Activation & onboarding review

MEMO

To: Onboarding squad Re: Where we lose people, and what to fix first

Where we lose people

• Signup to connect (100% to 41%): The largest loss. Most of our users are PMs and marketers who don't have database credentials, and the flow gives them no way forward without them.
• Connect to first chart (41% to 33%): About a fifth of people who connect never build a chart. Sync and the empty canvas are the likely causes.
• First chart to invite (33% to 9%): Expected, since invites come after value. Lower priority for now.

Retention is the reason to care about the first chart: 62% of users who reach one are active in week 2, versus 11% of those who don't. This is correlational, since engaged users may be the ones who build charts, but it is the strongest signal we have.

What to fix, in order

1. Add sample data and a skip at step 5. Users without credentials currently hit a dead end. Sample data gets them to a chart in minutes, which is the milestone that predicts retention. This also targets the top support ticket (41%). Reason: highest expected impact on the number that matters.

1. Let people in before email verification, and drop the role and team screens from the critical path. Step 2 blocks all progress, and steps 3–4 ask questions before the user has seen any value. Move role and team to after the first chart, where they can personalize templates. Reason: removes the earliest friction, when drop-off is highest.

1. Rebuild the credentials path. Replace the 14-page guide with short, per-warehouse steps for Postgres, BigQuery and Snowflake, and add a "send these instructions to your data engineer" link that delivers a connect-only invite. The person who has the credentials is often not the person signing up. Reason: addresses the remaining credential tickets and the step 6 drop-off.

1. Fix sync. Let leaving the page not cancel the sync. Show partial results as tables finish. Pre-select likely tables and add search to the step 7 list. Make CSV upload the fastest route to a first chart. Reason: sync failures are the second-largest ticket category (18%), and cancellation on exit loses users mid-setup.

1. Make the first chart appear without effort. Templates (step 9) should render a live chart immediately. Replace the empty canvas at step 10 with a suggested chart the user can edit. Reason: a blank canvas is the last hurdle before the retention milestone.

1. Move team invites after the first chart, and prompt for them there. Keep the skip link. Reason: invites are useful but come after value; test this once the upstream fixes land.

Measurement

Track signup-to-first-chart as the primary metric, with week-2 retention as the outcome. Run fixes 1 and 2 as separate experiments so we can see which one moves the funnel.

Onboarding: where we lose people and what to fix first

Sep 25, 2026 · @Martin Slaney

To: Onboarding squad

Bottom line

Onboarding works for people who reach a first chart: 62% of them are still here in week 2, against 11% of everyone else. Two-thirds of signups never get there, and most of those stop at "Connect your data", where we ask product managers and marketers for database credentials they don't have.

Where we lose people (last 90 days, 12,400 signups)

• Signup → connect data: 59% lost (\~7,300 people). The biggest leak by far. The top support ticket (41%) is "I don't have the database credentials."
• Connect → first chart: \~990 more lost, a fifth of those who connected. Syncs take up to 20 minutes, die if you leave the page, and "my sync failed" is the second-biggest ticket (18%).
• First chart → invite: 33% → 9%. Real, but downstream of the first two.

Caveat: we have no step-level data for screens 2–8, so we can't split the 59% between verification, the two questions and the connection form. The ticket data points hard at the form.

What to fix, in order

1. Let people reach a first chart without their own data. Add a sample dataset and a skip on screen 5, and land them on a ready-made chart. Why: it goes around the step that loses most people, and the first chart is the moment that predicts retention.
2. Let them hand off the credentials step. "Ask a teammate to connect" sends an engineer a secure link to fill in the connection form. Why: it answers the 41% ticket directly and turns the blocker into an invite.
3. Make syncing forgiving. Run it in the background, keep it going when they leave, email them when it's ready, and load a small sample of rows first so a chart appears in minutes. Show plain-language errors with a retry. Why: a 20-minute wait that punishes leaving the page is where connected users die, and it's the 18% ticket.
4. Don't start from an empty canvas. Use the role answer from screen 3 to suggest two or three templates instead of twelve, and draw the first chart automatically. Why: we already ask the role and do nothing with it; a blank builder is hard for non-analysts.
5. Trim the steps before connecting. Let people in before they verify their email (verify within a day), and drop or defer the team-size question. Why: cheap to do, but unmeasured, so instrument first rather than assume it's a big win.
6. Move the invite prompt to after value is shown, such as when someone saves or shares a chart. Why: people invite colleagues to something useful, not to a setup screen.

How we'll know it worked

• Instrument every screen, 1 to 11, before shipping anything.¹
• Judge fix 1 by how many sample-data users go on to connect real data and are retained in week 2, not by first-chart rate. A chart of sample data can lift that metric without anyone sticking around, so don't assume the 62% carries over.

To: Onboarding Squad Subject: Fixing Our Activation Funnel: Priorities & Action Plan

Where We Are Losing People

Our biggest drop-off occurs at Step 5–8 (Connect Data): 59% of signups leave without connecting a data source.

We are asking PMs and marketers for database ports, SSL modes, and passwords they do not possess. 41% of our support tickets are “I don't have credentials,” followed by 18% for failed syncs. Furthermore, requiring a 20-minute, open-tab sync kills momentum.

Because Week-2 retention jumps from 11% to 62% once a user builds their first chart, our sole objective is getting non-technical users to that "aha!" moment in under two minutes.

---

What to Fix (In Priority Order)

#### 1. Add “Explore with Sample Data” & a Clear “Skip” Button (Step 5) * What: Let users bypass database connection entirely using a fully populated e-commerce or SaaS demo dataset. * Why: This is the highest-leverage fix. It immediately unblocks the 59% who drop off at Step 5 and the 41% who lack DB credentials, giving them an instant path to create a first chart and experience the product’s value on Day 1.

#### 2. Introduce “Invite a Data Teammate” at Connection (Step 6) * What: Add a 1-click option: “Don't have credentials? Send an invite link to your data lead to connect your warehouse.” * Why: Replaces a dead end with collaboration. The PM stays in the app using sample data while the engineer handles credentials asynchronously.

#### 3. Make Syncing Asynchronous (Step 8) * What: Run syncing in the background. Notify users via email/in-app alert when ready instead of forcing them to keep a 20-minute tab open. * Why: Directly addresses our second-biggest support issue (18% sync failures/cancellations) and eliminates the drop-off caused by tab abandonment.

#### 4. Pre-Populate the Dashboard; Kill the Blank Canvas (Steps 9 & 10) * What: When a user selects a template, generate it automatically with 3–4 standard charts instead of dropping them onto an empty canvas. Prompt them to tweak one existing chart. * Why: Bridges the 8% drop-off between connecting data (41%) and creating a first chart (33%). Editing an existing chart is far lower friction than building from scratch.

#### 5. Defer Email Verification and Profiling Questions (Steps 2, 3, & 4) * What: Move verification to a non-blocking banner at the top of the app, and cut or defer the role/team-size questions until after the user creates their first chart. * Why: Every barrier before core value reduces trial-to-paid conversion. Let users reach the product before demanding administrative tasks.

---

Next Step

Let’s implement Priority 1 (Sample Data) this sprint. It requires zero infrastructure changes to our warehouse sync logic and directly targets our biggest activation leak.