The math is simple — interpretation drives decisions.
Start with link ratio stability across the triangle. Jumps, “special years,” or process changes suggest a shorter lookback or portfolio splits. Decide if a tail factor is warranted and how much weight it carries versus data.
Check paid vs incurred coherence; persistent gaps can be more telling than the IBNR itself. If you enable Mack, look at interval width — is precision sufficient for reinsurance or capital decisions?
A practical flow: triangle heatmap → link ratio distribution → ultimate/IBNR vs history → variants (window, tail) → sanity checks (seasonality, reserve policy).
Takeaway: Don’t anchor on a single output. Judge stability, uncertainty, and coherence (paid/incurred) together.