Follow the recipe once, then keep the parts that help.
Use the situation as your starting point, follow the steps, and save the parts that become a habit.
Try saying
“I am in this situation. Walk me through the recipe and tell me what good looks like.”
Name the situation
Start with the real moment you are in.
Follow the recipe
Do one step, then check the result.
Make it yours
Save the workflow when it works.
Track feature impact with outcome loops
Situation
You shipped a feature three weeks ago. Someone in the board meeting asks: "Did it move the number?" You don't know. The experiment data is in PostHog, the customer feedback is in Slack, and the OKR is in a Google Sheet nobody updates. Specky keeps the outcome, the work behind it, and your verdict in one place.
What you need
- An outcome the feature should move, saved as an objective in Goals & OKRs
- The feature as an initiative or workstream in Specky
- Optional: PostHog connected, so Specky can count events for you
Steps
1. Create the experiment before you ship (ideal path)
Open Experiments and click New Experiment:
- Hypothesis: "Adding bulk CSV export will increase activation rate for operations teams by 15%"
- Link to Opportunity (optional, shown once you have opportunities): pick the opportunity this tests
- Target outcome (optional): what should improve if this works
- Success criteria / pass signal: the result that would change your decision, for example "activation for operations teams goes from 20% to 23%"
Click Add Experiment. The pass signal locks once the experiment starts running, so the rule can't move after you see the data.
To have Specky measure it for you, click the experiment's hypothesis to edit it, click Suggest PostHog metric under Evidence after launch, then click Update. After the experiment is completed, Specky counts that PostHog event over the window and brings you a review proposal in your Inbox. It never decides the verdict for you.
2. Tie the work to its outcome
Open Did it work? (listed as Outcomes under Measure in the sidebar). In the Outcome map, find the feature under Unaligned initiatives and workstreams, click Align, choose the outcome, and click Align to outcome. The feature now sits under that outcome.
3. If you've already shipped, start from chat
Open the chat:
I shipped [feature] on [date].
Query PostHog for [event] in the four weeks before and after that date,
then create an experiment testing whether it improved [metric].
With PostHog connected, Specky can run the query and create the experiment in Experiments. Check the numbers before you rely on them.
4. Record the outcome
When the experiment is completed, or the delivery work is marked done, it shows up in Did it work? under Awaiting verdict or evidence. Click Validate now (or Add evidence) and fill in:
- Metric movement: the metric name, Before (at ship), and Now (measured)
- Evidence receipt: Metric movement, or Customer / qualitative evidence
- Verdict: Worked, Mixed, Didn't work, or Too early
- What did you learn?: one or two sentences for the next planning cycle
Click Save outcome →. Not ready yet? Click Snooze 1 week.
5. Use the outcome for next quarter's bets
Once you have measured three outcomes, open AI pattern synthesis in Did it work? to see patterns across them. You can also ask in chat:
Summarise all experiments from Q1 2026.
Which hypotheses were validated, which were invalidated, and what patterns
do you see in what's working vs. what's not?
Output
- A record that links the work, its outcome, the evidence, and your verdict
- Your batting average over time: the share of measured outcomes that were a win
- An AI pattern synthesis across outcomes to inform next quarter's strategy
Variations
No PostHog? Type the before and now numbers into Metric movement yourself, or choose Customer / qualitative evidence under Evidence receipt and describe what you saw.
Leading indicators only? Set the pass signal on a leading indicator (e.g., "feature used in first 3 days") rather than the lagging metric (retention). You get faster signal, even if it's not the final answer.
Ready to apply this?
Start with your own product and keep the first read grounded. You'll create an account before the guided result so it can be saved to your workspace.