Situation

Something changed — signups, activation, retention, support volume, adoption, or revenue — and the team is already forming explanations before anyone has checked the evidence.

What you need

• One current product question
• A connected source that can answer the “what” (PostHog, Amplitude, Mixpanel, Stripe, or another live MCP)
• One source that can explain the “why” (Slack, Help Scout, JustCall, Gong, Intercom, Zendesk, or research)
• A decision boundary: what you may do if the evidence is strong, and what you will not do yet

Steps

1. State the question precisely

Open Product intelligence and ask one question with a time range:

Why did activation change for new workspaces in the last seven days compared with the previous seven?

Avoid “tell me what is happening” until you know the decision you are trying to make.

2. Let the live source answer first

Use Ask Specky. The production assistant has the full built-in tools and connected MCP tools available; there is no keyword intent gate deciding whether it is allowed to investigate. Ask it to query the current analytics source first, name the tool/source it used, and show the relevant comparison.

3. Add the human explanation layer

Ask a follow-up:

Now compare that change with releases, support conversations, Help Scout threads, and customer calls from the same period. Separate evidence from hypotheses and quote the strongest examples.

4. Choose the smallest useful next action

If the evidence is strong, create a priority, experiment, or PRD. If it is mixed, create an investigation brief or interview campaign. If it is weak, record the missing data rather than pretending the answer is known.

5. Close the loop

When the change ships, create an outcome validation with the baseline, target, segment, and review date. Return to Product intelligence until the result has a verdict.

Output

• What changed, with the comparison and time window
• Which sources were checked and what each one supports
• The strongest customer language or delivery context
• Unknowns, contradictions, and confidence
• One recommended next step with a human review point

Variations

Support spike: start with Help Scout or Zendesk, then ask PostHog whether the affected workflow shows a matching behavioural drop.

Voice-heavy product: start with JustCall, Gong, fonio, or ElevenLabs, then link repeated call language to an opportunity.

Engineering-heavy product: start with Shortcut, Trello, Jira, Linear, or GitHub to see whether a release or incident lines up with the customer signal.