Frameworks you can put your own numbers into.

1. 01 · prioritization · portfolio management · evidence

   Value Confidence Calculator

   Score the evidence behind a value estimate, not the estimate itself. Decompose the number into factors, rate each by what you can prove, and see which single unknown decides the outcome — and what it costs to resolve.

   Open the calculator →
   evidence per factordrag to rate

   Reachable accountsmeasured × 0.95Attach ratestated × 0.50Annual contract valuetraded off × 0.70

   0.95 × 0.50 × 0.70 = 0.33$620,525

   adjusted annual value · from a $1.87M raw estimate

2. 02 · sops · chrome extension · privacy

   Switchback

   A Chrome extension that turns a browser workflow into a step-by-step SOP: one instruction and one annotated screenshot per click, with sensitive fields redacted before you export to Word, PDF or Markdown. No account, no server, no network access.

   Add to Chrome, free →
   you click → it writes the SOP0 network requests

   HomeOrdersReturns

   #BC-48213jane.d@████████Issue refund

   Refund an order1Click Orders in the sidebar2Search for the order number3Click Issue refund

   Emails, names, and card numbers are redacted before export..docx.pdf.md

3. 03 · experimentation · causal inference · measurement

   Experiment Designer

   For changes you can't A/B test — the copy lives on the item, pricing has to be identical for everyone, or it ships to twelve sites on a fixed schedule. Pick the design that survives the constraint, size it, and read the result without overclaiming.

   Open the designer →
   unit of analysispick one

   Each sessionEach item · ICC 0.02Each item · ICC 0.05

   −3%0+16%

   95% interval · × 2.98 design effect+0.6% to +12.2%

   interval clears zero: report the lift

4. 04 · ai evaluation · measurement · llm judges

   The Eval Stack

   Design an LLM evaluation you can defend. Pick a stack for your situation, see what each mix of data, grader and cadence can and cannot catch, watch judge bias move a ship decision, and size the test set before anyone quotes the number.

   Open the Eval Stack →
   judge sensitivityspecificity 0.90

   Judge sensitivity

   true gain

   +2.0

   judge reports

   +1.6

   sensitivity 0.90 · share of every gain kept× 0.80

   acceptable attenuation

5. 05 · requirements · bdd · product discovery

   From Ask to Spec

   Nobody hands you a clean problem. Take the solution, goal or bug you were given apart into the problems behind it, write requirements against those, and leave with a .feature file that traces every scenario back to the problem it solves.

   Open Ask to Spec →
   asked for

   Add a split-shipment button to the order screen.

   P1One backordered item holds the entire order; agents split by hand, ~10 min eachobservedP2Customers aren't told their order will arrive in piecesreportedP3Split orders are charged shipping twiceobserved

   @req-R1 @problem-P1
   Scenario: Backorder beyond threshold triggers a split
     Given an order with 2 in-stock lines and 1 line backordered for 10 days
     And the hold threshold is 5 days
     When the order is released for fulfillment
     Then the in-stock lines ship as one shipment
     And the backordered line is held as a second shipment

More on the way. This page grows as frameworks graduate from writing to something you can use.