RAN Flow Experiment

A RAN engineer describes what to test in plain language or builds it in a GUI: the change, the test cases, the cells or clusters, the schedule and the method, interleaved, sequential or A/B. RAN Flow Experiment runs the trial and shows a rollout dashboard confirming whether the configuration was applied across all target nodes or only part of them. The engineer is notified by email when execution completes and at set follow-ups, for example one, four and seven days later. Context can be added too, describing the goal of the optimisation or attaching a descriptive document, which tells the AI what to look at and which KPIs matter. No scripting, no manual KPI wrangling.

Instead of a raw before-and-after that leaves you guessing, RAN Flow Experiment measures the real effect of the change on the KPIs that matter. Every trial ends in a ranked verdict: Improved, Degraded or Unproven, each with a confidence level. It never says Improved without evidence, and an honest Unproven tells you the effect could not be shown and the change should be re-scoped. The result is a straight, defensible read an operator can take to a change board, and the evidence needed to safely accelerate optimisation, energy and feature rollouts.

When the trial completes, the agent pulls the KPIs, works out what the change did and writes the report for you: which KPIs moved, by how much, and how confident the result is. No pulling counters, no pivot tables, no stitching dashboards together. The engineer opens a finished report and decides, instead of spending days assembling one by hand.