Evidence is organized so its support—and its limits—remain visible.

Define the question first

Useful research begins with a bounded service, customer segment, U.S. geography, current stage, and decision question. A vague or conflicting scope is clarified or narrowed before deeper work.

The question determines which evidence is relevant. We first identify the decision the customer is trying to inform, then select sources that can address it; collecting more information is not a substitute for defining the question.

Preserve source context

• —Source name and URL or identifier
• —Retrieval date
• —Reference period or year
• —Geography
• —Limitation or confidence note where relevant

How Harborfield selects and uses data sources

Compare like with like

Comparability depends on service scope, geography, reference period, unit, population or category, and pricing basis where relevant. A fixed project price, an hourly rate, and a starting price may describe materially different work.

Differences are recorded before combining observations. A credible source may still describe a population too broad for the question; that observation can provide context without supporting a direct comparison.

Separate evidence from assumptions

Customer-provided assumptions are labeled as assumptions. Public observations remain source-linked. Financial figures, where an approved product includes them, are illustrative planning scenarios rather than forecasts or promises.

• —Evidence is information observed in an identifiable source, interpreted within that source's limits.
• —An assumption is a stated condition used when direct evidence is unavailable or when a scenario requires it.
• —A calculation is a deterministic transformation of defined inputs. Calculating from an assumption does not turn it into evidence.

Evidence, assumptions, and decisions

Use deterministic controls

Eligibility, rejection, product access, prices, scores, calculations, payment and order states, QA requirements, and delivery authorization belong to deterministic server controls. Narrative tools cannot override them.

Do not fill gaps with speculation

When evidence is missing, stale, contradictory, or non-comparable, the conclusion is narrowed, the limitation is marked, more structured information is requested, or the unsupported portion is not presented.

Unavailable evidence is not treated as zero. Older observations retain their reference period. Incomplete observations are identified, and conflicting sources are examined for differences in definitions, periods, and coverage rather than silently choosing the most favorable result.

What market research cannot tell you

Review before publication and delivery

Harborfield Business Consulting LLC is responsible for its public materials. Public descriptions must remain consistent with the defined service scope and distinguish factual claims from illustrations, assumptions, and limitations.

Customer report versions follow their applicable source, numeric, prohibited-claims, consistency, and completeness checks. Products priced $950 and above also require written human QA approval before delivery. These product-specific controls are not a claim that every public page or automated output has been independently reviewed by a named expert.

Corrections to public materials

Material factual errors in public materials can be corrected. A meaningful change should be identified rather than disguised as the original statement. If research publications are added later, their update and version information should distinguish the period of the evidence from the date of a correction.

Send a correction notice through Harborfield's existing written contact channel, identifying the page, the statement, and the supporting source where available. Public-content corrections are separate from the product-specific correction allowance for a customer report.

Contact Harborfield about a factual correction

Automation and AI within defined controls

Automation and AI may assist structured analysis, drafting, transformation, or workflow. Their outputs remain constrained by source evidence, deterministic calculations where used, stated assumptions, limitations, and applicable quality controls.

Generated narrative does not create a source, validate an unsupported number, or override product-specific review and delivery controls. The presence of automation does not make uncertain evidence certain.