Customer insights can ¨live¨ in various places
I was leading a complex data product, a voice of the customer platform, an internal RAG (retrieval-augmented generation) product covering user patterns, customer feedback and anomalies. That work shaped how I think about customer evidence. Voice of the customer data matters for the whole customer experience. That experience is everything a customer meets: the website, the product, anything behind login, and what happens before, during and after the purchase.
So when a team asks me to look at an ad, a page or an onboarding step, I start with the wider path. I look at what people expected, what they experienced and whether they were the right people to bring in. Sometimes the page does need work. In my experience, the evidence just as often points to a disconnect between steps, where one step sets an expectation and the next does not deliver it.
This article looks at Voice of Customer as part of broader customer evidence, how that evidence can inform marketing, messaging and product decisions, and why the learning continues after each change.
What Voice of Customer means
Voice of Customer (VoC) is customers' own description of their needs, expectations, experiences and objections.
One useful distinction comes from Griffin and Hauser's 1993 paper in Marketing Science, which describes a customer need as "a description, in the customer's own words, of the benefit to be fulfilled". A need is not a solution or a feature specification. A customer asking for a feature is usually telling you about a benefit they are missing, and the work is to understand which one.
The material comes from what people say on Trustpilot and other review platforms, and in customer service tickets. When someone returns an item to a vendor, they say why they are returning it and whether they want a refund. In a SaaS product or an app, people complain about the subscription or about expectations the product failed to meet, so feedback from cancellation and churn belongs in the same place. Add any third-party information, surveys and the customer interviews the user experience team runs, and the picture becomes much fuller.
All of this information needs to be digested and organised together. Everyone in a company, bigger or smaller, has access to customer insights like these. Gathered into one corpus, they become something marketing, product and customer teams can all draw on.
New companies may not have much of their own yet. They can gather adjacent customer objections from competitors and from the market, and turn them into a corpus they use alongside their audience investigation and matching. Those voices belong to someone else's customers, so I treat them as leads to check against the people the business wants to serve.
How to collect and organise that corpus, including where AI helps, deserves its own article.
Customer evidence: what people say, what they do and what happens next
VoC is one part of the picture. It is important to see what people say, what they do and what the outcomes are, and to check whether those customers are the ideal customers.
-
What people say
Reviews, support tickets, return reasons, churn feedback, surveys and customer interviews explain what people expected, what they doubted and how they describe their situation.
-
What people do
Website and product data from tools such as Microsoft Clarity, Hotjar, PostHog and Amplitude shows what people do on a site or within a product, from heatmaps to user flows. Mapped to the customer journey, it helps locate where friction sits.
-
What happens next
Outcomes show whether people bought, returned an item, upgraded, kept using the product or cancelled.
Each kind of evidence answers a different question. Nielsen Norman Group's guide to research methods separates "what people say" from "what people do", and notes that self-reported data is limited by what people are aware of and willing to report. Behaviour shows where people struggle, but it rarely explains why on its own.
The ideal customer check is easy to skip. Someone can come into the funnel and still end up unhappy, because we failed their expectation after targeting the wrong person in the first place. Their feedback is real, but it may say more about targeting than about the product. Voice of the customer data therefore connects to the whole customer journey and every touchpoint in it, not to a single page.
Internal subject matter expertise matters too, in an established company or a startup. Everyone inside knows the product, and that knowledge helps interpret the evidence. We still need to observe user behaviour, because what the team believes and what users actually do can disagree.
Using customer evidence across the journey
Based on that evidence, we can look at user patterns and apply them to several connected decisions.
Paid marketing
Customer evidence shapes audience hypotheses and angles around recognisable situations, desired outcomes and objections. AI can help cluster user patterns and turn them into new avatars, and we may discover new avatars worth testing in paid marketing. The clustering needs to start from real customer data, because averages hide differences. In one example from Nielsen Norman Group, a 65% bounce rate split into 83% for one segment and 37.5% for another.
Messaging
The same patterns show how to explain the offer in customer language and how the messaging on the website should change. They also show where a promise needs to hold throughout the journey.
Imagine a booking journey where customers expect a personalised tour. The messaging promises this all the way through, but at the checkout step the flow looks like any other website. There is no human touch, no relationship and no personalised experience at the point where the customer commits.
Commercial search journeys
Customer data shows which commercial queries people use and how they use them. That helps map each query to the right page, and shows where the same language plays a part in microcopy for conversion and for adoption across the journey.
I used customer language in this way when I worked with a hair supplements brand. The way customers described their concerns helped us shape our search and content strategy and turn it into SEO.
Website conversion and onboarding
A high-volume B2C SaaS job interview assistant was getting a big top of the funnel from template-type queries, through SEO and paid search traffic. What I discovered was a disconnect between the paid search ad and the landing page. The ad used the messaging customers needed, interview questions, but the page did not reflect that promise. People really wanted to get ready for their interview, and then we overwhelmed them with onboarding steps. The better route was to push them into the flow and show them in the messaging how they could start their interview with no anxiety.
Google Ads looks at the same connection when it rates how relevant a landing page is to the people who click the ad. Extra steps before the real task also tend to cost more than they give. In a usability study of 70 users, people who saw an upfront tutorial did no better at the tasks and found them harder.
In service businesses, I have seen examples where we had enough evidence that a human touch, such as a brokerage-style one, was necessary in the journey. What we needed to change was the sequence and the positioning: the section itself and its content had to move further up. Where that human contact sits matters. A study in MIS Quarterly found that a mix of self-service and personal service was linked to the lowest customer defection, especially early in the relationship.
Website journeys like these are the focus of my Website Strategy & Optimisation work.
Product adoption
Customer evidence also compares what attracted people with what onboarding helps them achieve, and where they first see value.
I had a client with a funnel from freemium to paid, and their main challenge was converting free customers to paid plans. I gathered information from third-party platforms and looked at what hobby users would use the product for, compared with large organisations that needed rendering work. I mapped that language so we could change the product's aha moments, helping customers see the value of the product, spend credits and convert. Onboarding, activation and trial-to-paid conversion are the focus of my Product Growth & Monetisation work.
Evidence points to a hypothesis: judgement sets the priorities
Customer evidence supports hypotheses, but it rarely proves a cause by itself. Several signals pointing the same way make a hypothesis stronger, because each source covers what another one misses.
AI can help organise large volumes of feedback and behaviour, but the priorities still need human judgement. I relate each pattern to the ideal customer profile, the offer and what the business needs to improve. A frequent complaint from people outside the target market can matter less than a serious barrier for the customers the business most wants to serve.
Continuous optimisation: why learning continues after each change
Customer expectations change and traffic sources shift. SEO versus AI search is a clear example, but the product or website also evolves. Every change a team makes therefore introduces new assumptions about how users are expected to behave.
Those assumptions are often less certain than they feel. In a 2009 case study, Microsoft's experimentation team reported that of well-designed experiments meant to improve a key metric, only about one-third did. That was at a large company with plenty of data and experience.
Each change needs a continuous learning loop. An experimentation and optimisation programme helps, but it is not only A/B testing. A website or journey can have plenty of traffic overall, while an A/B test needs enough traffic for the specific feature we want to test. Where that traffic exists, we align an A/B test with it. A lack of it does not mean we skip optimisation. We handle it in a different way, through an experimentation programme that still learns from each change and improves the experience and the commercial impact.
It is a loop. We learn from each experience, even when a change does not move the needle for the right KPI.
Measure the outcome against business
Volume alone can mislead. More form submissions can create extra work without better opportunities, and more signups say little if people never reach a meaningful result in the product.
Customer journey diagnostics brings these pieces together: what people say, what they do, what happens next, which customers the business wants to serve and what it needs to improve. The useful output is a prioritised decision, supported by evidence, with a clear way to learn from what happens after it ships.
Frequently asked questions
What is Voice of Customer?
Voice of Customer is customers' own description of their needs, expectations, experiences and objections. It comes from sources such as reviews, support tickets, return and refund reasons, cancellation and churn feedback, surveys and customer interviews. It is most useful when it describes the benefit people are looking for, rather than a specific solution.
How is Voice of Customer different from behavioural data?
Voice of Customer captures what people say. Behavioural data, from tools such as Microsoft Clarity, Hotjar, PostHog or Amplitude, shows what they do. What people say helps explain expectations and doubts, and behaviour helps locate where people struggle. Looking at both, alongside outcomes, gives a more reliable basis for a hypothesis.
Can customer feedback inform marketing as well as product decisions?
Yes. The same evidence can shape audience angles in paid marketing, the language on a website and the commercial search queries a site answers. It can also shape the sequence of a page and the moments in a product where people first see value. Each step needs to set and fulfil a consistent expectation.
Do you need A/B testing to optimise a customer journey?
No. A/B testing works where a specific feature has enough traffic to test. Without that traffic, an experimentation and optimisation programme can still learn from each change, gather new feedback and improve the experience. Many well-reasoned changes do not improve the metric they were meant to, so the learning loop matters either way.
Turning customer evidence into product decisions
Customer feedback becomes most useful when it is treated as the start of a decision, rather than the decision itself. Before changing the ad, the page or onboarding, look at what people said, what they did and whether they were the right people to bring in.
If your team has customer feedback, behaviour data and a list of possible changes, I can help turn them into prioritised product and growth decisions. Let's discuss a fractional or interim product role.