Business question
How do we measure whether our position improved?
By running the same requirement sequence again and comparing against a starting point. Improvement is a measured change in how AI recommends you, not an assumption that an action worked. Because results vary run to run, movement only counts when it stands out against that normal variation.
The problem
Most changes are made on faith. Without a starting point and a repeatable measurement, you cannot tell whether a change moved your position or the result simply varied on its own.
Why it matters
Measured movement is what separates a change that worked from one that only felt productive. It is also what lets you stop spending on things that do not move the outcome.
What the research studies
The components behind this question.
The evidence so far
What we have run, and what we have seen.
- E-045Video conferencing software: When the field survives and keeps its names
- E-041CRM software: When the recommendation set survives but its members do not
- E-043Accounting software: When the whole field is wiped out and replaced
- E-044Accounting software: When the same kill repeats on a second evaluator
- E-042Project management software: The unexplained recommendation transition
- E-034Clinical trial management software: When the recommendation set collapses
- E-040Project management software: When the recommendation set holds
- Observation · Candidate · 2026-07-20Survival appeared in two modes, and their cause is confounded
- Observation · Candidate · 2026-07-20Survival depended on the category, in at least four ways
What we can conclude
We have observed that AI recommendations vary across repeated runs, which is exactly why measurement has to be against a starting point and across many runs. Movement reported against a starting point is a real result; a single run is not.
What remains unknown
How much variation is normal for a given category, and whether an observed movement was actually caused by a specific change, needs a controlled before-and-after test rather than a single comparison.
How Upstream Zero helps
From the evidence to your answer.
Upstream Zero sets a starting point, runs the sequence again after a change, and reports the movement against it: how often you make the shortlist, how often you lead it, and how often you survive to the end. The number is a measured outcome, reported with its limits.