Azure App Insights → Marketing & Technical dashboards in 2 minutes.
AI-mapped schema. 22 KQL templates. Nothing raw ever leaves your tenant. MIT.
Marketing Analytics
Visitors, sessions, top pages, geo distribution, navigation paths, browser and device breakdown — with period-over-period delta chips.
Technical Monitoring
Response time percentiles, error rates, slow endpoints, frontend perf — all from existing telemetry, no agent to deploy.
Smart Readiness
0–100 readiness score, signal breakdown, and AI-ready prompts you can paste into Cursor or Copilot to fill the gaps.
How it compares
| Keren Analytics | Azure Portal | Datadog | Power BI | |
|---|---|---|---|---|
| Time to first dashboard | ~2 min (Docker) | 30+ min (write KQL) | 1–2h (agent + setup) | Hours (data prep) |
| Marketing vs Technical separation | Built-in | Manual workbook | Add-on | Manual report |
| Readiness scoring + AI prompts | 0–100, LLM-ready | — | Limited | — |
| Custom-dimension auto-mapping | Alias + regex (LLM optional) | Manual | Manual | Manual |
| Data residency | No raw data leaves your tenant | Native | New endpoint | New endpoint |
| License / cost | MIT, free | Included w/ Azure | Per-host $$$ | Per-user $$ |
Launch-time positions; the gaps narrow as each tool evolves. The columns we're least kind to (Datadog, Power BI) are also the most mature and have features we don't.
Privacy & security
User telemetry rows never leave your Azure tenant via this service. Only aggregated metrics (counts, percentiles, geo / browser distributions, top-N pages) and setup metadata (mapping, schema profile, dashboard payload) ever cross the wire.
That promise is encoded as automated checks in
scripts/security-audit.mjs.
Seven controls run on every push & PR plus a Monday cron —
sensitive-data logging, session-cookie hardening, CSP purity,
no-raw-telemetry-persistence, committed env-files placeholder,
npm audit high+. Production refuses to boot without a
real SESSION_SECRET; per-IP rate limiting protects the
public demo. See the
security policy
for the auditable controls and the private reporting path.
FAQ
Does it store my telemetry data?
No raw rows. Only aggregated metrics and setup metadata
(mapping, dashboard payload). Two filesystem sinks exist on the
server, both documented in SECURITY.md; any other
fs.write in src/ fails the security
audit.
How does it connect to Azure?
OAuth (PKCE) against Entra ID. You sign in once, the service exchanges the code for a delegated token, and queries run with your permissions — Reader on the subscription, Log Analytics Reader on the workspace. No client secret to manage on your side. Full setup in docs/setup-entra-id.md.
Do I need to register my own Azure app?
No. The host of the Keren Analytics instance you're using (this demo, or whoever deployed it inside your company) did that once. You just click Connect your Azure and sign in with your normal Microsoft account — same flow as Slack, Loom, Notion. The first sign-in from your tenant may show a one-time consent screen ("This app wants to read your Azure resources on your behalf"); a single click and you're in. If your tenant restricts user-consent, an admin needs to consent once for the whole org.
Is the AI part required?
No. The launch ships with deterministic schema mapping (alias table + regex patterns covering ~80% of real-world naming) and a deterministic "Environment analysis" panel that reads from the dashboard payload. Real Azure OpenAI integration is an opt-in post-launch layer — the dashboard works fully without it.
Traffic Trend
Top Pages
| Path | Views | Share |
|---|
Geo Distribution
Peak Hours
Content Performance
conversion contributionCampaigns & URL Parameters
| Source | Medium | Campaign | Visitors | Sessions | Conv. | Conv% |
|---|
Auto-detected URL parameters from your telemetry. Pin any parameter to add it as a dashboard filter.
Conversion Funnel
Traffic Sources
User Flow
| From | To | Transitions |
|---|
Browser
OS
Device
Frontend Performance
Slow Endpoints
| Path | P50 | P95 | P99 | Calls | Err% |
|---|
Session Timelines
recent sessionsReconstructed user journeys from telemetry events. No screen recording needed.
Signal Breakdown
Click How to fix on any missing signal to get a ready-to-use prompt for your AI code assistant.