Skip to content

Running a what-if membership analysis

What-if analysis (sidebar Financials → What-If Analysis, /rebate/whatif, gated to the internal ebiz role) models how membership changes ripple through rebate earnings across the portfolio: what happens to everyone else's rebates if a member leaves, a new member joins, or a member's turnover grows or shrinks. It is read-only — nothing it does touches live data.

This is different from the per-deal rebate simulation: simulation re-runs the real calculation for one deal with edited turnover and stores results; what-if uses a separate shadow evaluator, works across many deals, and changes membership/turnover rather than deal terms.

Steps

  1. Choose the scope — the whole portfolio or selected deals.
  2. Build an ordered list of mutations:
  3. Remove member (with an effective date),
  4. Add member (modelled from a comparable member's profile, scaled by a factor),
  5. Scale member (turnover × a factor over a date range).
  6. Set the policy toggles (how a leaver's baseline is treated, whether a leaver reduces the member count, per-member detail) and the projection horizon (defaults to year end).
  7. Run the scenario. It computes in the background with a progress bar; results load when ready.

Result

The analysis shows baseline vs scenario deltas: headline cards (impact on remaining members, group turnover removed, effective rate, net change), by-deal and by-member tables with drill-down to element/member/ period rows, tier-band shifts, and plain-language explanations of why remaining members were or weren't affected. Each row carries a confidence flag (exact / projected / modelled / frozen) and warns when a deal has no projection data.

The parity check compares the shadow evaluator's baseline against live rebate payments, flagging matches and discrepancies — useful for trusting the model before acting on it.

Results are held in a short-lived in-memory cache only (about 30 minutes); nothing is written to the database or the search index.

If it fails

  • Numbers look low-confidence — deals without projection data fall back to modelled estimates; the row flags this.
  • A re-run returns instantly — identical scenarios are served from cache; change an input to force a recompute.

A note on the linear impact report

A related printable report (the linear-rebate impact analysis) is generated by the same engine. Be aware it is served from a public/whitelisted endpoint that requires no login — it exposes deal names, supplier names, and projected rebate figures for the tenant (scoped only by the request hostname). It is read-only, but treat its URL as sensitive.