Onboarding tool or redesign: what does your B2B SaaS onboarding actually need?
How to tell whether a B2B SaaS onboarding problem needs an in-app tool, a redesign or a product change, with a symptom-to-fix table and a checklist for briefing an agency.
Activation is flat. One vendor offers a tour-and-checklist tool for a monthly fee. An agency offers a redesign for a project fee. A third option sits in the backlog, usually unnamed: a change to the product itself. Any of the three can be the right answer, and each wastes money when it is aimed at the wrong problem. A tour laid over a broken import step points politely at the failure. A redesign of screens whose real problem is a missing integration produces better-looking screens that fail the same way.
This guide gives you a triage table you can fill in from your own data in about a week, then a short list of what to send an agency if the answer turns out to be a redesign. Humbleteam, a design agency, publishes it, so read its lean toward design work with that in mind. The table is an editorial framework, not a research finding.
Three kinds of fix, and what each can change
An in-app guidance layer. Appcues, Userpilot, Pendo and Chameleon sell tools in this category. Their own pages describe flows or tours, checklists, tooltips and resource centers that a product team builds without code, shows to chosen segments, and measures. Userpilot lists flows, checklists, tooltips, modals, spotlights, banners and a resource center. Chameleon lists tours, tooltips, checklists, launchers, banners and microsurveys. Pendo describes walkthroughs and tooltips targeted at the user or account level, and Appcues documentation names flows, pins, checklists and launchpads. These are guidance and measurement features. They do not add an import, an integration or a permission model. That stays product work.
A redesign. This changes what the screens ask, in what order, and what a new user sees first: step order, wording, empty states, role-specific paths and the handoff between people. It needs a designer, and engineers to build the result.
A product or data change. This covers the activation event itself, a missing import or integration, sample data, and permissions for the second user. Product and engineering own it. A designer can help specify it, but the work is not design work.
What research says about tours
Page Laubheimer’s 2023 Nielsen Norman Group article on onboarding tutorials argues that tutorials shown at launch interrupt people who want to start, are often skipped, and in the author’s studies frequently did not improve task performance. Help that appears when a person is likely to need it avoids most of those problems. Walkthroughs can still earn their place for genuinely new interactions.
Two consequences follow. A tool is at its best delivering short, contextual help, not a long welcome tour. And no tool compensates for a screen that is confusing in itself. The tool vendors agree on the first point in their own guidance: Appcues documentation advises keeping flows to about 3 to 5 steps and warns that modals can feel like an obstacle. The research is general, not specific to B2B, so check it against your own users.
The triage table
Start from the symptom, not from the solution. Each row names the evidence to pull, the likely fix type and who should do it.
| Symptom | Evidence to check | Likely fix | Who should do it |
|---|---|---|---|
| Large drop at one setup step | Funnel by step, split by role, plan and signup source; then recordings of that step | If people hesitate over wording or order: redesign the step. If they hit errors: fix the errors | Design agency or in-house designer for hesitation; engineering for errors |
| Setup finishes, but the key action never happens and users look lost on the first screen | First-session recordings; support questions that start with “where do I” | Guidance on the existing screen: a checklist or tooltip pointing to the key action | Product team with a tool |
| Users arrive at an empty product and leave | What the account contains at first login; recordings of the first two minutes | Redesign the empty state with a learning cue and a direct path to the first task; add sample data if needed | Design agency or in-house designer; engineering if sample data must be built |
| Setup stalls when connecting data or importing a file | Import failure and retry counts; support tickets on this step; whether people return after a failure | Product change to the import or integration first, then design of the recovery states | Engineering first, designer second |
| The admin finishes, but invited colleagues do nothing | Invitations sent versus accepted versus first useful action; what an invitee sees and which role they get | Change the default permissions, then redesign the invitation and the invitee’s first screen | Product and engineering, then a designer |
| Activation looks healthy, but accounts still churn | Retention of accounts that hit the activation event versus those that did not | Redefine the activation event before touching any flow | Product and data, no agency yet |
| Tour completion is high, but the key action rate is not | Tour completers versus skippers, same cohort, on the key action | Cut the tour, move help next to the step, or fix the screen the tour explains | Product team with a tool; designer if the screen is the cause |
| One segment activates and another does not | The same funnel split by role, company size and plan | Segment-specific messages if the paths are right; a separate path if they are not | Product team with a tool for messages; design agency for separate paths |
Two rows have their own guides on this site: recovering from a failed import and what the second person sees when they join.
The empty-state row rests on Kate Kaplan’s Nielsen Norman Group work on empty states. She notes that empty states are especially common in complex applications the user has not yet configured, and recommends communicating system status, giving learning cues, and offering a direct pathway to the task that fills the space.
The churn row is easy to overlook. Elena Verna argues in her guide to measuring activation in B2B that logins and views are not activation, that setup steps should not count as delivered value, and that multi-user products should measure activation at the team level. Lenny Rachitsky’s post on choosing an activation metric (partly paywalled) walks through testing candidate milestones against retention. If your “activated” accounts do not retain better than the rest, no onboarding fix is safe to buy yet.
A worked example (illustrative)
A reporting product sees its biggest drop at the “Connect your data source” step. Five recordings of new accounts show the same thing: four users open the documentation in another tab and leave when their warehouse is not on the list.
That is not a tour problem, because an arrow pointing at a connector that does not exist helps nobody. It is not a screen problem either. It is a missing integration. The order of work: engineering decides whether to build the connector or offer a file import, then a designer shapes the step, and a guidance layer comes last if people still miss the new option. Had the connector existed and the recordings shown people scrolling past the button, the first move would have been a tooltip and the tool would have been the cheap, correct answer.
Run the check in one week
- Pull the funnel. Name every step from signup to the first useful action and split it by role, plan and source. Any product analytics tool with funnels and breakdowns works; Mixpanel’s documentation shows the idea. Keep the counts, not just percentages.
- Watch five first sessions. Recordings explain the “why” behind a drop-off that a report only locates; Mixpanel Session Replay links recordings to events. Jakob Nielsen’s advice on test size is that five people per round of qualitative testing surface most problems, and more rounds beat bigger rounds. Recordings of real sessions are not a usability test, but the logic carries over: you are looking for repeated confusion, not a rate. Check consent and privacy rules before recording.
- Read the tickets. Tag recent support questions by onboarding step. A cluster of “where is” questions points at guidance. A cluster of “it failed” questions points at engineering.
- Walk the second user’s path. Create two test accounts with different roles, invite one from the other, and note what the invitee can and cannot see.
- Test the activation event. Compare retention for accounts that reached it with accounts that did not.
With only a few hundred signups, the funnel will be noisy. Lean on steps 2 to 4 in that case.
When a tool is enough and when it hides the problem
A guidance layer is the right first move when the product works, people reach the right screen, and the key action is simply hard to find. It lets the team change copy and targeting weekly without a release. It also needs an owner, because flows break when the interface underneath them changes. That maintenance point is an editorial judgment, not a vendor claim.
A tool hides the problem when it becomes a stack of tooltips on a screen nobody understands, or a checklist whose first item fails for a technical reason. If the same step keeps needing another overlay, the step needs redesigning.
A redesign is the right move when recordings show the same hesitation at the same screen, when one path is forced on very different roles, or when the order of steps works against the user. The three often come in sequence: fix the product gap, redesign the flow, then use a tool to keep the guidance current.
What to send an agency if the answer is redesign
If the table points to a redesign, send the diagnosis along with the request. It can shorten discovery and keeps the agency from designing around a problem that is not design.
- The funnel with step names, date range and counts, and the step you believe is failing.
- Clips of the recordings behind that belief, with timestamps and consent confirmed.
- A screenshot of every tour, checklist or tooltip currently live. An overlay is part of the experience the agency will redesign.
- A list of problems you have classified as product work, with an owner and a date, so nobody asks the agency to solve them with a prettier screen.
- Your activation event, how you chose it, and whether you have tested it against retention.
- A sample of tagged support tickets from the failing step.
- Whether you plan to keep your guidance tool, and which parts of the design must be buildable inside it.
For help with the selection itself, Humbleteam’s comparison of agencies for B2B SaaS onboarding covers shortlisting and what to agree before work starts.
Limits of this method
The table is a triage tool. It narrows the choice; it does not prove a cause, and several rows can be true at once. The vendor descriptions above come from their own pages and state capabilities, not results, and they were not tested for this guide. Small samples, a recent product release or a change in your signup sources can all move a funnel without any onboarding change.
Before you sign anything, fill in the table for your own product, one row per symptom you can find evidence for. The “who should do it” column will tell you which budget to open first, and sometimes it will say none of them yet.
