← All articles
ComplianceFair HousingApplications

Consistent Screening Criteria: The Quiet Fair Housing Risk in Your Process

C
Castellan Team
October 1, 2024 · 6 min read

Discrimination usually isn't a decision

When people picture a fair housing violation, they picture intent: a landlord who refuses to rent to a protected group on purpose. Those cases exist, but they are not where most operators get into trouble. The far more common exposure is quiet and unintentional. It comes from applying the same criteria differently to different applicants, often without anyone meaning to.

This is the inconsistency problem, and it is the single most overlooked fair housing risk in a typical leasing operation. The criteria themselves might be perfectly lawful. The danger is in the variation: the exception made for one applicant and not another, the extra document requested from one and waived for the next, the bar that quietly moves depending on who is asking. That variation is where a discrimination claim is born, and it rarely looks like discrimination from the inside.

How inconsistency creates liability

A fair housing claim does not require proof that you intended to discriminate. It can rest on disparate treatment, applicants in similar situations being treated differently, or on disparate impact, a neutral policy that falls harder on a protected group. Inconsistency feeds both.

Consider how it happens in practice:

None of these feel like discrimination to the person doing them. They feel like normal judgment calls. But a fair housing tester, or a plaintiff's attorney comparing two application files, sees a process that produced different treatment for similar applicants. The intent does not matter. The pattern does.

Where inconsistency creeps in

Inconsistency is not usually one big decision. It is a hundred small ones, scattered across people and time. The common sources:

Multiple people, multiple processes

When three leasing agents handle applications, you have three slightly different processes. One always verifies income upfront. One sometimes skips it when busy. One asks for references that the others never request. Each agent is reasonable on their own. Together they produce a process that varies by who picked up the phone.

Pressure and fatigue

Criteria get applied most consistently on a calm Tuesday morning and least consistently at the end of a chaotic day. When an agent is rushing, steps get skipped, and the steps that get skipped are not random. They tend to get skipped for the applicants who seem easiest, which is its own kind of bias.

Well-meaning flexibility

The most dangerous inconsistency is the kind that feels like kindness. Making an exception, waiving a requirement, giving someone the benefit of the doubt. Every exception you make for one applicant is a standard you did not hold another to. Generosity applied unevenly is still uneven treatment.

What consistency actually requires

The fix is not stricter criteria. It is identical application of whatever criteria you have. That requires a few disciplines:

And, separately but importantly, the criteria themselves must be lawful. Consistency applied to a discriminatory rule does not save you. The process must never touch protected classes: no questions about familial status, disability, national origin, or, in jurisdictions that protect it, source of income. In many places you also cannot ask about criminal history before a conditional offer, or about immigration status at all. Consistent application of lawful criteria is the goal. Both halves matter.

Why automation is structurally suited to this

Here is the part that is easy to miss. The thing fair housing law most wants, identical treatment of every applicant, is exactly the thing software is best at and humans are worst at.

A human applies criteria with natural variance. They get tired, they make exceptions, they read the room. An AI agent applies the same criteria in the same order to every applicant, with no fatigue, no end-of-day drift, and no well-meaning exceptions. The qualification questions are the same questions, every time, for everyone. That is not a marketing claim. It is the core mechanical property of a configured process.

This is why automating qualification can reduce fair housing risk rather than increase it, provided two conditions hold. First, the criteria built into the system have to be lawful, screened for anything that touches a protected class. Second, the system must not make the final accept-or-reject decision on its own.

That second condition is firm. HUD's 2024 guidance on AI in tenant screening is explicit that an automated system should never auto-deny without human review, that liability is shared between the housing provider and the software vendor, and that decision records should be retained and tested for disparate impact. The right division of labor is consistent automated intake and pre-screening, with the formal decision made by a human at the application stage, accompanied by any required adverse-action notices. A phone or email pre-screen is informational. It should never produce something that sounds like a denial.

Putting it into practice

If you want to tighten consistency, a practical sequence:

  1. Audit your current process. Pull a sample of recent applications and check whether the same documents were requested, the same criteria applied, the same steps taken. Variation is your finding.
  2. Codify the criteria. Write the standard, lawful and uniform, and make it the single source of truth.
  3. Standardize intake. Whether through training or automation, ensure every applicant moves through the same steps in the same order.
  4. Keep the final decision human, with records. Pre-screen consistently, decide deliberately, document thoroughly.

The bottom line

The fair housing risk in your process is probably not a decision you would ever consciously make. It is the drift: the same rules applied a little differently across people, pressure, and good intentions, until two similar applicants got two different experiences.

Consistency is the antidote, and consistency happens to be the one thing automation does better than any human team. Build lawful criteria into a process that applies them identically to everyone, keep the final decision with a person, document it all, and you have turned your biggest quiet risk into one of your strongest defenses.

See Castellan respond in under 2 minutes

We'll map your real lead flow and show you exactly where the response gap is costing you leases.

Book a demo