Leaders Insights
Leaders Insights

Rester au meilleur niveau, un peu chaque jour.

DomainesMarketingDataFinanceIA
RessourcesApprendreTestOutilsBlogGlossaire
© 2026 Leaders Insights — Tous droits réservés.
Formations/Data in real estate/Governance, privacy and checks/Running a recurring data audit that catches drift before deals do
4/4+150 XP

Governance, privacy and checks

10How privacy law actually touches property data+15011Fair housing and anti-discrimination checks in scoring models+15012Building an access and permissioning model for property data+15013Running a recurring data audit that catches drift before deals do+150

Running a recurring data audit that catches drift before deals do

# Running a recurring data audit that catches drift before deals do

A regional mall's occupancy feed said 94% leased. The actual number, confirmed three weeks later during due diligence, was 81%. The gap wasn't fraud. It was a property management system that hadn't synced a batch of lease terminations for two quarters. By the time anyone noticed, the asset had already been marketed to buyers at the wrong cap rate assumption.

This is data drift: the slow, silent divergence between what your systems say and what is actually true on the ground. In real estate, where deals rely on stitched-together feeds from property managers, brokers, lenders, and municipal records, drift is not an edge case. It is the default state of the data unless someone audits it on a schedule.

What "drift" actually looks like in real estate data

Drift rarely announces itself. It shows up as:

  • Stale occupancy feeds: A property management system (PMS) like Yardi or RealPage exports leasing data monthly, but a lease-up or move-out from week two never makes it into the quarterly investor report.
  • Mismatched entity names: "123 Main St LLC" in the loan servicing system versus "123 Main Street Holdings LLC" in the title record versus "Main St Property Owner LPLPA standalone web page built for a single campaign goal, designed to maximise conversions by removing distractions and focusing visitors on one action." in the rent roll. Same asset, three names, zero automated way to know they match.
Voir la définition complète →
  • Silently-changed vendor fields: A third-party datathird-party dataData purchased from external aggregators, collected from audiences you don't own. It is bought or licensed rather than gathered through your own direct relationships.Voir la définition complète → provider (say, a comps or market-rent vendor) redefines "gross leasable area" or switches from calendar-year to fiscal-year reporting, and nobody documents the change. Your trend line now compares apples to oranges without anyone deciding to do that.
  • None of these are dramatic. All of them compound. A cap rate model built on drifted occupancy data doesn't just misprice one asset, it recalibrates every comparable valuation that references it.

    Why this is a governance problem, not just an IT problem

    Data governanceData governanceData governance is the set of policies, roles, and processes that ensure data is accurate, secure, well-defined, and used responsibly across an organization.Voir la définition complète → is the set of rules, roles, and processes that determine who is accountable for data accuracy, how it's defined, and how changes get tracked. In real estate, this matters for two concrete reasons.

    Regulatory exposure. If your fund reports occupancy or income figures to investors and those figures are wrong due to unaudited drift, you risk misrepresentation claims. In the US, the SEC (Securities and Exchange Commission) has brought enforcement actions against real estate fund managers for inaccurate reporting to limited partners, under general anti-fraud provisions of the Investment Advisers Act. In the EU, the Alternative Investment Fund Managers Directive (AIFMD) requires periodic, accurate reporting to regulators and investors, and national regulators (like Germany's BaFin or France's AMF) can sanction managers for material misstatements even when unintentional.

    Privacy exposure. Tenant data, rent rolls with individual names, payment histories, and access logs from smart building systems all count as personal data under the EU's General Data Protection Regulation (GDPR) and, increasingly, US state laws like the California Consumer Privacy Act (CCPA). A drifted or duplicated tenant record isn't just an accuracy problem, it's a data minimization and accuracy problem under GDPR Article 5, which requires personal data to be "accurate and kept up to date." Regulators have fined companies for retaining or processing outdated personal data long after it should have been corrected or purged. (See the European Data Protection Board's guidance on accuracy for the underlying principle.)

    Building the quarterly audit: a practical routine

    You don't need a data science team to run this. You need a checklist, a schedule, and someone accountable for closing each finding.

    Step 1: Reconcile entity identity across systems

    Pull entity names from your loan servicer, PMS, title records, and investor reporting platform. Run a fuzzy match (even a simple spreadsheet function works for small portfolios) to flag near-duplicates.

    python
    from rapidfuzz import fuzz
    
    entities = {
        "loan_system": "123 Main St LLC",
        "title_record": "123 Main Street Holdings LLC",
        "rent_roll": "Main St Property Owner LP"
    }
    
    pairs = [("loan_system", "title_record"), ("loan_system", "rent_roll")]
    for a, b in pairs:
        score = fuzz.token_sort_ratio(entities[a], entities[b])
        print(f"{a} vs {b}: similarity {score}")

    A similarity score below roughly 70 to 80 (on a 0 to 100 scale) is a flag worth a human look, not an automatic fix. The point is triage, not automation replacing judgment.

    Step 2: Check feed freshness, not just feed presence

    Ask: when was this occupancy or rent roll figure last updated, and does that date match the reporting period label? A "Q3 occupancy" figure last modified in July is stale by definition. Most PMS platforms log a last-modified timestamp; if yours doesn't expose it, that absence is itself a governance gap worth escalating to the vendor.

    Step 3: Diff vendor field definitions quarter over quarter

    Keep a simple data dictionary (a one-page log of field names, definitions, and units) for every external data source: comps providers, market-rent indices, ESG (environmental, social, governance) scoring vendors. Each quarter, request the vendor's current methodology note and compare it to last quarter's. CoStar, Green Street, and similar providers do periodically revise methodology. It is your job to catch when they do, not theirs to flag it loudly.

    Step 4: Sample-test against ground truth

    Pick a random 5 to 10% sample of units or leases each quarter and verify against a primary source: an actual lease document, a site visit report, or a utility billing record. This is the equivalent of an internal audit spot-check, and it's the single best defense against systemic feed errors that pass every automated check but are simply wrong.

    Step 5: Log findings and assign owners

    An audit that produces findings nobody acts on is theater. Every flagged discrepancy needs a named owner and a resolution deadline, tracked in whatever system you already use for issue management. This is the governance layer, not the technical layer, and it's usually the one that's missing.

    Vérification des acquis

    1. In the mall example, what was the root cause of the occupancy discrepancy discovered during due diligence?

    2. Which best defines 'data drift' as described in the lesson?

    3. Why does the lesson argue that data drift is a governance problem rather than purely an IT problem?

    4. A comps vendor quietly changes its definition of 'gross leasable area' without documentation. What is the most significant downstream risk this creates?

    CHOIX MULTIPLES

    5. Select ALL correct answers about how data drift typically manifests in real estate data.

    Sélectionnez toutes les réponses correctes.

    CHOIX MULTIPLES

    6. Select ALL correct answers about why drift is especially dangerous in real estate deal-making.

    Sélectionnez toutes les réponses correctes.

    Who should own this, and how often

    For a portfolio of meaningful size, this shouldn't sit solely with IT or solely with asset management. The best-functioning setups split it:

    • Asset management owns the ground-truth checks (Step 4) because they know the properties.
    • A data or operations lead owns the systems reconciliation (Steps 1 to 3) because they know the platforms.
    • Compliance or legal reviews findings that touch investor reporting or tenant personal data, because that's where GDPR and SEC/AIFMD exposure lives.

    Quarterly is the right default cadence for most institutional portfolios: frequent enough to catch drift before it compounds into an annual valuation cycle, infrequent enough to be sustainable without dedicated headcount. Funds preparing for a sale, refinancing, or annual audit should tighten to monthly in the two quarters before that event.

    🎬 [VIDEO: "Data GovernanceData GovernanceData governance is the set of policies, roles, and processes that ensure data is accurate, secure, well-defined, and used responsibly across an organization.Voir la définition complète → Fundamentals" - youtube.com/results?search_query=data+governance+fundamentals+explained - a plain-language walkthrough of governance roles, data ownership, and accountability structures applicable across sectors including real estate]

    Key Takeaways

    • Data drift (stale feeds, mismatched entity names, silently-changed vendor definitions) is the default outcome of multi-system real estate data, not a rare failure. It requires a scheduled audit, not a one-time cleanup.
    • Governance failures here carry real regulatory weight: SEC anti-fraud rules and AIFMD reporting obligations in fund contexts, and GDPR's accuracy principle (Article 5) wherever tenant personal data is involved.
    • A practical quarterly audit has five steps: reconcile entity names across systems, check feed freshness against reporting period labels, diff vendor field definitions quarter over quarter, sample-test against ground truth, and log every finding with a named owner.
    • Split ownership across asset management (ground truth), data/operations (systems reconciliation), and compliance (regulatory and privacy exposure). No single team should own the whole audit.
    • Tighten the cadence from quarterly to monthly ahead of a sale, refinancing, or annual audit, when the cost of undetected drift is highest.

    Précédent

    Building an access and permissioning model for property data