Normal Distribution Bundle Laravel Package

Product Decisions This Supports

• Pricing and discount engines. A normal CDF lets you convert a raw score (creditworthiness, churn risk, demand signal) into a percentile you can act on — e.g. "offer retention credit to the bottom 15% of accounts." That is a product rule a PM can own and tune without re-shipping code.
• Experimentation velocity. Sample-size and significance calculations for A/B tests need an inverse normal CDF (z-score lookup). Owning that math in-house unblocks the "how long must we run this test?" question that currently stalls roadmap debates.
• Forecasting with explicit uncertainty. P50/P90 delivery and capacity forecasts (Monte Carlo over task estimates, safety stock, SLA headroom) are all normal-distribution arithmetic. A PM can show stakeholders a confidence band instead of a single optimistic date.
• Threshold setting for anomaly/fraud alerts. Choosing alert cutoffs by false-positive rate (2σ vs 3σ) is a product tradeoff, not an engineering one — but only if someone can compute it cheaply.
• Build vs. buy framing. The need is real and recurring; the package is commodity math. That distinction drives the decision below.

When to Consider This Package

Adopt only if all of the following hold:

• You are on Symfony, not Laravel. This is a Symfony bundle (Happyr/NormalDistributionBundle) — it registers a service via the DI container. It will not drop into a Laravel app without hauling in Symfony HttpKernel/DI and writing your own service-provider shim, which is more work than the math itself.
• You have a legacy Symfony app already running PHP 7.x and need PDF/CDF/inverse-CDF for a short internal spike.
• You accept it as frozen, unsupported code you may have to fork.

Look elsewhere — in practice, almost always — because:

• Zero stars, zero dependents, last release September 2020. No adoption signal means no peer review of the numerics, no security disclosure path, and no evidence of PHP 8.x compatibility. Numerical code with no users is numerical code with unverified edge cases (tails, underflow, inverse-CDF convergence).
• Tiny surface area. The functions you actually need are roughly 50–100 lines of well-documented PHP, or one call into a maintained stats library. Vendoring 2019-era framework glue to get them is a bad trade.
• No roadmap influence. You cannot file an issue against an abandoned bundle and expect a fix; you inherit the maintenance burden silently.

How to Pitch It (Stakeholders)

To executives: There is a genuine product need here — percentile-based pricing, experiment sizing, and confidence-banded forecasts. But this specific package is abandoned (last release 2020, no users, no maintainer activity) and is built for Symfony, not our Laravel stack. Adopting it converts a one-day task into an indefinite maintenance and security liability with no support path. Recommendation: don't adopt; fund a one-day in-house utility module (or a maintained statistics library) instead. We get the same capability, we own the tests, and we control the upgrade path.

To engineering: Treat this as a reference for the API shape, not a dependency. Implement pdf, cdf, and inverseCdf (Acklam or Wichura's AS241 for the inverse), add property-based tests against known z-tables, and wrap it behind a small service so the underlying library stays swappable. If someone later wants the bundle, the migration cost is near zero — which is exactly why we shouldn't take the dependency now.