Public CRE benchmarks,
k-anonymized.

Quarterly CRE benchmarks aggregated from anonymized Ilora.ai customer portfolios across 11 canonical sectors and 91 property types. No cell is published below five properties; below-floor coverage is disclosed publicly as “coverage building” so you can see exactly where the dataset still has weight to gather.

Benchmark categories

Methodology & trust

K-anonymity floor
Every published cell aggregates at least five distinct properties. Cells with fewer are returned with NULL aggregates and a “coverage building” status — the database CHECK constraint aggregates_match_status_check enforces this; a buggy aggregator cannot leak a partial median.
Source
Anonymized aggregations of Ilora.ai customer metrics rows. Customer identity, property identity, and individual financials are never disclosed. Only quarter-grain medians + the five-number summary (p25/p50/p75/IQR bounds).
Cadence
Drafts compute on every publish_benchmarks.py run. Promotion to status='published' happens only after a human review pass. Subscribe via the Atom feed to be alerted when new cells publish.