left arrow icon

Onboard your first users by hand

Concierge onboarding means setting up the product, or doing the job, by hand for each early user, then automating what repeats. At Superhuman, over 65% of new customers fully moved their email after a human-led session, more than double the self-serve rate, though people who accept a call are not typical. The original concierge test also asked for payment, a stronger signal than goodwill.

Evidence C, limited evidenceStrong when it fits

Stage
Stage 1. No proven demandIs the problem real, and does anyone want this solved?
Category
Demand tests
Time
2 to 4 weeksAwiser's estimate of the active work
Cash
No cash
Live calls
Needs live calls
First result
Within weeks
Risk
LowHours per user, which grow with each one. The risk is mistaking goodwill for demand.
Needs
A few early users who have the problemTime for a session with each, and for follow-upsA real ask: a payment or a firm commitment

When does it fit?

  • You have only a handful of users: Paul Graham says you can sometimes do by hand what you plan to automate later.[1]
  • Getting started means leaving another tool or moving data: Superhuman's onboarding lead names both as situations a human session handles.[4]
  • You sell to businesses and have one engaged early customer: Paul Graham sometimes advises acting as consultants building something for that single user.[1]

When should you skip it?

  • You want the product judged on its own: Gaurav Vohra notes that customers willing to take an onboarding call are a subset of those who self-serve.[4]

How do you run it?

  1. Pick a few early users

    People who have the problem and agreed to try, not friends doing you a favor.

  2. Do the job with them, by hand

    Set up their workflow live, move their data, and do by hand what the product cannot do yet.

  3. Ask for something real

    A payment or a firm commitment, as the original concierge test did with a weekly check.

  4. Log every friction

    Write down each question, bug and workaround. It is your backlog.

  5. Watch for unprompted returns

    Count who comes back, or uses it, without being asked.

  6. Automate what repeats

    When you can predict what users will say and repetitive tasks slow your learning, automate that step.

What usually happens?

  • Many will say no: in the original concierge test, retold by Coleman McCormick, every interview ended with a sale attempt and most were rejected.[3]
  • You will know when to move on: Food on the Table's founder took the next step once the team could anticipate what early adopters would say.[2]

Who has done it, and what happened?

Each case is one team's own report: what happened to them, not a typical result.

  • Superhuman, reported by Gaurav Vohra, who designed its onboarding2025, Positive result

    Required a human onboarding session for new customers, set up each workflow live, and logged every feature request, bug and friction point.

    Gaurav Vohra reports that over 65% of new customers fully moved their email to Superhuman after the session, more than double the rate without any onboarding.[4]

    The author states the selection bias: customers who take a call differ from those who self-serve, and the comparison group had no in-app onboarding at all.

What does the evidence say against it?

  • Engagement alone is a weak signal: in the original concierge test, retold by Coleman McCormick, the team collected a check for $9.95 each week.[3]
  • No recent founder report with numbers was found for a concierge test leading to paid conversion.

How do you read the result?

What to count
Users who came back or used it without being asked, and users who paid or committed
When to read it
Two to four weeks after the first sessions
How to decide
Automate and widen only when users come back on their own and accept a real ask. Food on the Table's founder put it bluntly: if people do not adopt with high-touch service, a web page will not do better.[2]

What goes wrong?

  • Being paid for attention rather than the product: Paul Graham warns that users who pay for attentiveness expect you to do everything.[1]
  • Never automating: Food on the Table's founder automated once repetitive tasks slowed learning.[2]
  • Skipping the ask: in the original concierge test, every interview ended with an attempt to make a sale.[3]

How was it graded?

Evidence
C. Limited evidenceThe method is widely taught and one recent case reports numbers, Superhuman roughly doubling full adoption with human onboarding, but its author states a selection bias, and the older concierge cases predate the five-year window.
Verdict
Strong when it fitsWorth it while you have a handful of users and getting started is hard for them. Ask for payment or a commitment, or you are measuring goodwill.

How the grades and verdicts are decided

Sources

  1. 1
    Do Things that Don't Scale (opens in a new tab)paulgraham.com, Paul Graham. July 2013. Method.
    When you only have a small number of users, you can sometimes get away with doing by hand things that you plan to automate later.
    Sometimes we advise founders of B2B startups to take over-engagement to an extreme, and to pick a single user and act as if they were consultants building something just for that one user.
    But when they start paying you specifically for that attentiveness ... they expect you to do everything.
    Read on October 20, 2026.
  2. 2
    Concierge MVP: Learning from early adopters at Food on the Table (opens in a new tab)SlideShare, Manuel Rosso. March 12, 2011. Founder report.
    Move on to the next step when you can anticipate what your early adopters are going to say before they say it.
    If you can't get them to adopt your idea with high-touch, face-to-face service, they sure as hell are not going to buy into your cold web page.
    Automate when you are spending most of your time doing repetitive tasks that slow down your learning.
    Read on October 20, 2026.
  3. 3
    Learning from the Concierge MVP (opens in a new tab)Res Extensa, Coleman McCormick. March 22, 2023. Analysis.
    As they met potential customers in those settings, they would interview them the way any good market researcher would, but at the end of each interview they would attempt to make a sale.
    Most times they were rejected.
    Most important, each week they would collect a check for $9.95.
    Read on October 20, 2026.
  4. 4
    Obsessing Over Onboarding for 10+ Years - The Architect of Superhuman's Onboarding Shares His Playbook (opens in a new tab)First Round Review, Gaurav Vohra. April 23, 2025. Founder report.
    Maybe the customer gets the job done today with different tools. Maybe they have existing data that needs migrating.
    Due to selection bias, the segment of customers willing to show up to a 30-minute call was always a subset of the segment willing to self-serve into a product.
    Over 65% of new customers fully transitioned their email to Superhuman after their human-led onboarding. This was more than double what we observed by simply dropping users into the product, absent any in-app onboarding, and letting them figure it out themselves.
    Read on October 20, 2026.

Sources last checked on October 20, 2026. First published on October 20, 2026.

Awiser is the builder network where founders show their skills, find collaborators, and grow their projects with an AI copilot.