Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Marketing Research › Research Methods

Commissioning a Research Agency or Freelancer

How to commission research from an agency or freelancer: when outside help pays, the brief, judging proposals, agreeing data handling and checking fieldwork.
Guideintermediate13 min readUpdated

When outside help pays

The research design guide weighs in-house against external work in its section on choosing the method family. Running a study yourself is cheaper and faster for small qualitative work; an agency brings sample access, fieldwork capacity and distance from the sponsor's hopes, which matters most on large or politically charged questions. The comparison of in-house and agency work makes the same trade in general terms: agencies offer specialist depth and flexible scale, the in-house team holds business knowledge and keeps what it learns.

For research, a few situations tip the balance towards buying help. Reach: the study needs people the business cannot contact itself, such as non-customers, competitors' customers or a business audience it has no list for. Independence: the result will be contested internally, or a sponsor's preference is known, and findings from outside carry more weight. Skill: the method needs experience the team lacks, such as moderating group discussions, designing a measurement survey or analysing weighted data. Capacity: the study is larger or faster than the team can run alongside its other work. Language and place: the customers are in another country and speak a language nobody on the team does, the case the guide to researching customers in a new market covers, from translated guides to native-speaking moderators.

The case is weaker when the question is about the business's own customers, the volume is small and the team has done similar work before. A handful of interviews run by someone who knows the product, as the customer interviews tutorial describes, often answers more than a commissioned study the team does not engage with. And the stakes matter: a decision that will set a KPI target for the year or commit a large budget justifies more care, and often more independence, than a reversible one. Where a small study handed to an agency is now set to report after its decision, the diagnosis for research taking too long weighs bringing the work in-house or buying only the part the team cannot do.

What to buy: a whole study or one part

Research can be bought whole or in parts, and the choice shapes everything that follows. Being clear about which part is bought prevents the most common dispute, in which each side assumed the other was doing the design or the analysis.

What is boughtWhat the supplier doesWhat stays with the businessSuits
A whole studyDesign, materials, recruitment, fieldwork, analysis and reportThe brief, sign-offs and the decisionTeams with little research experience, or questions needing independence
Recruitment onlyFinding and screening participants to the business's profileDesign, materials, sessions, analysisTeams that can run interviews or tests but cannot reach the audience
Fieldwork onlyRunning interviews, groups or a survey to the business's materialsDesign, materials, analysis, reportTeams with design skill but no capacity or reach
Analysis onlyCoding, tabulation, weighting or synthesis of data the business collectedDesign, collection, interpretation for the decisionTeams with data but without analysis skill or time
Advice and reviewReviewing the business's own design, questionnaire or reportEverything elseSmall studies that need a second pair of experienced eyes

Buying a part means the business carries the parts it keeps. If the business writes the questionnaire and the supplier fields it, the questionnaire's faults are the business's, which is a reason to pretest it before handing it over, as the tutorial on pretesting a questionnaire describes. If the supplier recruits and the business moderates, the screener needs joint sign-off.

The brief and the invitation to propose

The brief is the business's half of the conversation. The research brief template sets out its fields: the decision and its deadline, the research questions, what is already known, who must be heard and how they might be found, the method guidance, the topics the instrument must cover, budget and timing, consent and data handling, and the deliverables. The design guide adds that method guidance should be a preference open to challenge, not a specification, because the people who execute research often know a better route to the question.

The invitation to propose wraps the brief in a process. The agency RFP template shows the general form: the same information and the same questions to every supplier, in the same order, with the scoring criteria published in advance. For research, the questions to ask every supplier are particular:

  • How would you answer the brief's primary question, and what would the findings let us decide?
  • What design do you recommend, and why that rather than the alternatives?
  • Where will participants come from, how will they be screened, and who will that source miss?
  • How will you reason the Sample Size or the number of sessions, and what will the smallest group we need to report on be able to show?
  • Who will design the materials, who will moderate or field, and who will code and analyse? Are those the people we will meet?
  • What quality checks will you run on participants and on the data?
  • What will the analysis plan contain, and what will the deliverables include?
  • How will you handle participants' personal data, and which professional code do you work to?

A budget range belongs in the invitation. Without it, proposals describe different studies at different scales and cannot be compared. The general process around the invitation, from the shortlist to chemistry meetings and contracts, belongs to the guide to selecting, briefing and reviewing agencies.

Judging proposals on the question, not the presentation

Research proposals are easy to judge on the wrong things: the confidence of the presentation, the size of the sample, the familiarity of the method. The useful test is narrower. Does the proposal answer the brief's question, with a design that can, from a sample that reaches the people who matter, analysed by people who will be doing the work?

The weighted scoring matrix, as its page describes it, supplies the discipline: criteria and weights agreed before anyone reads the proposals, every score backed by evidence from the proposal's own text, must-haves applied as pass or fail gates rather than added up, and a check that the ranking survives a reasonable change of weights. Lawful handling of participants' data is a gate, not a criterion to trade against price.

Criteria that separate research proposals well:

  • Fit to the question. Does the method suit the kind of answer needed? The qualitative versus quantitative comparison is the reference: interviews and groups for why and how, surveys for how many and how much, and both in sequence when the brief has both kinds of question.
  • The sample and its gaps. A proposal that names its source and says who it misses is more trustworthy than one that promises a "representative sample" without saying of whom.
  • The materials. For a survey, whether the proposal reflects the wording, order and sampling points in the survey design guide; for interviews, whether it plans to ask about past behaviour rather than opinions and hypotheticals, as the customer interviews tutorial does.
  • The people. Who moderates, codes and analyses, and whether they are the people presenting.
  • Quality and analysis. How participants are checked, how data is cleaned, how open answers are coded, and what the analysis plan contains.
  • Deliverables. Whether the report will carry bases, question wording, a method statement and limits.

Price matters, but as one criterion among these. The cheapest proposal is often cheapest because it has dropped the parts of a study that are hardest to see: screening, pretesting, a second coder, the method statement.

Participants' data and the professional codes

A commissioned study usually means a supplier handles Personal Data about participants: contact details, screener answers, recordings and free-text answers. Under the GDPR, as the GDPR glossary entry puts it, the organisation that decides why and how data is processed is the controller, and a supplier processing data on its behalf is a processor. Which role each party takes in a given study, and the contract terms that follow, are for whoever owns data protection and for legal; this guide does not set them out. What the person commissioning the study can do is raise the questions early: what data the supplier will collect, where it will be held, who will see it, whether it goes to any further supplier, and when it will be returned or deleted.

The participants' side of the study is prepared as the tutorial on participant information and consent describes: every use of what people say listed, information in plain words, separate consent for optional uses such as clips or re-contact, and a deletion date with a process behind it. Agree who drafts each document, who approves it and whose name participants will see as the organisation running the study.

The professional codes the site's research guides describe from the MRS and ESOMAR set the standard for the work itself: participation is voluntary and informed, people are told who is collecting the data and why, answers stay confidential unless people agree otherwise, research is never a pretext for selling, and findings are reported honestly with their methods and limits. As general practice, asking a supplier which professional code it works to, and how it applies it to this study, is a reasonable part of judging a proposal. This guide makes no claim about membership, accreditation or what either body requires of its members. One consequence matters in practice: because research is not a disguise for selling, a commissioned study cannot be used to build a sales list, and confidentiality may mean the business does not learn which customer said what.

Signing off materials and watching fieldwork

The business should sign off every document participants will see or hear before fieldwork starts: the screener, the invitation, the participant information and consent, the discussion guide or questionnaire, and any stimulus. Read each against the brief. For a survey, also read the questionnaire against the analysis plan, so that every question feeds an output the report needs. The usual faults are a screener that gives away the qualifying answer, as the recruitment tutorial warns, an invitation that reads as marketing, and questions that lead. Sign-off is the last cheap moment to fix them; once fieldwork starts, every fix is a break in the evidence.

Watch some of the fieldwork. Observing interviews or groups, live or recorded, shows whether the moderator follows the guide, whether participants fit the profile, and what the evidence actually sounds like before it is summarised. For a survey, ask for a soft launch and an early look at the data: completion by question, the mix of respondents against quotas, and a sample of open answers. Agree in advance what happens if a check finds a problem, such as a quota filling with the wrong people, so the decision is not improvised under time pressure.

Keep the business's role in fieldwork to observing and answering questions. Sponsors who join sessions and explain the product, or who ask for the questionnaire to change halfway through, damage the independence the study was bought for.

Accepting the deliverables

A research deliverable is accepted when the business can see how the findings were reached, not when the slides arrive. The survey design guide's reporting rules set the standard: a base for every figure, the exact question wording beside each result, differences between groups shown only where the sample supports them, and the sample's limits stated plainly near the top. For qualitative work, findings should carry their evidence: anonymised quotes attributed by group, and a note of how widely each theme was shared among the people interviewed.

Ask for the materials behind the report as well as the report itself: the final questionnaire or guide, the data tables with bases and weighting noted, anonymised verbatims, the codeframe where open answers were coded, and a method statement that says who took part, how they were found, when, and who was missed. The guide to reporting research findings covers how findings should be presented with their evidence and limits, which is the standard a commissioned report can be held to.

Check the deliverables against the brief's questions before the readout, and send questions in writing. A finding that rests on a small base, a chart without its question wording or a recommendation with no evidence behind it is a reason to ask, not a reason to reject the study. If a commissioned deck was accepted without those materials and the study has since changed nothing, the diagnosis for research findings that are not acted on shows how to recover the tables and reconnect the findings to a decision. Agree, too, what the business may do with the material afterwards, which belongs in the contract the general agency guide describes. To review bought studies beside the research the business runs itself, the marketing research audit scorecard rates how agencies are briefed, judged and signed off alongside every other stage of research and ranks what to fix first.

An insurer commissions a switching study (illustrative)

Illustrative example — not a real result

Skarvolde Insurance commissions a study of switching at renewal

Skarvolde Insurance, an invented home insurer, loses about a third of the customers who shop around at renewal and wants to know why, to decide whether to change its renewal pricing message or its claims communication. It commissions a whole study, sending the brief to 3 suppliers with a budget range of £30,000 to £40,000 and a deadline 12 weeks away.

The weighted matrix, agreed before the proposals arrive, weights fit to the question at 30%, the sample and its gaps at 25%, materials and analysis at 20%, people at 10% and price at 15%, with lawful data handling as a gate. Supplier A scores highest on presentation but promises a "nationally representative sample" without saying of whom; Supplier C is the cheapest but proposes a survey only. Supplier B proposes 16 interviews with recent switchers followed by a survey of 1,200 people who renewed or switched, names its source as a panel screened for home insurance renewal in the past 6 months, and states that people who renew automatically without shopping will be missed. B scores 7.6 against 6.9 and 6.1, and still leads when the weights for price and the sample are swapped.

At sign-off, the screener's first question reads "Did you leave your insurer because of price?"; it is rewritten to ask which of several things happened at the last renewal. The business observes 4 interviews. The data tables arrive with a base on every figure, and they show that only 38 respondents switched after making a claim, too few to report that group's reasons with confidence. The report says so, and the business commissions a small follow-up of claim-related switchers rather than acting on 38 answers.

Common mistakes

  • Buying a method instead of an answer. A brief that asks for "a survey" gets a survey, whether or not it answers the question.
  • No budget range. Proposals describe different studies at different scales and cannot be compared.
  • Judging the presentation. The most confident pitch is not the best design; score proposals on evidence in their text.
  • Accepting a sample without its gaps. A source with no stated limits is a source whose limits will surface in the findings.
  • Unclear division of work. Each side assumes the other is writing the questionnaire or coding the open answers.
  • Leaving data handling to the end. Who processes participants' data, and on what terms, is settled before fieldwork, with data protection and legal.
  • Skipping sign-off. A leading screener or a selling invitation approved by default.
  • Interfering in fieldwork. Changing the questionnaire mid-way or explaining the product in sessions spoils the independence that was paid for.
  • Accepting slides without tables. Findings without bases, wording and a method statement cannot be checked or reused.

Frequently asked questions

Should the brief tell the agency which method to use?

State a preference and the reasons for it, and invite challenge. The research design guide makes the point that the people executing a study often know a better route to the question, and a brief that specifies the technique closes that off. What the brief must fix is the decision, the questions, who must be heard, the deadline and the budget range; the method is what the proposals compete on.

How do we compare proposals that recommend different methods?

Judge each on how well it answers the brief's questions, not on which method you expected. Agree the criteria and their weights before reading the proposals, apply must-haves such as lawful data handling as pass or fail gates, and score each proposal with evidence from its own text. Then ask each shortlisted supplier the same follow-up questions: where the sample comes from and who it misses, who moderates and analyses, and what the deliverables will contain.

Who owns the data and the materials from a commissioned study?

Agree it in the contract before fieldwork, with whoever handles legal and data protection. As general practice, the business should expect to receive the questionnaire, guides, data tables, anonymised verbatims and the report, and to be able to reuse them. Participants' personal data is a separate matter: a supplier processing it on the business's behalf is a processor under the GDPR, and what it may keep, for how long and what it must return or delete belongs in the terms that data protection agrees.

Terms used on this page

Frameworks

Where this page is used

Sources

  1. Market Research Society — MRS
  2. ESOMAR — ESOMAR
  3. Regulation (EU) 2016/679 (GDPR) — EUR-Lex

Related