How to directly access your PM data
This is less a tool choice than a habit change: the goal is that no product question waits two days for an analyst.
What “direct access” means for a product leader. You can answer these yourself, today, without a ticket:
What share of active accounts touched this feature last month?
Which step in the flow loses the most people?
Did retention move after the launch?
Who are the ten accounts most likely to churn, and what do they have in common?
If any of those needs someone else, you don’t have direct access — you have a dashboard.
The three layers, and you need all three:
Product analytics — Amplitude, Mixpanel, PostHog. Event data: who clicked what, funnels, retention curves. Nearly all of them now have a natural-language query layer. Good for the first three questions above. Weak for anything joined with revenue, support, or sales data.
The warehouse — BigQuery, Snowflake, Postgres, wherever the real tables live. This is where the fourth question gets answered, and where the retention simulator lives. It needs SQL, or a tool that writes SQL for you.
The unstructured layer — support tickets, sales calls, NPS comments, reviews. No dashboard covers this; it’s where your interview synthesizer and status drafter get their raw material.
How to pick, or rather, how to get access:
Analytics: you usually don’t choose; you inherit. Learn what’s there deeply before asking for anything new. The tool is rarely the bottleneck; instrumentation is. Your first project is checking whether the events you care about are even being tracked.
Warehouse: ask for read-only access to the warehouse or a replica. This is the single most valuable request you can make in your first month at any company, and it’s usually granted if you ask for read-only. Then use your agentic coding tool to write the SQL — it’s good at it, and it’s the fastest way to learn the schema.
Unstructured: get an export path. A weekly dump of tickets to a folder is enough to start.
The test for any data tool:
Can you ask a question in plain language and get a number with the query shown? If the query is hidden, you can’t check it, and you’ll eventually present a wrong number to leadership.
Can you join across sources — usage plus revenue, usage plus support? Single-source tools give single-source answers.
Can you save and rerun? A question you’ll ask monthly should be a saved query, not a repeated conversation.
Can it push to you? A metric that changes should tell you, not wait to be looked at. This is the seed of the metric watcher.
Two cautions:
Natural-language-to-SQL is confidently wrong more often than people expect, especially on joins and date logic. Always look at the generated query, and sanity-check the number against something you know.
Read-only means read-only. Never ask for write access to production data as a PM. You don’t need it and it removes the safety net.
For you: the retention simulator and portfolio generator you already built are this layer, done well. The CV bullet is honest. The upgrade is making it standing rather than something you run — which is exactly item four’s “push to you” test, and Part 4 of your series.