Customer Satisfaction Research Agency: Services & Scope
What a customer satisfaction research agency does, what to ask before hiring one and how to read its scores and service recommendations, with a worked example.
A customer satisfaction research agency helps collect and interpret feedback about a service customers actually received. Its work may include study design, contact planning, fieldwork, analysis and recommendations. Before hiring one, check which tasks it will own, how answers connect to real customers and what decision the findings should change.
The proposal needs to identify who will do that work and what evidence will make the recommendation usable. A dashboard leaves the decision incomplete unless someone can connect its answers to the customers and visits being studied. For a broader supplier shortlist, our market research companies guide compares providers by the evidence and deliverables your decision needs.
Example: investigating a repair service’s feedback
Mereford Repairs is hypothetical. It is considering arrival updates before repair follow-up or aftercare changes. The study covers customers with a repair visit in a defined fortnight. Completed means a technician attended and the visit ended; it does not mean the appliance was fixed.
The team chooses one customer as the reporting unit, referring to that customer's first qualifying visit in the fortnight. Someone with two visits can explain both experiences, but those visits do not become two separate customers. A later visit stays in the operating file, linked to the same customer.
The agency must link each answer to the selected visit. Familiarity with appliance repair is not enough. Verify relevant service experience before using a general panel. People who never received this service may assess a proposed offer, but cannot report satisfaction with receiving it.
Ask the shortlisted agency to describe how its recruitment and questions will keep that distinction. Our research sampling guide covers wider choices about frames and analysis units. Here, the supplier must show how those choices affect the actual customer file and fieldwork it will deliver.
For a new inquiry about an agreed service, our mystery shopping research guide explains a trained evaluator's unannounced encounter. That check remains separate from satisfaction feedback about customers' actual visits.
Existing records determine the first piece of work
Describe available records before accepting a survey plan. Can visits, customers, contacts and prior feedback be connected? Agree the contact route with the client owner responsible for it.
Three starting positions should produce different proposals:
What the client actually has | What to commission first |
|---|---|
Completed visits exist, but customer links or usable contacts are missing. | Reconcile the visit and customer files, identify the unreachable eligible group and establish a feasible contact plan. Accept a documented frame and fieldwork design before promising a satisfaction readout. |
Eligible customers can be contacted, but suitable recent feedback has not been collected. | Commission the visit-specific instrument, approved invitations and follow-ups, human data collection, analysis and service-team readout. Keep the contact and response records with the findings. |
Current human feedback and operating records already cover the intended decision adequately. | Review that coverage, then commission the remaining diagnosis or comparison. Use targeted customer follow-up where an explanation is missing; do not automatically repeat the whole rating exercise. |
Whether existing evidence is sufficient depends on its use, coverage, timing and analysis. The small file below illustrates checks, not adequate real-study coverage. Ask the agency to explain that judgment before recommending another survey wave.
AAPOR's survey guidance recommends considering timely existing evidence and administrative records when planning research. It also connects collection mode to the available frame and contact information. For Mereford, this means the source inspection belongs in the proposed work, rather than becoming a surprise after the questionnaire is approved.
If the only available material is unrelated category opinion, ask the provider how it will obtain evidence from eligible Mereford customers. If no workable route exists, change the scope or stop the proposed measurement. Relabeling other opinions does not supply the missing experience.
Agencies can leave different work with your team
Two agencies can both offer customer satisfaction research while leaving different tasks with your team. Ask each to revise its proposed engagement around the same available files and service decision.
Arlington Research describes bespoke satisfaction projects covering questionnaire and discussion-guide development, KPI advice, fieldwork, analysis and reporting. Its service description supplies concrete stages to discuss. Confirm which stages the proposed Mereford engagement includes and who handles access to eligible customers.
Kadence describes combining qualitative and quantitative work to investigate expectation and performance gaps, with customer research that can be linked to a CRM. If that linkage matters, ask how the proposed study will connect a particular visit, customer, answer and later service record. A general integration description does not establish the contents of your project.
The client data owner supplies visits, status definitions and an approved contact route. The research lead checks eligibility and selects each invitation’s visit. Name who programs and sends invitations, manages reminders and opt-outs, saves the questionnaire, reviews answers and presents findings. Record any stage the client must perform.
Ask to see the returned evidence: eligible customer counts, contact and response outcomes, exact questions, source-linked explanations and missing information. A dashboard alone may leave that unclear. Our market research services guide covers engagement models; this study needs named responsibilities for actual service visits.
Five customer records sit behind three ratings
Suppose the agency receives this worked Mereford file after the source reconciliation. There are six completed visits, J01 through J06, belonging to five customers. C1's J06 concerns another appliance; J01 remains that customer's first qualifying visit. J06 is a legitimate visit, not a duplicate to delete.
The example contact rule is an invitation 48 hours after the selected visit. That is a chosen timing for a visit-specific question, not a universal recommendation. Ask the agency how the timing fits your service and what a later outcome question would require.
Customer and selected visit | Invitation, answer and evidence retained |
|---|---|
C1: J01 selected; J06 retained separately | Invited; rating 5/5. The customer reports a late arrival and a working appliance. A separate dispatch record confirms arrival outside the agreed window. |
C2: J02 | Invited; rating 1/5. The customer says the appliance still leaks. The completed-visit status does not verify a successful repair. |
C3: J03 | Eligible visit, but no usable contact in the supplied file. No invitation or rating. |
C4: J04 | Invited; rating 2/5. The customer says the aftercare instructions were unclear. No observation of their understanding is available. |
C5: J05 | Invited; no rating returned. No satisfaction judgment is inferred. |
The supplier should reconcile six visits to five eligible customer records, four invitations and three answered ratings. C3's missing contact and C5's nonresponse remain different dispositions. Neither becomes a neutral rating, a dissatisfied answer or an invented response.
Use a question tied to the specified visit, such as “How satisfied were you with the repair visit identified in this invitation?” Agree the ordered response scale and any inability-to-judge option before collection. Keep an overall judgment ahead of detailed arrival questions if the intended measure is an unprompted overall evaluation.
Qualtrics describes a common CSAT convention that counts the two highest ratings on a five-point satisfaction scale. Under that convention, this unweighted example has one positive answer among three ratings: 1 ÷ 3 × 100 = 33.3%. It is a description of these three invented answers. It is not one satisfied customer out of all six visits, nor evidence about the two customers without ratings.
Keep the scale and every rating, not only the percentage. Three answers cannot measure how common the service problems are, support a reliable customer-group comparison or benchmark another firm.
Make the recommendation survive a contrary case
The accounts challenge arrival updates as the universal fix. C1 reports lateness but rates highly. C2 reports a repair still needed. C4 describes aftercare. Arrival updates may matter, but these answers do not justify treating them as the remedy for all three.
Ask the agency to show what it would recommend differently because these cases exist. For C1, investigate how the reported repair outcome and late arrival contributed to the judgment. For C2, examine the reported fault and appropriate repair records before classifying it as a communication issue. For C4, check the instructions supplied and whether the customer can understand the required action.
Those are distinct follow-up assignments. A summary that turns all three into “customers want better communication” would remove the very differences the service team needs. Our mixed methods guide explains how accounts and measures inform one another. The agency acceptance question here is whether the returned evidence changes the proposed service action.
Mereford can therefore decline an unsupported instruction to roll out arrival messages as the overall satisfaction fix. It can ask for a revised readout that separates arrival reliability, repair effectiveness and aftercare, with an owner and next check for each. The records demonstrate why a recommendation needs revision; they do not establish which problem is most common or which investment will improve satisfaction most.
Assign the next service checks
Agree what the final recommendation must show. The data owner checks the eligible file and exclusions. The research lead explains contact coverage, answer counts, timing and supporting accounts. The service owner sees what can change and what remains unverified.
Keep arrival updates as one candidate. Assign the repair problem to operations and aftercare material to an understanding check. Keep missing experiences visible. The client can accept that diagnosis while declining a broad rollout.
If the business already has adequate, current human measurement, the next engagement could compare feasible arrival notices or revised aftercare explanations. Operations must first confirm what those alternatives truthfully promise. A message cannot create an appointment window the dispatch process cannot deliver, and clearer instructions cannot repair a leaking appliance.
An eventual evaluation should match the intended change. Arrival updates need evidence about the information customers receive and the service delivered. Aftercare revisions need a comprehension or use check. Repeat visits, complaints, later satisfaction and actual repeat business need their own observed records and an appropriate comparison. Stated intention to return is a different outcome from returning.
For a utility's future service plan, our water utilities customer research guide connects proposed improvements with the household bill and separates acceptability from affordability. Satisfaction with past service cannot settle that planning decision.
Feasible service messages can be compared separately
Sapien builds simulated audiences grounded in purchase, behavioral and research data. We compare alternatives, investigate differences between groups and explore the responses behind the findings. Our team builds the audience, runs the study and helps interpret the results; existing customer research can add context.
For Mereford, a simulated comparison could explore how customers interpret feasible arrival notices or revised aftercare explanations. Operations must confirm what each message can promise, and the study needs to represent the relevant experience. Keep that output separate from measuring actual customers' satisfaction or observing whether an appliance was repaired. Our synthetic vs human respondents guide explains that source choice.
Book a demo to compare the arrival notices or aftercare explanations your service can support. Use the customer accounts to define what each version needs to clarify, then keep the operational repair questions with the team able to resolve them.
