How to Brief a Research Assistant

A vendor decision waits for three missing facts. A competitor changes its packaging, but nobody records when. A market-entry meeting relies on the loudest opinion because the evidence is scattered across bookmarks and messages. The research is important enough to postpone and never urgent enough to own, so a price change, partnership or purchase proceeds blind.

Businesses rarely lack information. They lack a researchable question, an agreed source standard and a deliverable designed for a decision. An online research assistant can gather, structure and maintain evidence; they cannot supply the judgement the client has not defined. This guide shows how to brief the work so the result can be checked and used. Buyers ready to scope recurring support can review OVELITHUB’s online research assistant service.

Why the research never gets done

Research competes poorly with live client work. The founder can answer email, approve a campaign or join a sales call and see an immediate result. Comparing eight vendor security pages or reconstructing a market series requires uninterrupted time, careful notes and tolerance for ambiguity. It slips to the evening, is rushed through a general search and becomes a few tabs with no record of why they matter.

The consequences are concrete. A vendor is chosen from a confident demo without checking export rights. A market appears smaller because two sources use different definitions. A competitor’s promotional price is treated as its normal price. A partnership is discovered after its application window closes. The research cost is not measured in unread reports; it appears in avoidable decisions.

Delegation solves the capacity problem only when the requester owns the decision question. “Look into our competitors” transfers uncertainty, not work. The researcher must guess which competitors, which market, which period, which attributes and what the business intends to decide. A polished answer can still be irrelevant.

What desk research actually covers

Desk research uses information that already exists, whether public, licensed or held by the company. Common work packages include:

• Competitor and pricing monitoring: a dated tracker of plans, inclusions, promotional conditions, positioning and material changes.
• Market and segment sizing: a transparent estimate using published series, definitions, assumptions and range rather than one unsupported total.
• Vendor and tool comparison: a requirements matrix with evidence, gaps, contract questions and a shortlist for stakeholder review.
• Account research: a sourced company brief covering structure, current priorities, relevant public changes and questions for a sales conversation. List building itself belongs to prospect research for sales teams.
• Regulatory scanning: a change log linking to regulators and official text, with review by qualified legal or compliance professionals before action.
• Events and partnerships: a calendar of relevant programmes, eligibility, audience, deadline, cost and application source.
• Content sourcing: an evidence pack of primary research, datasets, expert material and counterarguments for a writer or subject-matter reviewer.

Each output is shaped around a decision. A researcher can identify and compare public evidence; they should not present an estimated market, regulatory implication or vendor claim as certainty. For deeper planning work, the separate market research assistant guide covers decision-shaped market briefs.

A search is not a piece of research

A search retrieves possible sources. Research connects them to a defined question through a reproducible method. Three elements create that difference:

1. A defined question. “Which customer-support platform meets our data, integration and regional requirements for a 20-agent team?” can be answered. “What is the best support tool?” cannot.
2. A source standard. The brief states which evidence is acceptable, how recent it should be and how vendor claims will be labelled.
3. A comparison structure. Every option is tested against the same criteria, missing values remain visible and the researcher does not silently move the goalposts.

Rigour is slower than collecting the first plausible answer. That is the point. It makes disagreement useful: a stakeholder can challenge a definition, source or assumption instead of arguing with an undocumented conclusion.

How to write a research brief that returns usable work

State the question, decision and deadline

Write one primary question, then name the decision owner and the decision it feeds. Add the true deadline and the consequence of lateness. For example: “By 18 September, compare the three shortlisted scheduling platforms so the operations director can select one for a security review and pilot.” The researcher now knows this is a shortlist, not a purchase recommendation.

List the decision criteria in priority order. Separate must-haves from preferences. If data location, API access or a minimum user role is disqualifying, say so before research begins. Include the current system and why the business is considering change.

Set boundaries and exclusions

Define geography, customer segment, company size, product edition, currency, tax treatment and publication window. Name excluded alternatives and why. Clarify whether promotional pricing, annual prepayment and negotiated enterprise rates belong in the comparison. A small boundary note prevents a large amount of irrelevant gathering.

State whether the assistant may contact vendors. Desk research and primary outreach have different consent, representation, time and record-keeping requirements. If outreach is excluded, unknowns should become questions for the next stage rather than guessed answers.

Define what counts as acceptable evidence

Use a source hierarchy. Product capability and price should come from official documentation, terms or a dated vendor response where possible. Company financial statements should come from regulatory filings. Market data should come from statistical agencies, credible research organisations or an industry body whose method and scope are visible. Independent reviews can reveal implementation problems but should not override contractual facts.

For every material claim, require title, publisher, publication or access date, direct URL and a short note explaining what the source supports. If the evidence is paywalled, store the citation and permitted internal extract, not a redistributed copy. Label vendor-sponsored research and distinguish publication date from the period measured.

Agree the output before work starts

Name the file format, columns, maximum memo length, update method and review meeting. Specify whether the researcher should recommend, identify a shortlist or only present evidence. Require a source log, assumption log, open questions and confidence level. Most decisions do not need a long report; a one-page comparison supported by a clean source table is often more useful.

Worked example: from vague request to researchable brief

Vague: “Look into CRM tools and tell me which is best.”

Researchable: “Compare Vendor A, Vendor B and Vendor C for a UK-based 12-person B2B sales team replacing spreadsheets. The decision is which two products enter a live trial next month. Must-haves are opportunity stages, email logging, role-based permissions, CSV export, documented API access and current UK personal-data terms. Compare published annual pricing for 15 seats, onboarding charges, contract term, export method, integration with our named email platform and independent implementation concerns. Use vendor documentation and terms for capabilities; label all review-site evidence. Deliver a 15-row matrix, one-page summary, source log and unanswered vendor questions by Friday. Do not contact vendors or assess legal compliance.”

This version makes missing evidence useful. If a vendor does not publish export limits, the cell reads “not confirmed” and becomes a trial question. The assistant is not pushed into inventing certainty to make the table look complete.

Source discipline is the quality control

Prefer primary sources for facts controlled by an authority: legislation from the legislature or regulator, economic data from the statistical producer, product behaviour from maintained documentation and company claims from filings or official disclosures. Secondary sources help explain context, compare interpretations and discover primary material. They should not be cited as the origin of a statistic they merely repeat when the original is available.

Useful confirmed starting points include the OECD Data Explorer for comparable economic and social series; the Eurostat Data Browser for EU statistics; US Bureau of Labor Statistics data; the SEC EDGAR filing search; and the free UK Companies House register. Use the source’s definitions and release notes, not only the headline value.

Niche information often exists outside general search results. Tenders Electronic Daily publishes EU public-procurement notices; WIPO PATENTSCOPE covers published international applications and participating patent collections. Trade bodies, regulator registers, planning filings, conference programmes, job postings and specialist press can reveal demand, investment and capability. Their coverage and incentives must still be stated.

Never detach a statistic from its measured period, population or unit. Record whether a figure is actual, forecast, survey estimate or vendor model. When two credible sources conflict, do not average them automatically. Compare definitions and methods, then explain the difference. Unsourced material should be returned for evidence or removed, not polished into publication.

Choose a deliverable that will be used

A comparison matrix supports vendor, competitor and location screening. Keep criteria in rows, options in columns, attach a source to each material cell and preserve unknowns. A source-linked summary memo supports a decision meeting: question, findings, implications, uncertainties and next action in one or two pages.

A maintained tracker supports recurring monitoring. Include entity, field, observed value, source, observed date, prior value, change, reviewer and next check. Preserve history instead of overwriting a price or policy. A recommendation note can identify the option best aligned with the stated criteria, but it must separate evidence from the analyst’s reasoning and leave the final decision with the owner.

Default to a living tracker when the subject changes. A quarterly presentation made from the tracker is disposable; the dated evidence underneath is the asset. Store it where the business controls permissions and version history.

Turn recurring research into a standing service

One-off research answers a moment. A rhythm reveals a trend. A weekly competitor check can distinguish a temporary promotion from a packaging shift. A monthly market scan can show which signals are persistent. A quarterly vendor review can catch contract, product or security changes before renewal.

Set a cadence for each field, not only each tracker. Price might be checked weekly during a launch and monthly afterward; financial filings follow their release schedule; regulations need event-driven alerts plus a periodic review. Record “checked, no change” so absence of an update is distinguishable from an abandoned tracker.

Use trigger thresholds. An assistant should escalate when a key competitor removes a plan, a regulator publishes a relevant consultation, a vendor changes data terms or a tracked price moves beyond the approved band. Routine reporting should summarise material change, not reproduce the whole database.

Tools and access the assistant needs

Provide advanced search operators, official-source bookmarks, an approved browser and spreadsheet or database template. Where the business holds subscriptions to industry databases, standards, analyst research or company information, create named user access under the licence rather than sharing passwords. A paid source is worth adding when the decision value, coverage, update frequency and permitted use justify it—not because it makes the memo look authoritative.

Internal evidence can be more relevant than published data. Give scoped access to analytics, CRM stages, loss reasons, support tags, prior vendor assessments and approved interview notes when the question requires them. Remove personal data that is unnecessary and respect contractual restrictions. The assistant should know which internal metric is authoritative and who can resolve a definition.

Automation and generative tools can accelerate query expansion, extraction, transcription and first-pass classification. They can also misstate, conflate or fabricate citations. Treat generated output as leads to verify, never as evidence by itself. The researcher remains accountable for opening the source, checking the relevant passage and linking it to the claim.

Quality control: test the evidence, not the formatting

At kickoff, calibrate five sample findings together. During production, a supervisor should sample high-impact rows, a random group and every low-confidence item. Open the source, reproduce the value, check the date and confirm the comparison criterion was applied consistently. Measure citation completeness, field accuracy, rework, on-time delivery and meaningful gaps discovered.

Require three annotations: confidence in the finding, gap that remains and disconfirming condition—what would need to be true for the conclusion to be wrong. A vendor may appear cheaper only if usage stays below a threshold. A market estimate may hold only if two categories are mutually exclusive. These notes expose fragile conclusions before a meeting turns them into fact.

OVELITHUB assigns a defined brief, source hierarchy and output template, then supervises work through calibrated examples and sampled citations. The client reviews domain judgement and the decision. Research quality improves when corrections become written rules for the next cycle rather than private feedback on one file.

What a research assistant cannot do

A research assistant cannot make the commercial, clinical, legal or strategic decision for the accountable owner. They cannot supply domain judgement that has not been explained, certify a vendor from public pages or manufacture a market figure when no suitable evidence exists. “No reliable public source found” can be the correct result.

They also cannot turn desk research into primary research without a new scope. Interviews and surveys require sampling, instrument design, consent, recruitment, recording, analysis and privacy decisions. An assistant can coordinate those steps under qualified leadership, but web searching is not a substitute for hearing from the target population.

Proper research sometimes narrows uncertainty rather than eliminating it. Its value is a traceable basis for the next action: trial two vendors, commission interviews, ask counsel a precise question or decide that evidence is too weak for investment.