PostHog product dashboard

August 26, 2026 · View on GitHub

How to read Paste It’s anonymous analytics: which journeys matter, which metrics to define, and how to build them in PostHog.

Event catalog and privacy disclosure: analytics.md.
This doc is an ops / product guide — not a list of additional collected fields.

Official builds send events to PostHog Cloud when a project token is injected. There is no checked-in dashboard JSON; create insights in the PostHog UI.

0. Principles

Rule	Why
North star = `clip_staged`	User took something from history back to the system pasteboard. We cannot reliably know “pasted into app X”.
Prefer Unique users / persons	Never treat raw event counts as “users”.
Active ≠ `app_open`	Menu-bar apps stay running; `app_open` only fires on process start (or analytics re-enable). Underestimates DAU.
`app_install` ≠ App Store install	Means first analytics-enabled run on this Mac (UserDefaults). Label charts accordingly.
Global filter	`app_name = Paste It` (or `$app_name`) if the project might mix products.
Default range	Last 30 days; Overview also use Weekly.

Recommended metric definitions (use these names on charts)

Metric	Definition (events)	Notes
DAU (recommended)	Unique persons with `panel_opened` or `clip_staged` that day	Primary health
DAU (legacy / underestimate)	Unique `app_open`	Optional secondary; do not trust alone
WAU	Unique `panel_opened` or `clip_staged`, weekly	—
New enables	Unique `app_install`	Proxy for installs
Activation (7d)	`app_install` → `clip_staged` within 7 days	True “got value”
Panel→Stage conversion	Sessions with stage / all panel closes	Use `panel_closed` + `did_stage`
Empty-open rate	`panel_closed` where `did_stage = false` ÷ all `panel_closed`	Inverse of conversion
Retention	Cohort start `app_install`; return = `panel_opened` or `clip_staged`	Not `app_open`

1. Journeys to watch

Focus on these five paths. Everything else is secondary until these are readable.

Path A — Activation

Question: After first enable, how soon does someone actually reuse history?

app_install
  → onboarding_started (source=first_launch | update | settings)
  → onboarding_step_viewed… / onboarding_completed
  → panel_opened
  → clip_staged   ← activation success

Watch: drop-off by onboarding outcome / last_step; time-to-first-clip_staged.
Do not treat “finished tutorial” as activated.

Path B — Daily reuse (core loop)

Question: Of people who open the timeline, how many stage something?

panel_opened (source, history_count_bucket)
  → [typed search — searches + zero-result flag on session_summary only]
  → clip_staged (trigger, clip_type, tab, age_bucket)
  → panel_closed (did_stage, duration_ms_bucket)
  → session_summary (stages, searches, search_had_zero_results)

North-star session KPI: share of panel sessions with did_stage = true.

Search quality (no new event volume): on session_summary, chart share with search_had_zero_results = true among sessions with searches > 0. Type filters are not tracked.

Path C — Interaction preferences (within stage)

Question: How do people take clips out? Fresh vs old? Which types?

$ \text{clip\_staged} \times \text{trigger} (\text{double\_click}, \text{hotkey\_1\_9}, \text{return}, …) \times \text{clip\_type} \times \text{age\_bucket} \times \text{tab} $

Use for: shortcut teaching, default retention, plain-text defaults.

Path D — Onboarding quality

Question: Where do people bail, and does finishing correlate with later stage?

onboarding_started
  → step views
  → onboarding_completed (outcome, last_step)

Cross-check: completers vs skippers → 7d clip_staged (cohort or funnel).

Path E — Paste Stack

Question: Is Stack discovered? Does Accessibility block Paste Next?

paste_stack_session   ← one event per open→close (quota-friendly)
  direction, collected_count_bucket, paste_next_count / attempts,
  accessibility_trusted_at_open / _at_close, paste_next_without_ax,
  empty_paste_next_count, last_fail_reason?

Charts: Unique users with paste_stack_session; share with paste_next_count > 0; share with accessibility_trusted_at_close = false or paste_next_without_ax > 0.

Path F — Update adoption

Question: Do auto-updates land? Where does the Sparkle funnel stall?

update_interaction
  action: check → found → download → install
  (+ dismiss / fail)
  source: auto | menu | settings

Caveat: download may fire more than once (will/did); filter or split with result using live event samples.

Paths not instrumented yet (do not build charts for these)

Path	Status
Pin / folder / edit / delete	Not wired (defer — more events)
Capture pause / ignored apps / MCP toggle	Not wired

2. Dashboard layout

Create one dashboard:

Name: Paste It — Product

Sections (top → bottom):

Overview (health)
Activation & onboarding
Core loop (panel → stage)
Stage breakdowns
Paste Stack
Updates

Optional later: pin/folder/MCP once those events exist.

3. Insights to configure

3.1 Overview (health)

Chart name	Type	Configuration
DAU (panel)	Trends	Event `panel_opened` → Unique users → Interval Day → Last 30 days
DAU (stage)	Trends	Event `clip_staged` → Unique users → Day
WAU (panel)	Trends	Same as DAU panel, interval Week
New enables	Trends	`app_install` → Unique users → Day. Subtitle: first analytics enable, not App Store install
Retention (install → panel)	Retention	Start: `app_install`. Return: `panel_opened`. (Alt return: `clip_staged`)
Version mix	Trends	`panel_opened` → Unique users → Breakdown `app_version` (or `app_open` if needed for process-start cohort)
DAU (app_open) — optional	Trends	`app_open` Unique users → Day. Label clearly as underestimate

Formula / paired series (empty opens):

Series A: panel_closed where did_stage = true
Series B: panel_closed (all)
Formula A / B → name Panel sessions with stage
Empty-open rate ≈ 1 - A/B, or chart B−A / B separately

3.2 Activation & onboarding

Chart name	Type	Configuration
Install → Stage (7d)	Funnel	Step 1 `app_install` → Step 2 `clip_staged`. Conversion window 7 days. Primary activation KPI
Onboarding steps (first launch)	Funnel	Filter `source = first_launch` on start where applicable. Steps: `onboarding_started` → `onboarding_step_viewed` (`capture` → `paste` → `browse` → `organize` → `stack`) → `onboarding_completed` (`outcome=completed`)
What's New (update)	Funnel	Filter `source = update`. Content version 3 is Stack-only (`stack`). Older packs used the full install sequence.
Onboarding outcomes	Trends	`onboarding_completed` → Breakdown `outcome` (`completed` / `skipped` / `dismissed`)
Bail-out step	Trends	`onboarding_completed` where `outcome` is `skipped` or `dismissed` → Breakdown `last_step`

3.3 Core loop

Chart name	Type	Configuration
Panel opens	Trends	`panel_opened`: Unique users and Total count (two series or two charts)
Open by source	Trends	`panel_opened` → Breakdown `source` (`hotkey` / `status_item` / `menu`)
History size at open	Trends	`panel_opened` → Breakdown `history_count_bucket`
Session duration	Trends	`panel_closed` → Breakdown `duration_ms_bucket`
Panel → Stage	Funnel	`panel_opened` → `clip_staged`. Prefer short window (e.g. 10–30 min). Ideal: same `session_id` if PostHog funnel supports property matching; else person + time window
Sessions with stage	Trends + Formula	See Overview empty-open pair (`did_stage`)
Session intensity	Trends	`session_summary` — average of `stages` / `searches` (as PostHog allows)
Search zero-result rate	Trends + Formula	Among `session_summary` with `searches > 0`, share with `search_had_zero_results = true`

3.4 Stage breakdowns

Chart name	Type	Configuration
Stage by trigger	Trends	`clip_staged` → Breakdown `trigger`
Stage by clip type	Trends	`clip_staged` → Breakdown `clip_type`
Stage by age	Trends	`clip_staged` → Breakdown `age_bucket`
Stage by tab	Trends	`clip_staged` → Breakdown `tab`

3.5 Paste Stack

Chart name	Type	Configuration
Stack sessions	Trends	`paste_stack_session` → Unique users + Total count
Stack collected size	Trends	Breakdown `collected_count_bucket`
Used Paste Next	Trends / Formula	Sessions with `paste_next_count > 0` ÷ all `paste_stack_session`
AX friction	Trends	Filter `accessibility_trusted_at_close = false` or `paste_next_without_ax > 0`
Direction mix	Trends	Breakdown `direction` (`fifo` / `lifo`)

3.6 Updates

Chart name	Type	Configuration
Update funnel	Funnel or Trends	Filter `update_interaction` by `action`: `check` → `found` → `download` → `install`. Validate enum values in Live events first
Update by source	Trends	`update_interaction` → Breakdown `source`
Update failures	Trends	`update_interaction` where `action`/`result` indicates fail (from live samples)
From → to version	Trends	Breakdown `from_version` (and `to_version` when present)

4. Step-by-step: create the dashboard in PostHog

4.0 Confirm data

Left nav → Data / Events (or Activity / Live events).
Confirm recent app_open, panel_opened, clip_staged.
Open one event and verify property names (did_stage, source, trigger, …) match analytics.md.

No events → use a Developer ID–signed build with Secrets/posthog.env injected; Settings → Privacy → analytics on.

4.1 New dashboard

Dashboards → New dashboard
Name: Paste It — Product
Optional: pin to project home

4.2 Minimum viable board (build these first)

Do these six before polishing:

DAU (panel) — Trends / panel_opened / Unique users / Day
New enables — Trends / app_install / Unique users / Day
Install → Stage (7d) — Funnel
Onboarding outcomes — Trends / breakdown outcome
Sessions with stage — panel_closed + did_stage formula (or two series)
Stage by trigger — Trends / breakdown trigger

Then add open-by-source, age/type breakdowns, retention, and update funnel.

4.3 Adding any Trends insight

Dashboard → Add insight → Trends
Pick event → set aggregation to Unique users (unless you intentionally want Total count)
Interval Day/Week → date range Last 30 days
Optional Breakdown by property
Optional Filter (e.g. did_stage = true, is_duplicate_skip = false)
Rename clearly → Save & add to dashboard

4.4 Adding a Funnel

Add insight → Funnel
Add ordered steps (events + property filters per step)
Set conversion window (activation: 7 days; panel→stage: minutes–hours)
Save to dashboard

4.5 Adding Retention

Add insight → Retention
Start event: app_install
Return event: panel_opened (or clip_staged)
Save

5. How to read the board (decision cues)

If you see…	Likely product action
High panel opens, low `did_stage`	Empty-open / discovery problem; check duration buckets and `history_count_bucket = 0`
High `search_had_zero_results` among searchers	Search / empty-state UX
Many `paste_stack_session`, low `paste_next_count`	Stack opened but unused — teach Paste Next / ⌘V
`paste_next_without_ax` or AX false at close	Accessibility onboarding / empty-state CTA
Activation funnel dies before `clip_staged`	Onboarding or first-open UX; check bail `last_step`
Most stages are `age_bucket = <1m`	Users treat it as “recent clipboard”; retention defaults less critical
Many stages are older buckets	History depth / search / pin matter more
`source` dominated by `status_item`, little `hotkey`	Hotkey discoverability
`trigger` mostly `double_click`, little `hotkey_1_9`	Shortcut education
Update `found` high, `install` low	Sparkle / permission / UX stall
DAU(panel) ≫ DAU(app_open)	Expected for menu-bar app — trust panel/stage

6. Privacy reminders for analysts

Never request clipboard text, OCR text, paths, or full search queries in new events.
Search quality: searches + search_had_zero_results on session_summary only — never send the query string. Type filters are not tracked.
Paste Stack emits one paste_stack_session per open→close (not per paste-next).
Clipboard capture is not reported (quota); infer inventory from history_count_bucket / stage types.
Opt-out: Settings → Privacy; client stops sending (optOut).
Uninstall is not observable; churn = silence on panel/stage.

Event list: analytics.md
Client: Sources/PasteIt/Analytics/
Packaging / token inject: mac-packaging.md