Retiring 100+ legacy scripts

$350K saved in professional services. At least a week back on every major upgrade.

Role: Architect · Platform: ServiceNow

Legacy script refactor: before and after Before: more than 100 legacy third-party scripts built up on the platform made upgrades slow and risky. Every script was reviewed against out of the box, then split, merged, kept or deprecated, with much reverted to out of the box, done in house. After: a smaller custom layer, 350 thousand dollars saved in professional services, and at least one week cut from every major upgrade. BEFORE 100+ legacy third-party scripts, built up over five years Out-of-the-box platform • Upgrades skipped new features, or broke things • Script upkeep stretched every upgrade cycle HOW Every script reviewed against out of the box Split Merge Keep Deprecate Much of it reverted to out-of-the-box features. Done in house. AFTER A smaller, reviewed custom layer Custom scripts Out-of-the-box platform, doing more $350K saved in professional services 1 week+ cut from every major upgrade Faster system, smaller codebase

Context

Five years into the platform's life, customisation had piled up. More than 100 legacy third-party scripts sat on top of out-of-the-box behaviour, and the number kept growing.

Problem

Every upgrade became a trade-off. Either we held back new platform features, or we upgraded and things broke. The team spent a huge amount of time maintaining scripts instead of delivering, and the upgrade cycle stretched.

Decision

We priced vendor help, and the quote was high. So we did it in house: rewrite the scripts we still needed, and return everything else to out of the box.

What we built

  • Reviewed and listed every script.
  • Compared each one with out-of-the-box behaviour to find what it actually changed.
  • Made one of four calls on each: split it into a separate script, merge it with others, keep the existing change, or deprecate the custom behaviour.
  • Rewrote what stayed and retired the rest.

Result

  • $350K saved in professional services.
  • At least one week cut from the development cycle of every major upgrade.
  • A faster platform with a smaller codebase.

What I'd do differently

We underestimated how long it would take. Next time I'd plan buffers for releases already in flight and for changes landing mid-project.