Redesigning Harvest's user onboarding
Redesigning Harvest's user onboarding
Redesigning Harvest's user onboarding
From welcome flow to in-product getting started. An end-to-end redesign that lifted trial-to-paid by 8%.
From welcome flow to in-product getting started. An end-to-end redesign that lifted trial-to-paid by 8%.
Role
Design strategy, UX, UI, research, plus hands-on frontend where it moved the work forward.
Design strategy, UX, UI, research, plus hands-on frontend where it moved the work forward.
Team
Me
PM
EM
FE
BE
What this case study covers
Harvest's onboarding had two problem areas: a welcome flow losing 30% of signups before they reached the product, and an in-product getting-started experience only 1.5% of trialists ever found. We rebuilt both.
Harvest's onboarding had two problem areas: a welcome flow losing 30% of signups before they reached the product, and an in-product getting-started experience only 1.5% of trialists ever found. We rebuilt both.
Impact
Welcome flow completion
Welcome flow completion
70% → 90%
70% → 90%
Projects created during welcome flow
Projects created during welcome flow
60% → 85%
60% → 85%
Activation rate
Activation rate
6.6% → 25.2%
6.6% → 25.2%
Trial-to-paid conversion
Trial-to-paid conversion
+8%
+8%
One guardrail metric (repeat sessions) moved against us. Covered later in the case study.
The problem
Losing half our signups after their first session.
Losing half our signups after their first session.
Losing half our signups after their first session.
What the funnel looked like:
What the funnel looked like:
30% of signups dropped out of the welcome flow itself, before even reaching the product
30% of signups dropped out of the welcome flow itself, before even reaching the product
5,500 of 10,000 monthly signups gone after session one
5,500 of 10,000 monthly signups gone after session one
75% of team trials never invited a teammate
75% of team trials never invited a teammate
57% never tracked time
57% never tracked time
93% never created an invoice
93% never created an invoice
<1.5% ever found the existing getting-started checklist
<1.5% ever found the existing getting-started checklist
Research
How we diagnosed the problem.
How we diagnosed the problem.
How we diagnosed the problem.
Methods:
Methods:
Journey-mapping every phase: sign up → explore → use → decision
Journey-mapping every phase: sign up → explore → use → decision
Quant analysis via Pendo, Heap, and Mode: conversion funnel, feature adoption, repeat sessions
Quant analysis via Pendo, Heap, and Mode: conversion funnel, feature adoption, repeat sessions
User interviews with churned prospects and successfully converted trialists
User interviews with churned prospects and successfully converted trialists
Usability testing on the welcome flow to diagnose the 30% drop-off
Usability testing on the welcome flow to diagnose the 30% drop-off
UserVoice and NPS analysis: voice-of-customer patterns
UserVoice and NPS analysis: voice-of-customer patterns
I led design research alongside the growth team's PM, data partner, and CSM.
What we heard
"Please, just tell me where to look and what to do."
"Please, just tell me where to look and what to do."
"Please, just tell me where to look and what to do."
— Camille, consultant
— Camille, consultant
"Maybe I'm not quite ready."
"Maybe I'm not quite ready."
"Maybe I'm not quite ready."
— Daria, architect
— Daria, architect
We had interview sessions with more than 20 prospects and customers.
The insight
Not all actions are equal - some predict conversion.
Not all actions are equal - some predict conversion.
Not all actions are equal - some predict conversion.
Clients, tasks, and projects were 2–3× more predictive of conversion than time tracking.
Clients, tasks, and projects were 2–3× more predictive of conversion than time tracking.
This reshaped the brief: onboarding shouldn't point users at the feature Harvest is named after. It should point them at the actions that turn trials into paying teams.
This reshaped the brief: onboarding shouldn't point users at the feature Harvest is named after. It should point them at the actions that turn trials into paying teams.
Feature completed
Feature completed
Conversion rate
Conversion rate
Clients
Clients
15.5%
15.5%
Tasks
Tasks
12.9%
12.9%
Projects
Projects
11.9%
11.9%
Integrations
Integrations
10.6%
10.6%
Reports
Reports
8.1%
8.1%
Invoices
Invoices
5.7%
5.7%
Time tracking
Time tracking
5.4%
5.4%
Task creation became our primary activation metric: it predicts conversion, and it's a leading indicator we could measurably move.
Solutions
Streamline the welcome flow
If we make the initial quiz more relevant, users are more likely to return for a second session, because people tend to be consistent with their previous actions.
If we make the initial quiz more relevant, users are more likely to return for a second session, because people tend to be consistent with their previous actions.
Activate via an in-product checklist
If we frame onboarding items as a grouped set and expose how-to references, users are more likely to complete them, because tasks that are part of a group are more tempting to complete (pseudo-set framing).
If we frame onboarding items as a grouped set and expose how-to references, users are more likely to complete them, because tasks that are part of a group are more tempting to complete (pseudo-set framing).
First bet
The re-designed welcome flow
The re-designed welcome flow
The re-designed welcome flow
We kept the multi-step structure but rewrote every question and gave the flow its own identity, so users stopped mistaking the project-creation step for the actual product.
We kept the multi-step structure but rewrote every question and gave the flow its own identity, so users stopped mistaking the project-creation step for the actual product.
Welcome flow completion
Welcome flow completion
70% → 90%
70% → 90%
Projects created during welcome flow
Projects created during welcome flow
60% → 85%
60% → 85%
Second bet
In-product starting point
In-product starting point
In-product starting point
Harvest used Pendo to host feature tours and a getting-started checklist. Fewer than 1.5% of trialists ever found it.
Harvest used Pendo to host feature tours and a getting-started checklist. Fewer than 1.5% of trialists ever found it.
v1 — Leverage what we had
Two fast iterations, shipped solo, no engineering dependency.
Two fast iterations, shipped solo, no engineering dependency.
Resource Center interactions
Resource Center interactions
+400%
+400%
Feature interaction at account level
Feature interaction at account level
+25%
+25%
Resource Center visits
Resource Center visits
+38%
+38%
Checklist engagement
Checklist engagement
+30%
+30%
v2 — Built into the product
A dedicated first-run page replacing the popup.
A dedicated first-run page replacing the popup.
Results
Core feature engagement significantly up
Core feature engagement significantly up
Core feature engagement significantly up
Action
Action
Result
Result
Task created
Task created
+282% rel. ✅
+282% rel. ✅
Client created
Client created
+48% rel. ✅
+48% rel. ✅
Project created
Project created
+26% rel. ✅
+26% rel. ✅
Time entry created (guardrail)
Time entry created (guardrail)
Held steady ✅
Held steady ✅
Reflection
Activation isn't a single feature. It's a system.
Activation isn't a single feature. It's a system.
The best next step isn't always iterating on what you built. Sometimes it's looking upstream. The welcome-flow redesign moved more numbers than any in-product change did.
The best next step isn't always iterating on what you built. Sometimes it's looking upstream. The welcome-flow redesign moved more numbers than any in-product change did.
Top-of-funnel decisions shape mid-funnel outcomes. You can't optimize a checklist if the welcome experience is already losing 30% of users before they see it.
Top-of-funnel decisions shape mid-funnel outcomes. You can't optimize a checklist if the welcome experience is already losing 30% of users before they see it.
Appendix
The tradeoff
A guardrail we knew we were risking, and decided to ship anyway.
A guardrail we knew we were risking, and decided to ship anyway.
Repeat sessions: 🔻 -10.55% (53% → 47%)
Repeat sessions: 🔻 -10.55% (53% → 47%)
Why:
Users landed on a checklist instead of going straight to the app
Additional friction before they could explore freely
The checklist didn't create enough value for users who weren't interested in Harvest to begin with
Why:
Users landed on a checklist instead of going straight to the app
Additional friction before they could explore freely
The checklist didn't create enough value for users who weren't interested in Harvest to begin with
The squad and leadership discussed the likely tradeoff before the test. Core engagement lifts and trial-to-paid were the primary success metrics; repeat sessions was a guardrail we were willing to risk if conversion moved enough. It did. Trial-to-paid +8%, core actions up decisively. So we shipped to 100% of account owners with a clear hypothesis for the next iteration.
The squad and leadership discussed the likely tradeoff before the test. Core engagement lifts and trial-to-paid were the primary success metrics; repeat sessions was a guardrail we were willing to risk if conversion moved enough. It did. Trial-to-paid +8%, core actions up decisively. So we shipped to 100% of account owners with a clear hypothesis for the next iteration.
Where I'd take this next
Personalizing the hub based on user intent. Richer setup paths for users who want guidance; lighter-touch exploration for users who don't, recovering the repeat-session cost from v2.
Personalizing the hub based on user intent. Richer setup paths for users who want guidance; lighter-touch exploration for users who don't, recovering the repeat-session cost from v2.
A natural place to apply intelligent defaults: use signal from the welcome flow answer to tailor the getting-started experience without another question.
A natural place to apply intelligent defaults: use signal from the welcome flow answer to tailor the getting-started experience without another question.
The welcome flow we didn’t ship
Before landing on the redesigned flow, we explored a more ambitious concept: an interactive welcome experience where the UI behind the questions would respond to the user's answers in real time. As you selected what you wanted to accomplish with Harvest, the app preview alongside would populate and change. Onboarding as education, built into the act of answering.
Before landing on the redesigned flow, we explored a more ambitious concept: an interactive welcome experience where the UI behind the questions would respond to the user's answers in real time. As you selected what you wanted to accomplish with Harvest, the app preview alongside would populate and change. Onboarding as education, built into the act of answering.
Why we rejected it: Too resource-heavy for the bet. We wanted to move fast on fixing the drop-off, and this concept would have needed significant engineering investment before we knew whether a lighter redesign could close the gap. It did (70% → 90%), which validated the call.
Why we rejected it: Too resource-heavy for the bet. We wanted to move fast on fixing the drop-off, and this concept would have needed significant engineering investment before we knew whether a lighter redesign could close the gap. It did (70% → 90%), which validated the call.
