What is the pipeline?
The pipeline is the sequence of steps the Flockr engine runs on every page load: read the live demand state of every visible product, generate candidate messages from each signal family, filter by eligibility, score the candidates, allocate the available slots, and return the selected messages. It runs end-to-end in milliseconds, on every request.
What are the stages of the pipeline?
Six, in order: read live demand state; generate candidates from the signal families; filter those candidates by eligibility; score the survivors; allocate the available slots to the highest scorers; return the result. Each stage narrows the field — from everything that could be said, to the few things worth saying here.
Why run it on every request?
Because demand is live. A product’s signals, lifecycle state and stock can all have changed since the last page view, so the pipeline recomputes from the current state every time rather than serving a cached decision. The message a shopper sees reflects the store as it is at that moment.
How do thresholds, eligibility and scoring fit into the pipeline?
They are its filters and ranker. Thresholds decide which signals qualify, eligibility decides which messages are appropriate, and the message score decides which appropriate candidate wins each slot. The pipeline is the order those checks run in, fast enough to happen inside a page load.
See the pipeline in action
The engine runs the full pipeline on every page load, in milliseconds — signals in, the right message out.