Not every business problem is a machine learning problem. Before a team commits data scientists, infrastructure and months of iteration, it helps to score the problem against five structural questions — the PIERS checklist. This scene renders your five scores as a five-sided radar "crystal": each pylon is a factor, its height is your score, and the faceted cap connecting the tops shows the overall shape of the fit. A dashed pentagon marks the pass threshold, and the central core glows purple and rises into a beam only when every factor clears the bar.
Most failed ML projects are killed not by bad modeling but by bad problem selection — teams build an accurate model for a decision nobody was actually going to act on, or for an event too rare to have learnable structure.
Score a candidate business problem against the five PIERS factors and watch a 3D radar crystal rise from a pentagon platform — its shape, color and central beam reveal at a glance whether the problem is genuinely ready for machine learning.
Each pylon's height encodes one factor score — Pattern, Impact, Examples, Repeatability, Simple action. The faceted cap and outline trace the overall profile, a dashed ring marks the pass threshold, and the core glows and beams green only when every factor clears the bar.
Drag the five sliders to score a real or hypothetical problem from 0–10. Toggle the threshold ring, or load a worked example — a well-framed churn-prediction brief, a one-off decision, a vague metric, or a textbook ideal case — to see the crystal reshape instantly.
Most failed ML initiatives aren't killed by weak models — they're killed by weak problem selection: predicting something nobody will act on, or something too rare to have learnable structure in the first place.