Add this key to Google Authenticator, 1Password or any TOTP app, then enter the code it shows.
We will email a single-use code to the recovery address on file for this account.
Choose columns, set widths in pixels, or move them left and right. Saved for your admin account in this browser.
| Actions |
|---|
Totals cover recorded lifetime usage. Calls include metered inbound and outbound calls. Plan value is not an API charge. Country comes from the latest sign-in IP, or the signup IP when no sign-in is recorded. IP Geolocation by DB-IP. Double-click Texts received to open that user’s conversation. Click Inspect or double-click another cell for details.
| User | LLM runs | Tokens | Est. $ | Emails | Texts | Calls (min) | Paid-API fallbacks |
|---|
Activity across all users, newest first.
Checked users are hidden from this event list.
| When (Pacific) | Username | Task | Event | Details |
|---|---|---|---|---|
| Loading events… | ||||
System services, prepaid balances, subscription windows and renewal forecasts.
| Service | Usage coverage | Requests | Tokens | Cost / value | Prepaid balance | Allowance left | Renewal (Pacific) | Forecast | Manage |
|---|
A subscription window belongs to its account and may also be used outside Lisa. Provider windows and monthly plan fees are separate. User-connected plans are counted above; their private credentials are never shown.
Development, lisa.my production and Advisor. Hardware, monthly estimates and usage history.
Estimated model spend, token volume, and usage by model and user.
| Model | Billing | Runs | Input tokens | Output tokens | Est. value | Last used |
|---|---|---|---|---|---|---|
| Loading model usage… | ||||||
| Date | Runs | Muse | Other metered | Subscription value |
|---|---|---|---|---|
| Loading daily usage… | ||||
| User | Runs | Tokens | Muse | Other metered | Subscription value |
|---|---|---|---|---|---|
| Loading user usage… | |||||
Scheduled review of agent token volume, cap hits, and opportunities to reduce repeated work.
Product changes recorded for yesterday and today, in Pacific time.
SMS delivery failures, limit requests and support mail.
Loading the last 7 Pacific days…
Every solver session that ran for a stuck task: what happened, how it was solved, why Lisa could not do it herself, and what was changed in the system as a result.
Growth, task volume, and agent activity over time, in Pacific calendar days.
Unique new users by Pacific calendar day. A user counts once after completing at least one real task; conversation, routing, visitor, merged, and onboarding records are excluded.
Loading registration cohorts…
Unique users with a Facebook ad click recorded at signup, grouped by registration day. The message count includes users who sent at least one message to Lisa.
Loading Facebook cohorts…
Unique new users by Pacific calendar day. A user counts once after completing at least one real task; conversation, routing, visitor, merged, and onboarding records are excluded.
Loading registration cohorts…
All registered users, using recorded lifetime onboarding milestones. The Analytics date range and Overview table search do not change this funnel.
Required counts include users who completed that step and every earlier required step. Lisa nudge and Phone call are optional and counted separately. This shows completion, not the order or timing of users’ actions.
Loading onboarding counts…
Every step from the ad click to a heavy user, and which of them are actually being sent. Match quality is Meta's own score for how well it can tie an event to a person, so a step Meta has never seen shows nothing there.
What users do on the Messages and WhatsApp choice screen (/new/3).
Current website path and earlier onboarding designs, with measured outcomes.
Loading signup flows…
Acquisition funnel by date and audience.
Loading signup interactions…
| Step | Attempts | % visits | % of related step |
|---|
Funnel rows count each action once per attempt. Repeated actions appear separately below; changing the phone or resetting starts a new attempt. Percentages use attempts first tracked in this rolling window, not unique people or Meta impressions. Optional taps and autofocus are separate. Related-step percentages count attempts with both events; a zero denominator is shown as —. Admin-marked tests and demos are excluded from regular traffic. Historical field interactions cannot be recovered.
| Diagnostic | Attempts | % visits |
|---|
Repeated actions count separately; network retries do not. Tap areas are a coarse 3 × 3 viewport grid, not screenshots. Edit means a field was edited, never its contents. Each attempt retains up to 2,000 actions. Counts are best effort when a browser blocks tracking or closes before upload.
| Page | Action | Occurrences | Attempts | % of active page attempts |
|---|
Every percentage is a share of accounts created, not of the row above. Replying is not downstream of the nudge: someone can text Lisa before she ever nudges them, so the reply count is not a conversion of the nudge count.
Tenant browsers have separate local relay ports. This view shows their last recorded upstream routes and paid traffic. Public exit addresses may be shared.
Both dates are included. Pacific calendar days.