6. approved for broader sharing;
7. published or externally distributed;
8. approved for implementation;
9. implementation initiated;
10. related systems or CRM records changed.

Approval of the analysis or recommendation does not automatically authorize any later state.

Do not treat client delivery, public sharing, sending, publishing, CRM changes, or implementation as implied by approval of the analytical artifact.

11. Preserve the reusable method

After the approach has proven useful, offer to preserve the reusable decision method as a custom skill, template, or analytical framework.

Reusable components may include:

• framing questions;
• source strategy;
• evidence standards;
• evidence-quality checks;
• decision criteria;
• option structure;
• scoring or comparison method;
• recommendation format;
• artifact structure;
• writing voice;
• review behavior;
• stress-test questions.

For reusable team methods, remove client-specific:

• facts;
• confidential evidence;
• calculations;
• permissions;
• artifacts;
• identifying information;
• engagement restrictions

unless explicit access has been agreed for the intended audience and destination.

12. Use recurring refreshes only when appropriate

Recommendation work is normally decision-driven rather than scheduled.

A draft-first recurring workflow may refresh a recommendation when all of the following are stable:

• the underlying decision;
• authoritative source set;
• evidence definitions;
• cadence or trigger;
• output format;
• review owner;
• destination;
• permissions;
• notification behavior;
• approval boundaries.

A recurring refresh must stop and request review when:

• a material source becomes unavailable;
• important sources conflict;
• permissions or access change;
• a critical calculation cannot be reproduced;
• evidence definitions materially change;
• a key assumption fails;
• the client's underlying decision changes;
• the recommendation falls outside the approved engagement scope.

Recurring analysis should remain draft-first unless explicit authority exists for further action.