B2B software (holdco) 700 FTEs · ~$150M ARR · 5-entity (US/UK/JP/IN/AU) Multi-entity close + intercompany matching

APAC holdco, 5 entities, ~700 employees (anonymized)

Per-entity materiality + FX-aware intercompany matching across US/UK/JP/IN/AU close cycle

An APAC-headquartered holdco with 5 entities adopted closegate's per-entity materiality overrides + intercompany matcher Protocol. Cut close-window overlap incidents by 90%.

Outcomes

  • Multi-jurisdiction close cycle: 14 days → 9 days
  • Cross-entity intercompany clearing accuracy: 73% → 96%
  • Multi-timezone close-window overlap incidents: from 4/quarter to 0/quarter
  • ECB daily FX feed adopted as audit-defensible source

The CFO of an APAC-headquartered holdco had a specific problem: 5 entities (US, UK, JP, IN, AU) running close cycles in 5 timezones, with significant intercompany activity across them, and a stretched team trying to manage materiality, FX rate sources, and audit evidence across all of them.

The starting point

  • 5 entities: US parent ($USD), UK sub (GBP), JP sub (JPY), IN sub (INR), AU sub (AUD)
  • 700 FTEs total; finance team of 12 split across 4 locations
  • Intercompany activity: ~$15M/yr in cross-entity charges (transfer pricing, shared services, license)
  • Close-window overlap problem: APAC entities close before US sometimes; reverse on quarter-end; intercompany entries were posting to the wrong period frequently
  • 4 close-window overlap incidents the prior quarter (where US-parent close was finalized before JP-sub had locked, causing reposting)

The CFO’s brief: “We need a single source of truth for materiality across our entities + a defensible FX source + audit-defensibility across all 5 jurisdictions.”

The architecture

  • Engine: closegate-engine on AWS, multi-AZ
  • Per-entity materiality overrides: US $10K, UK £8K, JP ¥1M, IN ₹500K, AU $15K
  • FX adapter: ECB daily rates (free, defensible source); fallback to OpenExchangeRates paid tier for non-major currencies
  • Intercompany matcher: AccountCodePairMatcher with chart-of-accounts conventions:
    • 1500-UK-* ↔ 2500-US-*
    • 1500-JP-* ↔ 2500-US-*
    • 1500-IN-* ↔ 2500-US-*
    • 1500-AU-* ↔ 2500-US-*
  • Multi-timezone calendar awareness: each entity carries (timezone: IANA, cutoff_time, calendar: enum). The pipeline surfaces multi-tz close-window overlap as a first-class signal in the workspace UI.

The per-entity materiality move

Before closegate: one global $10K threshold (USD). Result: a ¥500K transaction (~$3K USD) at the JP sub would auto-confirm even though it was meaningfully above JP-local materiality conventions.

After closegate: per-entity overrides. The JP sub’s ¥1M threshold (~$6.7K USD) fires HITL routing on the JP-local view; the runtime converts via the ECB daily rate for the US consolidated view; the audit event records both numbers + the FX source.

The JP-sub controller’s quote: “for the first time, the gate is enforcing the same standard our local audit firm would apply manually.”

The intercompany matcher win

Before: 73% intercompany clearing accuracy. Misses came from:

  • FX rate slippage between booking date and clearing date
  • Account-code variation across entities (US uses 2500-INTERCO-JP, JP uses 2500-INTERCO-US-PARENT)
  • Period-boundary timing (US-parent books in March; JP-sub books in April due to JST cutoff)

After closegate’s AccountCodePairMatcher with tolerance_pct: 0.5 for FX slippage: 96% accuracy. The misses are now genuine reconciliation discrepancies rather than mechanical errors.

The close-window overlap defense

closegate ships multi-tz close-window overlap detection as a first-class signal. Each entity’s (timezone, cutoff_time) config drives the calculation: if a state change would land in entity A’s locked period but entity B is still open, the UI shows an amber banner and the gate fires HITL.

In the 4 quarters before closegate: 4 overlap incidents. In the 4 quarters after: 0 incidents. Two near-misses, both caught by the HITL routing.

TCO

Cost lineAnnual
AWS infrastructure (multi-AZ, 5-node)$14,400/yr
Anthropic API (Sonnet 4.6)~$32,000/yr
ECB FX feed$0 (free)
Existing close-cycle tooling (continued)(existing)
One-time implementation (in-house eng, ~140 hours)$28,000
Year 1 incremental~$74,400
Year 2+~$46,400/yr

For a 5-entity holdco, a vendor like BlackLine + their professional-services engagement to wire multi-entity materiality would have been $400K+/yr. The savings funded the 4-headed close-cycle team’s tooling for the year.

What worked

  • Per-entity materiality as first-class config. The single most-cited win from the CFO + audit committee.
  • ECB daily FX feed. Free, public, defensible. The audit firm specifically called this out as “what we want to see from cross-border AI deployments.”
  • Multi-tz overlap detection. The 4-to-0 incident reduction was the cleanest before/after measurement in the pilot.

What was painful

  • JP-sub controller skepticism. The JP team had been burned by previous “cross-border tooling” promises. Took 2 cycles for them to trust the per-entity materiality + the JP-local audit firm walking the closegate controls. After that, the JP team was the most enthusiastic adopter.
  • Account-code conventions. Each entity had grown its own chart-of-accounts conventions; mapping them to the 1500-X ↔ 2500-X pattern required a 3-week project. Not closegate’s fault; would have been required for any tooling.
  • Timezone arithmetic edge cases. DST transitions in different countries broke our first close-window-overlap implementation. Took 2 weeks to handle correctly. Now well-tested.

What’s next

  • Adding a 6th entity (SG sub, Singapore) in Q3
  • Multi-currency consolidation workflows (currently entity-by-entity)
  • Integration with the JP-local FX feed for higher-precision rates on Japan-only transactions

This case study is published with the design partner’s permission. Company name + revealing details anonymized; numbers cited are accurate.

Case study published with the design partner's permission; company name and revealing details anonymized at their request. The numbers cited are real.

Inbound

Run a pilot like this one

Two design-partner slots open this quarter. One real workflow, your real policy.yaml, monthly 30-min sync, direct line to the maintainer.