Customer onboarding questions to ask 30 days in

The thirty-day check-in is the most wasted meeting in B2B SaaS. Not because it should not happen — the timing is close to ideal — but because of what it usually turns into: a status update, a light feature demo, and a closing "anything else we can help with?" that reliably produces "no, all good."

The customer says all good because that is what people say. Then they churn in month five.

What follows is a script. It is opinionated, it takes twenty minutes, and it is designed for one purpose: to leave you with facts you can act on rather than a sentiment you can report.

Three rules before the questions

Ask about the last time, not about usually. "How do you usually handle month-end?" produces a tidy, idealised description of a process that does not exist. "Walk me through what you did at the last month-end" produces the actual sequence, including the workaround and the spreadsheet. People are unreliable narrators of their habits and reliable narrators of their last Tuesday. This is also why the survey version of these questions fails: a form cannot tell that "usually" was an answer to a different question.

Do not defend the product. The moment you explain why something works the way it does, the interview is over. The customer now understands their job is to react to your explanations rather than describe their experience, and they will be polite about it. If you feel the urge to say "actually, you can do that in settings" — write it down, say it in the follow-up email, and keep going.

Say what the call is for, and mean it. "I'm not here to sell you anything or check a box — I want to understand what the first month was actually like, including the annoying parts." Then honour it. If you pitch at minute eighteen, you have taught them that the next one is a sales call, and the next one will be.

The nine questions

1. Before you bought this, what were you doing instead?

Open here, not with the product. You are establishing the baseline they are unconsciously comparing you against, and it is almost never the competitor you assume. It is usually a spreadsheet, an email thread, or a person. If the honest answer is "nothing, we just lived with it," your onboarding has a harder job than you think, because there is no existing habit to displace and no felt pain to relieve.

2. What was happening that made you go looking?

The trigger. Jobs-to-be-Done practitioners call this the push, and it is the single most useful thing to know about a customer, because it is what the product actually has to resolve. An audit finding, a resignation, a client complaint, a board question. If nobody on your team can name the trigger for a given account, that account is being served generically.

3. Walk me through the first thing you tried to do in the product.

Specific, sequential, and about them rather than about you. Let them talk. The failures live in this answer, and they arrive as asides — "so I exported it, and that took a while because the format was odd, and then I…" Stop them there. The aside is the finding.

4. Where did you get stuck?

Ask it directly and then wait. The silence is the important part. Most people's first response is "oh, nothing really," followed — if you say nothing for four seconds — by "well, there was the thing with the permissions, but that was probably us."

"That was probably us" is the most valuable phrase in onboarding research. It marks a place where the product confused someone badly enough that they blamed themselves, which is exactly the failure mode that never reaches your support queue.

5. What did you do when you got stuck?

The recovery path. Did they search the docs, ask a colleague, email support, or quietly stop? The last one is the one you need to count, and it is the one you will only ever learn by asking, because by definition it generated no signal on your side.

6. Who else was supposed to be using this, and are they?

Seat expansion is the leading indicator everyone tracks and the reason for it is the thing nobody asks about. If four people were meant to be in and one is, find out what happened to the other three. Frequently: nobody told them, or they were told once in a meeting they half-attended, or they tried it, found it slower than their current method, and reverted. Each of those has a different fix, and only one of them is a product problem.

7. What have you had to do outside the product to make this work?

This is the workaround question and it is the highest-yield item on the list. Customers build scaffolding around software constantly and almost never mention it, because from their perspective the spreadsheet is just part of how the job gets done now. Every workaround is a gap, and workarounds that appear in three separate interviews are a roadmap item that will pay for itself.

8. If you had to justify this renewal to your CFO next week, what would you say?

A specific, uncomfortable, concrete question that forces a real answer. If they can articulate a crisp case — hours saved, a risk closed, a number — the account is genuinely healthy regardless of what the health score says. If the answer is hedged, vague, or "well, we've only just started," you have found a renewal risk five months early, with time to build the case with them rather than for them.

9. Is there anything I should have asked and didn't?

The only open-ended question that reliably earns its place. Unlike "any other feedback?", it gives the person a defined job — spot the gap in my understanding — and people are good at that job. Roughly one call in four produces something here that reframes everything above it.

The five that waste the call

"Are you happy with the product?" Yes-biased, unanswerable honestly by a polite person, and even a truthful answer is a sentiment you cannot act on.

"How would you rate onboarding out of ten?" You already have that number from a survey, and it cost you nothing to collect. Spending live conversation time on it is the single clearest sign the call has no design behind it.

"What features would you like to see?" This outsources product design to someone who has been using the tool for four weeks. You will get a list of things adjacent to what they already know. The gap you actually need is behind question seven — what they had to build themselves — not behind a wishlist.

"Would you use it more if we added X?" Hypotheticals about future behaviour are weakly predictive at best, and the customer will say yes because saying yes is free and agreeable.

"Any other feedback?" As a closing question this is a formality and everyone in the call knows it. Question nine does the same work with a real job attached.

What one of these actually sounds like

Real fragment, lightly disguised, from a thirty-day call at a payroll product:

Interviewer: Walk me through the first thing you tried to do.

Customer: I set up the pay run for the fortnightly staff. That went fine.

Interviewer: And then?

Customer: Then I did the monthly ones. That was — yeah, that was fine too, in the end.

Interviewer: In the end?

Customer: Well, we have a few people on both, contractors mostly, and I couldn't work out how to stop them appearing twice. So I just deleted them out of the monthly one each time and added them back after. It's only about ten minutes.

Ten minutes, every month, forever, on a task the customer had already reclassified as normal. They had not raised a ticket. They rated onboarding an eight. On the health dashboard the account was green, because the pay runs were completing.

It came out because someone heard "fine, in the end" and asked two more words. The fix took a developer a day and a half, and it turned up in four more interviews over the following month — at which point it stopped being an anecdote and became the quarter's second-highest-value ticket.

How many of these do you need?

Fewer than you would guess. Abbie Griffin and John Hauser's 1993 Marketing Science paper "The Voice of the Customer" — still the strongest empirical answer to this question — found that twenty to thirty interviews surface ninety to ninety-five per cent of customer needs, and that twenty captured over ninety per cent of what thirty revealed.

So the target is roughly twenty-five conversations per cohort or per segment. At twenty minutes each that is somewhere over eight hours of talking, which is a fortnight of someone's part-time attention, once. You will know you are done when the fourth consecutive call produces nothing you have not already written down. If you are choosing where to start, the accounts that stalled are which customers to call first, and why.

Then do the part that matters

Tag every finding with a cause and an owner: product gap, process gap, expectation gap, or mis-sale. Count them — whether you keep the tally by hand or run these as a study instead of a spreadsheet. Fix the two most frequent. Tell the customers who raised them that you fixed them — this last step costs an email and buys more goodwill than the fix itself.

An interview you do not tabulate is a nice chat. Twenty-five interviews you tabulate is the most reliable onboarding roadmap you will ever have, and it costs less than one conference booth.

---

Where Nosie fits

A script is only worth having if the calls get made. Twenty-five interviews is a fortnight of someone's part-time attention, and that fortnight is the thing that never arrives.

Nosie runs this script for you. It calls the cohort you pick, asks these questions in this order, follows each answer instead of moving to the next item, and returns every response transcribed, tagged by cause and counted — so the tabulation step above is already done by the time the study closes.

Try it on yourself. Your first self-test interview is free, so you can hear the script run before a customer does. If you would rather trigger studies from your own systems, here is how the integration works.