Ghost Tech Detection — High-Revenue, Low-Headcount UK Anomalies
Most credit bureaus throw the headcount field away. We treat it as a primary anomaly axis, and the resulting Ghost Tech feed is one of the cleanest UK alt-data factors a quant desk can wire in.
Ghost Tech: Revenue / Headcount
When revenue per employee crosses the SIC peer 99th percentile, you're looking at one of three things, a shell company, a transfer-pricing artefact, or a genuinely AI-driven outperformer. Each is tradable; we just flag the anomaly.
(turnover / average_headcount) ≥ percentile_99(SIC_peer_set) AND average_headcount ≤ 10
Revenue-per-employee benchmarked against the live SIC peer-set distribution. Companies appearing in the 99th percentile *and* operating with 10 or fewer staff are tagged as Ghost Tech candidates and pushed to the alt-data feed.
Companies tripping this signal in the last 24h
Sample of records flagged by the Ghost Tech: Revenue / Headcount signal. The full feed updates within minutes of each Companies House submission.
Why Ghost Tech is the cleanest UK alt-data factor
The factor works because UK iXBRL filings are unusually disclosure-rich. Companies House asks every UK private company to declare, at minimum, turnover, average headcount and wages-and-salaries. That gives any analyst three independent inputs to a single ratio (revenue per employee) on the entire population, refreshed every business day, with full historical depth.
What you do with the anomaly depends on your strategy:
- Shell-company detection for compliance and AML teams, we cluster the population by ownership, registered office and directorship overlaps to surface structures of interest.
- Transfer-pricing surveillance for tax-research desks, recurring high-revenue/low-headcount entities owned by overseas parents are a known artefact of intra-group invoicing.
- AI-firm signature detection for venture and growth-equity teams, genuinely AI-leveraged services businesses are starting to dominate the 99th-percentile band, and their financials look unmistakable next to traditional consultancies.
We deliver the feed three ways:
- Daily JSON dump of every entity newly entering the 99th percentile
- Five-year Parquet history for backtesting (no look-ahead bias, point-in-time clean)
- Webhook fired on first crossing for desks running an event-driven strategy
Every record is auditable back to the originating iXBRL filing reference, so research compliance can verify the trigger.
Want the Ghost Tech feed point-in-time clean for backtesting?
The Data Market exposes five years of Ghost Tech signal events with no look-ahead bias, JSON, Parquet or CSV delivery. Each row is auditable to the originating Companies House filing reference.