Franchise customer research is customer research run across businesses your brand does not own. The brand sets the promise, the trademark and the standards. An independent operator runs the transaction, hires the staff and holds the customer record. The customer, meanwhile, believes they bought from the brand. Every research problem in a multi-unit business starts from that three-way split, and almost every research programme that fails in one fails because it was designed as though the split were not there.
This guide covers who actually owns the customer in a franchised or multi-unit brand, what that does to your sample frame, and how to run brand-level research when the brand does not hold the list.
The scale of the problem is larger than most research teams assume
The International Franchise Association's 2026 Franchising Economic Outlook, released on 19 February 2026, projects that the number of US franchise establishments will grow from 832,521 to 845,000 units, an increase of 1.5%. Franchise employment is expected to rise by more than 150,000 jobs to nearly 8.9 million, and franchise output to climb from $907.3 billion to $921.4 billion.
Two derived figures matter more for research design than the headline growth rate. The first is that 845,000 establishments and roughly 8.9 million jobs works out to about 10.5 employees per unit. The second is that $921.4 billion spread across 845,000 units is about $1.09 million of output per establishment. These are small businesses. They are not branch offices with a corporate IT department, a data protection officer and a research budget. They are owner-operated businesses that happen to carry your sign.
That is the population your research programme has to reach, and it is why a research design borrowed from a single-owner enterprise breaks immediately.
What the law says a franchise is, and why it decides your data access
The Federal Trade Commission's Franchise Rule defines the relationship in 16 CFR 436.1(h). A franchise exists where the terms specify that the franchisee obtains the right to operate a business identified with the franchisor's trademark, that the franchisee makes a required payment, and, in the middle element that matters here, that:
The franchisor will exert or has authority to exert a significant degree of control over the franchisee's method of operation, or provide significant assistance in the franchisee's method of operation.
Read that carefully, because it draws the line your research has to live inside. The franchisor is entitled, by the very definition of the arrangement, to control the method of operation. Nothing in that definition gives the franchisor the customer. The right that makes you a franchisor is a right over how the business is run, not a right over who walks through the door or what their email address is.
In practice this means the customer record sits wherever the system that captured it sits. If the loyalty app, the online ordering flow or the reservation platform is brand-operated, the brand holds the record. If the point-of-sale terminal, the walk-in booking or the service ticket is operator-operated, the operator holds it. Most multi-unit brands are a mixture, and almost none of them can produce a single, complete, deduplicated list of the people who bought from the brand last quarter.
What this does to your sample frame
A brand-level survey in a multi-unit business is not drawn from a population. It is drawn from the union of however many operator lists happened to be handed over, and that union is shaped by three things you did not choose:
- Which operators participate. Participation correlates with how the operator is performing and how they feel about head office this quarter. The operators who ignore the request are not a random subset.
- Which systems captured the transaction. Digital-first customers are over-represented because digital systems are usually the brand-operated ones. Walk-in and phone customers are systematically missing.
- Which records the operator is willing and lawfully able to share. Consent and lawful basis were usually obtained by the operator, in the operator's name, for the operator's purposes.
The result is a sample that looks national and behaves local. You can weight it, and survey weighting will fix the demographic skew you can see, but weighting cannot repair a frame that is missing whole channels of purchase. It adjusts the people you reached. It does not conjure the people no system recorded.
What this is not
Two adjacent methods get confused with this one, and keeping them separate saves a lot of wasted fieldwork.
It is not mystery shopping. A mystery shopper is a trained observer standing in for a customer, sent to check whether the standard was executed. That is compliance measurement, and it answers a question about the operator. Franchise customer research asks real customers what they experienced and why, which is a question about the brand promise.
It is not partner satisfaction research. Partner or franchisee satisfaction treats the operator as the respondent and the relationship as the subject. That is a genuinely useful programme, and larger brands should run one, but it tells you how your operators feel about you. It tells you nothing about what the customer experienced. Nor is this the same as proxy research, where you interview someone adjacent because the user is unreachable. Here the customer is reachable. Somebody else simply holds the phone number.
How to run it anyway
The design that works inverts the usual flow. Instead of asking operators to supply customers, the brand supplies the instrument and controls the draw.
- Recruit at the point of experience, not from the operator's list. A QR code on the receipt, the packaging, the service ticket or the confirmation screen puts the invitation in front of every customer of every participating unit, on terms the brand sets. The operator does not choose who sees it.
- Record the unit as a variable, not as a stratum. Capture which location the experience happened at, but do not build the sample by location quota. Location quotas make the brand look uniform when it is not.
- Ask about the experience, not about the operator. Customers cannot reliably tell you whether a standard was followed. They can tell you what happened, what they expected and what they did next.
- Interview, do not just score. A location-level number tells you which unit is different. Only a conversation tells you why, and in a multi-unit brand the why is usually a method-of-operation detail that head office can fix once and deploy 845,000 times.
- Run it continuously, not annually. Unit-level operations change fast. An annual brand tracker describes a network that no longer exists.
Step four is where most multi-unit programmes stall, because interviewing at brand scale used to mean hiring moderators in proportion to locations, which nobody can afford.
Where Koji fits
Koji was built for exactly the shape of this problem: a very large number of customers, spread across a very large number of places, none of whom will join a scheduled video call.
- AI-moderated voice interviews run at whatever scale the brand needs, in parallel, at any hour. A brand with 400 locations can run 400 conversations tonight without booking a single moderator. This is the difference between a satisfaction score and an explanation.
- No moderator bias, and no operator in the room. Every respondent gets the same instrument, asked the same way. In a multi-unit business that matters more than usual, because the alternative is a franchisee handing a tablet to a customer they have just served.
- Six structured question types run alongside the open conversation: open_ended, scale, single_choice, multiple_choice, ranking and yes_no. You get a comparable location-level number and the transcript that explains it, from the same session, rather than running a quantitative tracker and a qualitative study six months apart. The full breakdown is in the structured questions guide.
- Automatic thematic analysis and one-click reports roll several hundred conversations into themes without a coding team, so the finding reaches the operations director while it is still true.
- Customizable AI consultants let you brief the interviewer on your category. A quick-service restaurant brand and an automotive service network need different probes, and neither needs a research degree to configure.
The legacy alternatives were built for a different company shape. Qualtrics and Medallia are enterprise experience suites designed around a single corporate owner with a central data warehouse, and they price and deploy accordingly. UserTesting and dscout give you access to a panel, which is useful when you need strangers and useless when you need the person who actually ate at unit 412 on Saturday. Typeform and SurveyMonkey will collect a response, but a form cannot ask a follow-up question, and in multi-unit research the follow-up is the entire value. Koji is the AI-native option: from question to insight in hours rather than the six-to-twelve weeks a traditional multi-market fieldwork cycle takes, with no research expertise required to run it.
Brands running this in specific verticals may also want the sector guides for retail and for restaurants and hospitality.
The two things that break this next
Getting the sample frame right is the first problem, not the only one. Two more are waiting.
The first is what happens when you tie the resulting scores to the operator's scorecard, which is the obvious next step and the one that quietly destroys the data. We cover that in multi-location customer feedback.
The second is what you are allowed to do once the finding is in. In a franchised network the person who must act works for a different company, and the channel through which you can require action is narrower than most brand teams realise. That is covered in brand standards and customer research.
Frequently asked questions
Who owns the customer data in a franchise?
Whichever party operates the system that captured it, unless the franchise agreement says otherwise. The FTC Franchise Rule definition at 16 CFR 436.1(h) gives the franchisor authority over the franchisee's method of operation, not ownership of the customer relationship. Brand-operated apps, ordering flows and loyalty programmes generally produce brand-held records. Point-of-sale, walk-in and phone transactions generally produce operator-held records. Most brands have both and can reconcile neither.
How many franchise establishments are there in the US?
The International Franchise Association's 2026 Franchising Economic Outlook projects 845,000 US franchise establishments in 2026, up from 832,521, supporting nearly 8.9 million jobs and $921.4 billion in output.
Can head office survey franchise customers directly?
Yes, if the invitation reaches the customer at the point of experience rather than through the operator's contact list. A brand-controlled QR code or confirmation-screen link avoids relying on operator-held records and the consent that was collected in the operator's name. Pulling customers from an operator's database raises lawful-basis questions that vary by jurisdiction and by what the operator told the customer at the point of collection.
Why not just use the operator's own customer list?
Because you inherit both the gaps and the selection. Operator lists over-represent digital and repeat customers, exclude whole purchase channels, and are handed over by operators who chose to participate. The union of those lists is not a sample of your customers, and no amount of weighting turns it into one.
What is the difference between franchise customer research and mystery shopping?
Mystery shopping sends a trained observer to check whether a standard was executed, which measures the operator. Franchise customer research asks real customers what they experienced and why, which measures the brand promise. Both are useful, and they answer different questions. See the mystery shopping guide for the compliance side.
How often should a multi-unit brand run customer research?
Continuously rather than annually. Unit-level operations, staffing and local competition change on a timescale of weeks, so an annual tracker describes a network that has already moved. Running a rolling programme of AI-moderated interviews keeps the picture current at a cost that does not scale with the number of locations.
Run brand-level research without asking your operators for permission
Koji lets a multi-unit brand interview its own customers at scale, with a brand-controlled instrument and a brand-controlled sample, and get themes back the same week. No moderator bookings, no six-week fieldwork cycle, no dependence on which operators felt like participating.
Start a study with Koji and run your first hundred customer interviews this week.
