Part 3 · Operate
Reading a buying window
What the timeline shows, what confidence means, and how to disagree with it.
01What it is
A read is a conclusion about a company with a dated spine of the events behind it, a confidence number and a verdict. The timeline is the signature surface of the product.
02Why it exists
A conclusion you cannot check is a guess with a number on it. The spine exists so you can look at the three things that happened and decide whether you agree.
Confidence is always a number and never an adjective. "High confidence" hides the difference between 0.62 and 0.94, and those are different decisions.
03How to use it
- 01Read the spine top to bottom before the verdict.
- 02Check the dates. A window built on something from four months ago is weaker than the number suggests.
- 03Open the sources. Every node links to where it came from.
- 04Disagree in the interface when you disagree. Silence teaches nothing.
04What good looks like
- You give it
- A window with three dated nodes, two of them load-bearing, at confidence 0.78.
- You get
- You can see the hiring, the pricing change and the funding, with dates, and judge the conclusion.
- Why
- Load-bearing nodes are marked, so you can see which events actually carried the verdict.
Example data, from the demo workspace
05What weak looks like, and what it costs
- You give it
- A window you disagree with, closed without a response.
- You get
- The same pattern fires again next month.
- Why
- Negative outcomes feed allocation, and a read nobody marked produced no signal either way.
Derived from the rules in the code