How to Conduct B2B SaaS Customer Discovery Interviews

Customer research · Customer interviews

Customer interviews have always been one of the highest quality forms of information available to a business, and with the advent of AI they matter more rather than less: a model can synthesize everything already written about your market, but it cannot produce the sentence your customer would say about your product. They are also difficult to get right, and harder in B2B SaaS than almost anywhere else. Customer bases are typically too small to sample. Purchases can be rare, slow and complicated. And because of the politics involved in a purchasing process, a plethora of stakeholders can be involved, some of whose standing depends on how the purchase turns out.

What are Customer discovery interviews for?

Customer discovery interviews lay the groundwork for future customer and product understanding

Customer discovery interviews are one of the first tools to reach for when you are establishing a go to market approach for a B2B SaaS company. They are where you take the first steps toward your ICP, the personas you want to target, your value propositions and the whole of your positioning and messaging, and along the way they collect the texture that shapes product development and boardroom strategy.

When you run discovery depends on where your company is. Before you have paying customers, discovery means talking to prospects to test whether a problem is real and painful enough to build for. Those people are harder to find, and they cannot tell you anything about a product they have not bought, but at that stage they are the people you have. Once you have customers, you can talk to people who have already chosen you. They can tell you what they did: why they started looking, what they compared you against, and what nearly stopped them. As we explain in our overview of customer interviews, that is much stronger evidence than what a prospect says they would do.

Most of the companies we work with already have paying customers, so this guide is written for that later stage. Almost everything in it applies earlier on as well. Where a question asks about a purchase, ask a prospect how they handle the problem today instead.

So, how do you conduct one?

When Should you conduct customer discovery interviews?

Conduct discovery interviews when you have no hypothesis

You should conduct customer discovery interviews when you have no hypothesis, or one you distrust, and you need to understand the perspective of your market before you write your positioning or messaging. You should almost always run these interviews before anything else. They supply you with the vocabulary and the frameworks you need to do follow up research, whether that is more interviews, surveys or product testing.

Who should you talk to in a Customer discovery interview?

Customer discovery interviews should be aimed at existing customers

If you don’t know exactly which of your customers to interview, get a spread of customer types where possible: firms of different sizes, revenues, business models and geographies. When you are conducting research it is normal to worry about confirmation bias, and to feel that you need to interview not only your evangelists but also your detractors. We find it valuable to talk to detractors in order to understand your anti-ICP, and who you shouldn’t go after in future product or service go to market strategies, but once you have heard from a few of them you have the information you need. It is more important to talk to customers with whom you have positive relationships, because understanding them better will help you develop a firmer and more subtle grasp of your ICP so you can nail your niche. If you can get 10 interviews or so that is acceptable, but 20 or more is ideal. More on this in the FAQs.

What does a typical Customer discovery interview look like?

The standard 30-minute format

An ideal customer discovery interview has six sections. Aim for 30 minutes with a customer at the very least, although more is always better. I typically conduct a 45 minute version that runs through the same six sections with one or two more minutes each, mostly in sections 2 and 3. We'll run through all the sections, what their goals are, what you should be asking, and what you're looking for, below.

Section 1 · 3 minutes

Warm up and context gathering

Section 1 is where you both warm up and where you gather your initial context. You will introduce yourself and get to know your customer if you don’t know them already. This section is intuitive, and because of that people frequently wing it, missing the opportunity to do important foundational work for the rest of the interview. Be sure to determine your customer's proximity to your product, and their role in decision making and budget allocation within their firm. Proximity shapes the nature of their needs. A C-suite executive will want your tool to provide return, while a product manager who uses it daily might pine for a feature they know would fix a bottleneck in their team. At Kalungi we typically consider three types of persona within an ICP to categorize these types of proximities:

Your goal in these first few minutes is to learn which of these roles your interviewee occupies, and to understand the shape of gaps in your market knowledge. In customer discovery interviews earlier in the research process this section can run longer, as you gather information about the general market.

Ask two qualifying questions here as well (they'll inform which version of section 2 you use): 

Section 2 · 7 minutes

The switch story

Now that you have some general background it is time for you to dig into their switch story. The switch story is their narrative understanding of how they arrived at, or switched to, using your product. The goal here is to work out why your interviewee is using your product, or whichever product you are competing with or interested in for the sake of this research. This will be your most valuable section for positioning and messaging. We want to stay away from hypotheticals and your questions need to be grounded in things that have already happened.

There are four switching forces. Depending on the shape of the interview the four will not be dealt with equally, and some may be left out entirely. The workaround for this is to talk to different types of customers, so that when you aggregate your knowledge at the end of the process, you have insight into all four forces, as all four are necessary for a full understanding of how your product behaves and is perceived in the market. We've listed the four forces in the cards below.

the four switching forces

The four forces have distinct functions later on, so you want to do your best to keep answers to the four forces mentally separated, even though you might be asking questions and having conversations that cross all four at once. Push and pull help you focus your messaging, anxiety lets sales and onboarding update their processes, and habit helps you develop your roadmap and combat churn risk.

You will typically need two versions of section 2 ready, one for customers who switched to your solution recently and one for long-tenured customers. You usually cannot ask about the switching forces by name, so ground them in a specific episode instead.

I like to ask the following questions:

The thing to listen for across those is the difference between lock in and preference. A customer who stays because leaving is expensive and a customer who stays because you are the best will both say they are happy, but they have very different implications for your relationship and ability to deepen revenue streams.

Section 3 · 7 minutes

Current value and frustration

Section 3 deals with the current value of your product and customer frustrations. It is where you learn what the product is worth to your customers in practice. Again, we try to ground it in an episodic narrative.

The point is not to ask whether they are satisfied. A specific story is far more powerful than a general complaint or some praise. The particular steps, timing, and thought processes of the story are crucial. Without details like that, a complaint isn't very actionable. But feedback becomes a lot more useful if you know exactly where in the process the hurdles are. They might not even be where the customer thinks they are; you might pick up on little details they think are insignificant. If they just give you a general complaint or some compliments, you might even misattribute it to the wrong step or function in a solution or process.

If you can get a specific story about how your product saved them time or effort, you have a nascent business case on your hands. When someone says reporting is clunky you have nothing, but if you ask when it last happened, how often it happens and what they did about it, you have the makings of a feature request you can take to your product team immediately. 

Take your time with those first few questions, and feel free to ask lots of follow-up questions to flesh out your customer's narratives. Once you have dug into those two questions, close the section with a three part question about dependency.

These questions feed your value proposition and your return on investment language on one side, and the roadmap, the churn risk register and your objection prep on the other.

Section 4 · 5 minutes

Cross-sell and ecosystem

Section 4 covers cross sell and wider ecosystems if you offer multiple distinct solutions or have a platform, and it is where you try to place yourself in the customer’s budget without making them clam up by directly discussing price. We recommend a comparative question:

Nobody has to say a figure for you to learn a great deal. A product that sits below the coffee budget has a pretty clear ceiling (unless perhaps you work in the coffee business); a product that sits alongside payroll systems has a different one, and quite a different set of people paying attention to it when it has to renew.

This is an unavoidably vague question and the answers will be vague too, so listen for the cues. Do they describe you as one line item or several? If the customer holds three of your products in their head as three separate purchases with three separate justifications, that matters a great deal when you are trying to establish a platform narrative.

Next, you'll want to learn what drives their use of your solution. Why do they pay you? We call this the "engine". There is a chance it was already shared with you, but if it wasn't you should ask:

This is the section to sacrifice when the call runs long, which it often does. If you are a platform, though, or you hold market share inherited from a legacy product, these questions are essential for separating real adoption from noise in your usage numbers.

Section 5 · 5 minutes

Pricing perception

Section 5 is about pricing perception, and won't be necessary in every case, but it is hard to do without if you are developing a pricing strategy that has little precedent, as you might when you open a new segment or launch a new product. Depending on how similar your solution is to other options on the market, you may already have a ballpark idea of pricing, which you can use to bound these questions and to shape the quantitative research that follows to validate them, whether that is a survey, an alpha test, or something else.

This is also the section most likely to corrupt your data if you handle it carelessly. We try very hard to use a simple rule to avoid this corruption: never say a number first. Once you have named a price, every answer that follows is calibrated against your figure and, fatally, you have no way of knowing by how much. Unfortunately it is a rule you will have to break fairly often if you are charting unknown territory. If you genuinely have no alternative, lead different people in different directions across your interviews, high in one call and low in the next, write down which way you led each time, and account for the bounds you introduced by balancing the number of customers within certain segments that you present each price to. For example: if you have six interviews with SMB startups, ask three about price A and the other three about price B. It sounds obvious, but don't ask all of your SMB customers about price A, and all your enterprise customers about price B.

Since this is quite an emotional and sensitive topic, and since it is so vulnerable to bias, we want to pay special attention to behavior in this section. To do that, we first need to start with a grounding question:

Their initial reaction tells you about affect, which matters, but the action, and how they carried it out within the decision making structure of their company, is the real goldmine. Someone who called a partner, someone who went and got a competing quote, and someone who signed the very same afternoon have told you three completely different things about where your price sits relative to what they expected. Where applicable, follow up with questions about the negotiation they pursued inside their own company:

Now that we have some grounding, we move onto prices. Because we so severely want to avoid asking about just one price that we put in the customer's mouth, we ask four questions from the Van Westendorp price sensitivity meter. You ask about four boundaries, one extreme and one plausible case in each direction. Start with the extreme cases to ease the customer into the exercise:

Sometimes we supplement these pricing questions with questions that provide direct links to purchase probability and volume demand, which is called the Newton-Miller-Smith Extension, or NMS for short.

These questions are obviously sensitive, and they are the least reliable in the interview. The leading meta analysis of hypothetical bias puts average overstatement of willingness to pay at about 21 percent (Schmidt and Bijmolt, 2020). You will still need to venture the questions, because the alternative is more or less guessing at prices. Given enough volume, each customer only giving you an answer to one or two of these questions can be sufficient, because you can aggregate it with other customers’ responses. If you do not have enough responses to do that comfortably, pay particular attention to their body language, communication style, and how they react to the questions themselves. It is an imperfect science, but, again, it is far better than guessing. The Van Westendorp questions are directionally powerful, especially when millions of dollars are on the line.

In the worst case customers will be tight lipped about these abstract questions and you will have to present actual numbers and ask what they think. Treat that as a last resort, and even then only do it across multiple interviews so you can balance out the bias you are injecting.

If there is time, ask what unit the capability they would expect to be charged in: per seat, per site, per transaction, per outcome, and which of those would feel fair and which would feel strange. That will help you turn these pricing preferences into a pricing model you can work through.

At the end of section 5 you should have a set of directional price boundaries and a story about how this customer relates to money. You do not yet have a precise set of numbers. Statistical resolution on price comes from a survey or sets of test, which these answers will help you write.

Section 6 · 3 minutes

Validating the decision maker and wrapping up

Section 6 maps the decision makers, if you have not already done so, and closes the call. Ask:

The most interesting question here relates to the person who could have killed it and did not. Your sales team frequently won't hear about them, and your CRM won't record them either, while they are typically the main reason deals or product adoption take longer than hoped for.

People are not very reliable about events inside their own organization: across 127 studies and 11,874 observations, two informants from the same organization correlated only at 0.612 on average across all topics, and the thing they agreed on least was internal matters (Homburg et al., 2012), so Treat what you get here as a hypothesis rather than a map.

If you can, you want to interview two people (or maybe three) per account.

Round off with a referral question.

Then finish with an open closer.

Ask it, then stop talking and count to seven. Even if it is a bit awkward, this question frequently produces the single best insight of the interview.

limitations

What a customer discovery interview does, and doesn't, tell you

Customer discovery interviews tell you how the customer perceives your product, and themselves in relation to it, the four switch forces in sequence, workarounds nobody mentioned, and negative findings. Negative findings can feel like they are beside-the-point, but a competitor you fear going unmentioned across a whole cohort is a valuable result, even if it feels frustrating because it is unintuitive or cuts against the grain.

These calls do not give you give you firm numbers, but they allow you to shape the hypotheses and numbers you will test further down the line. They crucially tell you what your customers care about.

No matter how much we like to think we know exactly what our customers want, unless you ask them, you won't. Customer discovery interviews give you the qualitative attributes and directions that larger scale testing or surveying can then test against afterward.

It is very common for companies to simply misunderstand what their customers want, shaping hundred-million dollar product campaigns around it, only to be perplexed to find out, years later, that customers don't even use their product as intended, or that there is a friction where none was expected. Customer discovery interviews prevent this by giving you the language your customers use, and bounding your expectations based on theirs, not your own. 

If you were to survey 100 prospective customers on a price for a new platform feature, proper discovery would allow you to vet the bounds for the pricing ranges you give, greatly increasing the resolution of your data. You would now be able to understand which attributes your customers are likely to care about, so you know what values to test for in the future. Imagine you thought the interconnectivity of your platform was its key feature, but it turns out that is not only not the main pull, but it drives friction as potential adopters chafe against being forced to integrate it with existing tools you had never even heard of before. Now you can test for that moving forward by asking what tools people integrate with, and how much of a premium they are willing to pay for this interconnectivity, if any, as opposed to assuming what people care about and what they don't.

What you should watch out for

Dos and don’ts in a customer discovery interview

FAQs

Frequently asked questions about customer discovery interviews

How many customer discovery interviews do you need?

It depends what you are establishing. Around nine interviews to know what the problems are called, around 24 to understand what they mean, and 20 or more per segment if you are working across several (Wutich et al., 2024). A systematic review of 16 empirical tests found saturation between five and 24 interviews, with a core range of nine to 17, on populations far more homogeneous than a B2B buying committee (Hennink and Kaiser, 2022). Two rules matter more than the numbers: three independent mentions before anything counts as a pattern, and stop when you stop learning. If ten people give you ten different answers, suspect the segment rather than the sample.

How long should a customer interview be?

Thirty minutes if that is all you can get, 45 is comfortable, and 60 is perfect, but you'll need to prepare a longer battery of questions to make sure you use the time well. The ratio of talking matters more than the length. The customer should be talking about 80 percent of the time, and if you are talking more than a fifth of the call you are probably leading them.

Who should you customer discovery interview, and how many people per account?

Existing customers for discovery work, spread across sizes, models and geographies, weighted toward accounts you have a good relationship with, with a few detractors for the anti-ICP. Where you can, interview two people per account rather than one. Across 104 buying center decisions, models built from multiple informants significantly outperformed single informant models, and the authors concluded that single informant data should be used in organizational buying research only with care (Wilson and Lilien, 1992). One informant per account is the cheapest way to be confidently wrong.

Can you combine discovery with concept or pricing testing?

Not in the same call. The moment you show someone a concept they start describing their world in terms of it, and the discovery material from the back half of that conversation is unusable. Run discovery first, build concepts from what it gives you, then test the concepts in a separate round.

Should you record the interview?

Yes, with permission asked at the top of the call, and say what will happen to the recording. Notes taken live are a summary of what you already understood, which is precisely the wrong filter for a customer discovery interview. Ask separately, before you close, for permission to use their words anonymized in your marketing.

What if you do not know the customer’s industry?

That is an advantage if you use it. A consultant has a strong alibi to be ignorant, and a question that feels too naive to ask out loud often buys you a sentence you would not have gotten any other way. Some version of “we have only been in this space a week, how do you see this company?” costs five minutes of background you do not need and occasionally returns the best material in the call.