How to Conduct B2B SaaS Product Feedback Interviews

If your homepage copy was written from the founder's memory of the first five customers, this is for you.

‍

Most lean SaaS teams have never run a structured customer interview. Feedback arrives as support tickets, a Slack message from the AE after a lost deal, and whatever the biggest account said on the last renewal call. The roadmap, the onboarding flow and the messaging all get shaped by that noise. Then conversion on the site stays flat, sales cycles stretch, and churn gets explained as "pricing" because nobody asked.

‍

Ten structured interviews fix most of this. It takes two weeks, costs almost nothing, and needs one person who can hold a script and stay quiet. Here's how to run them.

‍

‍

What ten interviews should actually tell you

‍

Before you book a single call, be clear on what you're trying to learn. There are three questions worth the effort:

‍

1. What job did they hire the product for? Not the category you sell into. The actual situation that made them go looking.
2. What value do they get, in their own words? The phrases they'd use to describe it to a peer, not the ones on your pricing page.
3. Why do people leave? What changed, what replaced you, and what would have kept them.

‍

Notice what's not on the list: "What features would you like?" Customers are reliable narrators of what happened to them and unreliable predictors of what they'd do. Ask about the past, not hypotheticals.

‍

Surveys and NPS can't answer these questions. They tell you what (a score, a feature name) but never why. Support tickets skew toward friction. Sales call notes skew toward objections from people who didn't buy. Interviews are the only input that gives you the full story from someone who did.

‍

A note on fit: if you have fewer than ten paying customers or no active sales motion, don't formalise this yet. Founder-led sales calls are your interviews. This playbook is for teams with a defined product, real customers and enough churn to learn from.

‍

Pick three cohorts, not a random sample

‍

Random sampling gives you random findings. Pick three cohorts of three to four people each. Each one answers a different question.

‍

‍

Pull the lists from HubSpot using lifecycle stage, close date and last activity. If your lifecycle stages aren't reliable, pull from billing and your product analytics instead. Don't let a messy CRM delay this.

‍

Getting a yes. The founder sends the invite personally, from their own inbox, three lines long. Say it's a 30-minute conversation, not a sales call, and that you want to understand how they work. For churned accounts, offer a $75–100 gift card and mean it when you say there's no pitch. For current customers, a promise to share what you learn is usually enough. Expect roughly one in three to accept. Send fifteen invites per cohort.

‍

The script: ask about moments, not opinions

‍

Opinions are cheap and often wrong. Moments are specific, remembered and useful. The whole script is built around getting people to replay what happened.

‍

‍

Questions for recent wins and retained users

‍

• "Take me back to before you started looking. What was happening that made this a problem worth solving?"
• "What were you doing to handle it before? What did that cost you in time, money or headaches?"
• "What else did you look at? What made you drop the others?"
• "Walk me through the first week after you signed up. What did you do first? Where did you get stuck?"
• "When did it click? What were you doing when you first thought, this is worth paying for?"
• "What do you do in the product every week? What would you miss if it disappeared?"
• "If a peer asked why you use this, what would you say in one sentence?"

‍

Questions for churned and lost accounts

‍

• "When you first signed up or took the demo, what were you hoping would change?"
• "What happened in the first few weeks? Where did it fall short of that?"
• "Was there a specific moment you decided it wasn't going to work?"
• "What are you using now? What does it do that we didn't?"
• "What would have had to be true for you to stay?"

‍

What to avoid

‍

• Hypotheticals: "Would you use a feature that…" You'll get a polite yes that means nothing.
• Leading questions: "You probably found setup easy?" You've just told them the answer.
• Pitching or defending. If they criticise something, say "say more about that" and write it down. Explaining why they're wrong ends the interview.
• Feature wish lists. If they volunteer one, ask what they'd do with it. The underlying job matters; the feature request doesn't.

‍

Every follow-up should be one of: "What happened next?", "What did that cost you?", "Say more about that." That's enough.

‍

Running the call without contaminating it

‍

Two people if you can manage it: one asks, one takes notes. The interviewer should not be the AE who sold the deal or the CSM who owns the renewal. People soften feedback when the person who will be hurt by it is on the call. The founder or marketing lead is the right interviewer at this stage.

‍

Record it, with consent, and run it through a transcription tool. You'll need the exact phrasing later, and you can't capture it while listening properly.

‍

Hold the 30 minutes. Running long signals the next invite will be a time sink.

‍

‍

Two closing questions

‍

• "Who else do you know who deals with this problem?" You get referrals and a signal about where your real ICP lives.
• "If you were describing this to a peer in one sentence, what would you say?" This is where your next headline comes from.

‍

Within an hour of the call, the note-taker writes a five-line summary: trigger, alternative considered, the "click" moment, the value phrase, biggest friction. Do it before the next interview blurs them together.

‍

Turning transcripts into three decisions

‍

Ten transcripts are useless until they're tagged. You don't need research software. A spreadsheet with one row per quote and five columns works:

‍

• Trigger – the event that started the search
• Alternative – what they used or evaluated instead
• Value phrase – how they described the outcome, verbatim
• Friction – where they got stuck or nearly left
• Feature mentioned – what they actually use

‍

Go through each transcript and pull every quote that fits a column. Then count. Anything said by three or more people across cohorts is a pattern. Anything said once is an anecdote, no matter how compelling.

‍

The patterns feed three decisions:

‍

1. Messaging. Compare your homepage nouns to the value phrases in the sheet. Most teams discover they're selling "workflow automation" while customers describe "not having to chase people on Friday afternoons." Replace your internal language with the customer's verbs. If the recent-wins cohort consistently names one alternative, your positioning should say what you do that it doesn't.

‍

2. ICP. Look at the retained cohort. Role, company size, trigger event and tooling stack. That's your real ICP, and it's often narrower than the one in your pitch deck. If the churned cohort shares a profile that's different from the retained one, that profile belongs on a disqualification list, not in your outbound. This is usually the finding with the most revenue impact, and the one teams resist most.

‍

3. First-week onboarding. The retained cohort described the moment the product "clicked." The churned cohort described where it didn't. Map your current first-week path against both. Most onboarding flows push every feature in order of when the team built them. Rebuild it to reach the click moment as fast as possible and cut everything that doesn't get the user there.

‍

Write all three up in a one-page document with the supporting quotes. Share it with sales before you share it with anyone else. If the AEs read the quotes and say "yes, that's what we hear," you've got alignment. If they don't, you've got a different problem worth knowing about.

‍

Then put a repeat on the calendar. Three interviews a month keeps the sheet current without anyone having to run a research project again.

‍

Conclusion

‍

Ten interviews won't tell you everything. They'll tell you what customers actually hired you for, how they describe it, and why some of them left. That's enough to fix messaging that doesn't convert, an ICP that's too wide and an onboarding path that pushes the wrong things first. The budget is a few gift cards and twenty hours of someone's time.

‍