Tradly MemoryCompare · Web analytics
Umami is open source and private. It's also just page views.
Umami gives you control and simplicity. Tradly Memory gives you the same lightweight feel plus per-visitor intent and memory your agents can use.
The gap
Self-hosting solves control, not insight
Umami is popular for good reasons — open source, self-hostable, cookieless, fast. But a private dashboard of aggregates is still a dashboard of aggregates.
Umami earned its following as the lightweight, open-source alternative: you host it, you own it, no cookies, tiny script, clean dashboard. For teams that want analytics they control, it's a reasonable choice.
What you control is a counting system. Umami shows you page views, referrers, and visitors. It won't tell you which specific visitor is close to buying, why they stalled, or what to do for them — and it has no path to hand that to an AI agent. The insight you can act on is precisely what a self-hosted counter doesn't produce.
Head to head
Umami vs. Tradly Memory
| What you get | Umami | Tradly Memory |
|---|---|---|
| Open source / self-hostable | Yes | Managed, agent-ready |
| Cookieless, privacy-safe | Yes | Yes |
| Per-visitor profiles | No | Yes |
| Intent signals (search, cart, CTA) | No | Built in |
| Cross-channel recommendations | No | Yes |
| Feeds AI agents / MCP | No | Yes |
Where Umami falls short
Three things a self-hosted counter can't do
1. Remember a visitor across sessions
Umami measures visits, not journeys. The returning shopper who researched a product yesterday and returns today is invisible as a person.
2. Capture intent signals
Searches, CTA clicks, cart adds, and comparison behavior are the signals that predict buying. Umami's data model has no place for them.
3. Connect to agents and actions
Recovery emails, ad audiences, and AI agents need scoped visitor context. Umami serves aggregate counts — and counts don't drive a decision.
The difference
A counter vs. a memory
Counter (Umami)
You own the server and the page-view counts. The visitor behind the count stays invisible.
Memory (Tradly Memory)
maya-84f2: returned twice, compared plans, stalled at checkout — intent high, action recommended.
The alternative
Own the insight, not the server
Skip the server. Keep the privacy.
Install one script and get per-visitor intent and agent-ready memory — without running your own counter.
Install the pixel