Decision Models Prioritise Mixed SEO Signals
Jev proved most useful when it ranked a steady stream of varied SEO signals for deeper analysis, rather than adding another scoring layer to a low-volume workflow.
Developed from a conversation between Pete Winn and Andy David
Andy first added Jev, a decision model designed for fast classification and confidence scoring, to an existing candidate-scoring workflow. It worked, but the workflow didn’t produce enough candidates to justify the extra machinery. He left it running in shadow mode for occasional tests rather than making it part of the main process.
An SEO tool gave the model a more substantial job. The tool produced different signals that had to be weighed and ordered before an agent investigated them. Jev could rank those signals and turn them into a prioritised queue, allowing the agent to spend its time on deeper analysis instead of deciding what deserved attention first.
The distinction mattered because the previous confidence scores came from the agent itself, and Andy suspected they were largely fabricated. Giving Jev the bounded task of prioritising signals produced a more useful division of labour. The decision model handled repeated sorting inside an established workflow, while the agent concentrated on the open-ended work that followed. A new model layer earned its place when the volume and variety of decisions were high enough to make that separation worthwhile.