There are three proposals for the same study on the desk, with quoted prices differing by several dozen percent. The cheapest promises the same sample scope as the most expensive, the most expensive has the most polished presentation, and none explains directly where respondents will come from or how it will verify whether they provided reliable answers. Choosing a research agency is rarely decided by price – it is decided by what the provider has written (or omitted) about methodology, fieldwork, and data quality control.<\/p>\n
Research proposals have a strong tendency to resemble one another. This results from a simple mechanism: research agencies respond to the same brief, using the same vocabulary and often proposing the same technique – CAWI, CATI, IDI, or FGI. At the headline level, the differences disappear. They only emerge at the operational level, which clients usually do not know because they have never seen what fieldwork looks like from the inside.<\/p>\n
The real differences between research firms are concentrated in several areas. The first is the sample source. Two agencies declaring “n=1000, nationally representative sample” may work with entirely different materials: one using its own panel with a history of respondent activity and limits on survey participation, the other using a sample purchased ad hoc from a broker, without insight into how many surveys the same person completed in the past month. The stated outcome is identical, but data quality may be incomparable.<\/p>\n
The second area is the sampling approach. Quotas by gender, age, and region are a minimum requirement, not a complete methodology. The question is whether the market research agency explains how it handles difficult groups: older people in online research, B2B decision-makers in narrow industries, or users of niche product categories. A declaration is not enough here – the agency needs to describe the screening procedure, the expected incidence rate, and a contingency plan if the sample structure cannot be completed within the expected timeframe.<\/p>\n
The third area is the analytical layer. A proposal may end with data tables and descriptive commentary, or it may include specific techniques such as conjoint analysis, segmentation, price modeling, or factor analysis. This directly affects what questions can be asked of the data once the project has been completed.<\/p>\n
The fourth area is the team. This is not about a list of names in a presentation, but about who will actually lead the project. In practice, the person who convinced the client during the meeting may not appear in the project afterward. It is worth asking directly who will write the questionnaire, who will moderate the interviews, who will be responsible for analysis, and who will serve as the operational contact.<\/p>\n
Assessing a proposal requires reading it as a technical document rather than sales material. The following elements should be included in a proposal from a reliable research firm – their absence may be a warning sign.<\/p>\n
<\/br><\/p>\n
The section on quality control provides considerable insight into the provider’s actual standard. As Hume’s Institute experts point out, the description of fieldwork quality control and data verification says more about the quality of a proposal than the length of the presentation – an agency that can list specific data-cleaning procedures may explain its approach to quality management better than one that limits itself to generalities.<\/p>\n
What should this section include? In quantitative research: checks on completion time and rules for identifying speeders, detection of straightliners and response patterns, attention-check questions, verification of the logical consistency of responses, and, where justified, checks for duplicate devices or IP addresses, conducted in accordance with data protection regulations. It is also worth determining what percentage of completed surveys is typically rejected during data cleaning. A broader discussion of these mechanisms can be found in the article on data quality in quantitative research<\/a>. In telephone research: rules for monitoring or listening to recordings, the recontact procedure, and supervision of interviewers’ work. In qualitative research: the method of recruiting and verifying participants, the moderator’s experience in the relevant category, and the rules for transcribing and coding the material.<\/p>\n It is also worth verifying the team separately. Three questions are helpful: how many projects using this methodology the named researcher has worked on in the past year, who will write the questionnaire and whether the client will see it before the pilot, and who will lead the results presentation. The answers can quickly show whether the team presented in the proposal is the actual project team or merely a calling card.<\/p>\n Many unsuccessful research projects do not begin with a methodological error by the provider, but with a decision-making error on the commissioning side. Several patterns recur particularly often.<\/p>\n Comparing proposals solely by the price per interview.<\/strong> The unit price is a function of the assumptions: questionnaire length, the difficulty of the target group, the level of quality control, and the scope of analysis. A proposal that is one-third cheaper may differ in its assumptions, not its efficiency. If one market research agency assumes a 20-minute interview with an incidence rate of a few percent, while another assumes a 10-minute interview in the general population, these are two different studies being compared. Pricing is discussed in greater detail in a separate article on research costs – the only relevant point here is that a price without assumptions is not complete information.<\/p>\n Assessing an agency based on client logos.<\/strong> A reference list indicates that someone purchased a service at some point; it does not say whether the service was good or whether it was delivered by the same team. It is more revealing to ask for a description of a project with a similar methodological structure: what the research problem was, what technique was used, what difficulties arose during fieldwork, and how they were resolved. The description of difficulties is often more important here than the description of success.<\/p>\n Mistaking report length for analytical value.<\/strong> A 200-page presentation in which every slide contains one chart and a verbal description of what can be seen in that chart is not necessarily analysis. Value arises where there are cross-tabulations, relationships between variables, explanations of discrepancies, and an indication of which questions the data do not resolve. It is worth requesting an anonymized sample report from a similar project.<\/p>\n Overlooking data rights and access rules.<\/strong> The contract should clearly specify who may use the results and dataset, under what conditions, the format in which it will be provided, whether the client will receive the questionnaire script and codebook, and how long the agency retains the data. Access to the raw dataset may be restricted by personal data protection requirements or recruitment conditions, so it is worth agreeing on this before the project begins.<\/p>\nWhat mistakes most often undermine the selection of a research agency<\/h2>\n