Key points

A good research proposal restates the decision problem in its own words and shows how each method answers a specific objective.
Evaluation criteria are set before the proposals are opened, with weights for technical quality and for price.
The sample, fieldwork quality control and the named team say more about future data quality than any agency credentials presentation.
The lowest price is often the result of a smaller sample, an untested questionnaire or missing quality control, not of efficiency.
A clarification meeting before contracting resolves more ambiguity than any annex.

1. The problem: three proposals, three prices, no way to compare

A company wants to know why its sales have dropped in one region. A donor-funded programme needs a baseline study before launch. A local public authority wants to know which services residents are missing. In each case someone writes a brief or terms of reference, sends it to several agencies and gets back documents of 10 to 40 pages, full of methodology, tables, CVs and prices that sometimes differ several times over. Choosing between them is a research decision in its own right: the winning research proposal will determine what data you end up with, how reliable it is and whether it answers the question you paid for.

The usual mistake is to compare proposals on price and overall impression. Two proposals can both promise "a representative survey of 1,000 respondents" and mean entirely different things: one a probability sample, face-to-face interviews and 15% of interviews verified by revisit; the other an online link shared on social media and weighted at the end. On paper they look the same. In the data, they do not. This article helps you see the difference before you sign. For the step that comes before the proposals, writing the request itself, read how to write a research brief.

2. What a research proposal should contain

Formats vary, but a complete proposal covers every stage of the research process. If a research proposal leaves one of them out, that is not an editing slip; it is a signal of how the project will be run.

Understanding of the problem. The agency restates the decision problem in its own words, adds context the brief did not contain and, where relevant, offers an alternative reading. A proposal that copies the brief back without adding anything was not written for your project.
Research objectives. Framed as specific questions or testable hypotheses, not as broad themes.
Research design. Exploratory, descriptive or causal, with the specific methods, their sequence and how they connect: for example, in-depth interviews that feed the questionnaire of a survey.
Sampling plan. The target population, the sampling frame, the selection method, the sample size with a justification, and the expected precision.
Data collection. Who collects the data, how, with what training and with which quality control mechanisms. If any fieldwork is subcontracted, the proposal should say so.
Analysis. Which analyses will be run, from cross-tabulations to statistical models, and how the results will be interpreted against the decision.
Reporting. Interim reports, the form of the final report, the presentation, the datasets delivered.
Timeline and cost, broken down by phase.
The team, with roles and relevant experience, not just a list of names.

3. Evaluation criteria are set before the envelopes are opened

The simplest protection against an arbitrary choice is a scoring grid written before anyone reads the proposals. The evaluation criteria follow from the brief: if the key requirement is comparability across regions, sampling gets the heaviest weight; if the project is exploratory, the quality of the interview guide and the moderators' experience matter more.

For donor-funded programmes and public procurement, the grid is usually set by the procedure: a technical evaluation with a minimum qualifying score, followed by a financial evaluation, with the final score combining the two according to weights fixed in the terms of reference. Private companies are not bound by this discipline, but they gain from it. A hypothetical grid for a quantitative survey might look like this: understanding of the problem 15 points, methodology and sampling 30, fieldwork quality control 15, team 15, timeline and workplan 10, price 15. The exact weights are yours to set; what matters is that they exist before you see the prices.

If you evaluate as a panel, each member scores independently and large gaps between scorers are discussed openly. A 10-point gap on methodology usually means one evaluator noticed something the other missed.

4. How to read the methodology section

In a research proposal, methodology is where submissions differ most, and where a reader without research training is most easily impressed. Four questions separate a sound methodology from a decorative one.

Does the method fit the question?

If the brief asks for figures that can be projected to a population, the answer is quantitative research with a probability sample. If it asks you to understand motives, barriers or the language customers use, you need qualitative research. A proposal that answers a "how many" question with focus groups, or a "why" question with nothing but a closed survey, has misread the request. Our guide to which type of research solves your problem covers this matching in detail.

Is the sample justified or merely stated?

A number without a rationale is not a sampling plan. Look for a precisely defined population, the source of selection, the method (random, stratified, multi-stage, quota), and the margin of error for the total sample and for the subgroups you will compare. A sample of 1,000 people gives good precision nationally, but if you want to compare five regions, each one has only around 200 respondents and a much wider margin of error. A good proposal says so without being asked.

Is the questionnaire treated as a stage of work?

The proposal should show who builds the instrument, how it is approved, which languages it runs in and how it is tested before fieldwork. A missing pretest is a clear warning sign: an untested questionnaire produces errors that no analysis can repair afterwards. See our articles on questionnaire design and on questionnaire pretesting.

What happens to non-response?

Every survey loses some of the people selected. A mature proposal explains how call-backs are made, how unavailable households or individuals are replaced, how the response rate will be reported and whether, and how, the data will be weighted.

5. Quality control and the team: what they say about the data to come

Most quality problems arise in the field, not in the office: shortened interviews, respondents picked for convenience, questions read differently from how they are written and, in rare cases, fabricated interviews. A serious research proposal describes concretely how these are prevented: interviewer training, the share of interviews verified by telephone call-back or revisit, checks on duration and GPS coordinates for tablet interviews, listening to recordings in telephone surveys, logical checks on the dataset. If a proposal says only "we will ensure data quality", ask how.

The team matters just as much. Check who actually runs the project day to day, not just who signs the proposal, how much time each member is allocated and whether the named people have experience with this type of study and this population. For multilingual studies, ask who translates the instruments and who moderates or interviews in each language. In Moldova, where many studies run in both Romanian and Russian, an in-house multilingual team reduces the risk that the two language versions of a questionnaire end up measuring different things.

6. Price: what you are actually buying

Price differences between proposals almost always have a methodological explanation. A price far below the others usually comes from one of the following: a smaller sample than requested, or one not spread across the relevant subgroups; a cheaper collection method, for instance online instead of face to face, for a population that is not fully online; no pretest; reduced quality control; a report limited to tables, with no analysis or interpretation. None of these savings is visible in the proposal summary, but all of them are visible in the data.

A fair comparison uses the same basis: the same number of completed interviews, the same method, the same questionnaire length, the same deliverables. If the proposals are not comparable, ask the agencies to align them or to price their variants separately. A fuller explanation of cost drivers is in what market research costs and why.

7. Common warning signs

Generic text that would fit any client, with your company name dropped in here and there.
A promise of "national representativeness" for an open online survey, with no sampling frame and no discussion of people who do not use the internet.
A sample size with no margin of error and no breakdown by subgroup.
A timeline that leaves no room for questionnaire approval, pretesting or translation.
No mention of personal data protection or respondent consent.
A team that looks impressive on paper but has no roles and no allocated days.
Guarantees about the findings ("we will confirm that..."), which contradict the independence of research.

8. Solid proposal versus weak proposal

Section	Solid proposal	Weak proposal
Problem	Restated, with added context and explicit decision questions	Copied from the brief
Design	Methods tied to objectives, with their sequence and logic	A list of unconnected methods
Sample	Population, frame, method, size, margin of error by subgroup	"1,000 representative respondents"
Instrument	Stages for drafting, approval, translation and pretesting	"The questionnaire will be developed"
Fieldwork	Training, share verified, automated checks, subcontracting declared	"We will ensure quality"
Analysis	Named analyses linked to decisions	"Statistical analysis of the data"
Team	Roles, allocated days, relevant experience	Generic CVs, no roles
Ethics and data	Consent, anonymisation, data retention and deletion	Missing

The table does not replace the scoring grid, but it shows where to look when two proposals end up with similar scores.

9. What to ask the agency before you sign

For shortlisted proposals, a one-hour clarification meeting is the best investment in the whole selection process. It shows how the team that will actually do the work thinks, and it turns ambiguities in the text into written commitments.

How did you understand the decision the study has to support, and what would you change in the brief?
How did you calculate the sample size, and what precision do you have for each subgroup you will report on?
What frame do you select respondents from, and how do you handle non-response?
How, and with how many respondents, will you pretest the questionnaire in each language?
What share of interviews do you verify, and by what method?
Which part of the work is subcontracted, and to whom?
Who runs the project day to day, and how many days are they allocated?
How do you obtain respondent consent and protect personal data?
What exactly do you deliver: report, tables, dataset, presentation, recommendations?
What happens if fieldwork slips or the response rate falls below the estimate?

Choosing an agency is not an administrative exercise. A carefully evaluated research proposal is the first quality check on the study, and evaluation criteria fixed in advance protect you from picking the most persuasive proposal instead of the soundest one. If you would like a second opinion on the proposals you have received or on your terms of reference, our decision support team can help, and answers to the most common questions are on our FAQ page.

Do you have a decision to take?

Tell us the problem, not the method. We propose the right research design and explain what it can and cannot deliver.

The Perspicuus research team · 16 February 2026