I’d contend having user behaviour logged directly into your monolithic app db using the ahoy gem instead of some external grafana/etc system is much more powerful because then you can point an agent at a copy of your database and it can wire together all of your operational and business metrics at the same time without having to try and say, connect a user profile between grafana/salesforce/hubspot/intercom etc. eg. “Claude, tell me which customers activate and convert and which churn and why and make a plan to optimise the funnel”. Hooking that up has never been easier https://insidertrades.directory/built-with-rails/ahoy-blazer... ie. build, don’t buy in the age of AI
No, the right way is to just push the transactional data into your analytical database periodically, and then point the model at the analytical database.
On every place I worked, logs, traces and metrics represent approximately 100% of their data by size. The transactional data is a rounding error.
Ahoy tracks user behavior, but it can't give you spans. If you want to know exactly how many milliseconds an external API or DB query took, you need OTel.
Would love to see what people are doing to visualize in Grafana.
If you are using the OtelCollector, you can export logs to Grafana Loki, to have them visualised
see: https://grafana.com/docs/loki/latest/send-data/otel/
Right now we mostly use it for log-to-trace navigation and basic service health metrics (like error rates, p95 latency, and overall request volume).
I’d contend having user behaviour logged directly into your monolithic app db using the ahoy gem instead of some external grafana/etc system is much more powerful because then you can point an agent at a copy of your database and it can wire together all of your operational and business metrics at the same time without having to try and say, connect a user profile between grafana/salesforce/hubspot/intercom etc. eg. “Claude, tell me which customers activate and convert and which churn and why and make a plan to optimise the funnel”. Hooking that up has never been easier https://insidertrades.directory/built-with-rails/ahoy-blazer... ie. build, don’t buy in the age of AI
No, the right way is to just push the transactional data into your analytical database periodically, and then point the model at the analytical database.
On every place I worked, logs, traces and metrics represent approximately 100% of their data by size. The transactional data is a rounding error.
Ahoy tracks user behavior, but it can't give you spans. If you want to know exactly how many milliseconds an external API or DB query took, you need OTel.
Sending user data to Anthropic?