Voice of the customer has two meanings. In product work it is a research method that turns customer interviews into a ranked list of needs. In customer experience it is a program that gathers feedback from many channels and acts on it. The first meaning is older and better tested.

What is voice of the customer?

A marketing encyclopedia defines it as a process for capturing customers’ requirements (Gaskin et al.). The entry lists a detailed set of wants and needs. They are sorted into a hierarchy. Each is then ranked by importance and by how well current options meet it.

The chapter, by Steven Gaskin, Abbie Griffin, John Hauser and two colleagues, names four parts. They are the needs, their hierarchy, their priorities and the customer’s view of how well each need is met.

A customer need is a description, in the customer’s own words, of the benefit to be delivered. That rule keeps design guesses off the list. “A room that stays quiet” is a need. “Triple glazing” is a design answer.

Where does the voice of the customer come from?

The phrase comes from quality function deployment, or QFD. QFD takes the voice of the customer into research, engineering and factory work (Griffin and Hauser, 1993). Its main tool, the house of quality, began at Mitsubishi’s Kobe shipyard in 1972 (HBR, 1988). Yoji Akao co-founded QFD. A 1983 article in Quality Progress first took it outside Japan (Akao tribute).

How do you structure customer needs?

Discussions with customers usually produce 100 to 200 phrases that state a need. Hauser’s MIT note sorts them into three kinds. Basic needs are the ones a customer assumes. Articulated needs are the ones a customer says aloud. Exciting needs surprise and please the customer when met. The needs stay in the customers’ own words (MIT note).

No manager can work from 200 phrases. So the method builds a hierarchy. The top level holds 2 to 10 primary needs, which set strategy. Each splits into 3 to 10 secondary needs, which set tactics. Tertiary needs sit below them (MIT note).

For illustration, a hotel program finds one primary need, easy rewards. Its secondary needs are fast posting, a clear view of tier progress and partner purchases that count. Under fast posting sits a tertiary need: points that show up before the guest leaves. The structure tells the team what to fix first.

How many customers do you need to interview?

Griffin and Hauser studied the method in 1993. In a typical study, they wrote, 10 to 30 customers are interviewed one on one for about an hour. Their data showed that interviewing 20 customers finds over 90% of the needs that 30 customers would give (Griffin and Hauser, 1993).

Format matters less than hours. In their data, two one-on-one interviews found about as many needs as one focus group, 51% against 50%. Four interviews matched two groups, 72% against 67% (Griffin and Hauser, 1993).

Counting mentions does not rank needs. The same study found that frequency of mention does not appear to be a good surrogate for importance. Ask members to rate each need, then rank by the ratings (Griffin and Hauser, 1993). For illustration, a program codes 600 member comments. “Points never post” appears 120 times, or 20%. “Status benefits are unclear” appears 60 times, or 10%. A follow-up survey asks 300 members to rate each need from 1 to 9. Posting speed scores 6.8 and benefit clarity scores 8.1. The less-mentioned need ranks first.

Interviews cost real money. One major US firm put the cost of an interview, counting team time, at $1,000 to $2,000. Across 5 to 10 segments and 5 to 10 product lines, cutting from 30 interviews per segment to 20 saves $250,000 to $2,000,000 (Griffin and Hauser, 1993). Those are 1993 dollars.

What are the sources of voice of the customer data?

TechTarget splits the inputs in two. Indirect input is data that reflects customer behavior, such as transactions and product use. Direct input is data that reflects what a customer says, such as interviews and social posts (TechTarget). The table sets each source against its strength and its blind spot. Our measuring customer loyalty guide covers the survey scores.

SourceLoyalty program exampleStrengthBlind spot
InterviewsWhy a lapsed member leftNeeds in the member’s own wordsSmall samples, and no ranking of importance
SurveysImportance ratings, CSAT, NPSRank needs across many membersSelf-selection in satisfaction scores
Open channelsReviews, social posts, support contactsUnprompted and specificOnly people who choose to write
BehaviorEarns, burns, app eventsCovers every memberShows what members do, not why

Satisfaction scores carry a known bias. Griffin and Hauser found that a score taken from customers of the brand they chose contains self-selection. In their data, the brand with the highest satisfaction was a small niche brand. They warned that such measures can create counterproductive incentives (Griffin and Hauser, 1993).

How do loyalty programs use the voice of the customer?

Delta changed its SkyMiles plan after member feedback. In October 2023, CEO Ed Bastian wrote that many members had shared feedback on the September changes. Concerns centered on how reachable Medallion status was, limits on Sky Club access and recognition of long-term customers. Delta then cut its proposed 2025 status thresholds. Silver fell from 6,000 to 5,000 Medallion Qualifying Dollars. Gold fell from 12,000 to 10,000, and Platinum from 18,000 to 15,000. Diamond fell from 35,000 to 28,000 (Delta).

Starbucks ran an idea community, My Starbucks Idea, for five years. By March 2013 it had received 150,000 ideas and launched 277 of them. Starbucks counted digital rewards for using the Starbucks Card among the results (Starbucks).

Hyatt said its World of Hyatt program was designed based on listening and taking action (Hyatt).

J.D. Power runs the voice of the customer at scale for loyalty programs. Its 2016 hotel study weighted six factors. Account maintenance carried 23%, ease of redeeming 22% and ease of earning 18%. Variety of benefits and program terms carried 16% each. Customer service carried 5% (J.D. Power, 2016). The list shows where members put their weight.

The method also works on partners. Hauser notes that the same methods have been used to learn what an organization’s own people want, called the voice of the employee. They have also been used for the channel and the supply chain, called the voice of the collaborators (MIT note). A program with a partner ecosystem can interview its partners the same way. Partners are customers of the program too.

Forrester ties the habit to results. Its 2024 US index found that only 3% of companies are customer-obsessed. Those firms said revenue grew 41% faster, profit 49% faster and retention 51% better than at the rest (Forrester). The figures come from a survey.

How GRAVTY handles the voice of the customer

GRAVTY holds the behavior half of the picture. It records each earn, burn, tier change and offer response as a timestamped event tied to a member (retention economics).

Agentic AI Compass pulls KPIs, partner data and customer sentiment into one place (Compass). A team can ask it why a metric moved.

Frequently asked questions

What is the difference between voice of the customer and customer feedback?

Feedback is the raw input. Voice of the customer is the organized result. It sorts needs into a hierarchy, ranks them by importance and tracks how well each is met (Gaskin et al.). A pile of survey comments is feedback. A ranked list of needs is the voice of the customer.

Where does the phrase voice of the customer come from?

It comes from quality function deployment, or QFD. QFD took the voice of the customer into research, engineering and factory work (Griffin and Hauser, 1993). Its main tool, the house of quality, began in 1972 (HBR, 1988).

Are surveys or interviews better for voice of the customer research?

They do different jobs. Interviews find the needs in customers’ own words, and surveys rank how important each need is. Griffin and Hauser found that survey measures of importance can predict how customers react to product concepts (Griffin and Hauser, 1993). Use interviews first and surveys second.

Can voice of the customer data be biased?

Yes. Satisfaction scores come from customers of the brand they chose, which builds in self-selection. In one data set, the brand with the highest satisfaction was a small niche brand (Griffin and Hauser, 1993). Behavior data covers every member, so read it beside the survey scores.

Are NPS and CSAT part of voice of the customer?

Yes, as two of its metrics. TechTarget lists customer satisfaction score and Net Promoter Score among the important VOC metrics (TechTarget). They are signals inside the program. They do not replace the needs list. Our entries on Net Promoter Score and redemption show two places to ask.