The Right Match
Diagnose risk. Prescribe the next step. A hospital does not triage a patient by vibes — and neither should you.
Fifty Files and One Afternoon
A landlord opens fifty applications on a Sunday afternoon. By file eleven he is skimming. By file twenty-six he is looking at names and incomes only. By file forty he has stopped reading and started sorting by feel.
He is not lazy. He is a person, doing something people cannot do.
This is the failure nobody talks about, because it does not look like a failure. There is no fraud, no dispute, no bad tenant. There is just a tired man at a kitchen table making forty-nine decisions he could not possibly explain the next morning.
And here is what happens inside that fatigue. Instinct quietly takes over. Not as a deliberate choice — nobody decides to abandon rigour. It simply fills the space where attention used to be, and it feels exactly like judgement.
So the smooth file wins. The fast responder wins. The familiar-sounding name wins. The applicant who writes like the landlord writes, wins.
More information does not automatically create better decisions. Without structure, it usually creates worse decisions with more false confidence.
That is the paradox at the centre of this pillar. The rental market is not short of applications. It is short of a system that can read information properly, verify what is real, and explain why a decision was made.
Which brings us to a number I want you to sit with, from my own research:
Zero. The number of times "we had a bad feeling" has stood up as a real process standard.
Why This Pillar Comes Fourth
Screening is where every landlord believes the skill lives. It is the part we discuss with other landlords, the part we think we are good at, the part that feels like expertise.
It is the fourth step for a reason, and the dependency is absolute.
- If Identify did not confirm the person is real, the documents may belong to nobody.
- If Data did not organize and source-verify the file, you are assessing fragments and stories.
- If Engage did not deliver information properly, key facts arrived late, or never arrived, and the result is distorted before you begin.
Assess sits on top of everything upstream. Skip a pillar and you are not screening — you are performing a confident ritual over unverified material.
This leads to the foundational insight of this chapter, and it is worth stating as plainly as possible:
You cannot assess what you have not verified.
And in today's market that rule has a cruel corollary. The fake file is often cleaner, faster and more convincing than the real one.
Fake pay stubs are generated in minutes and frequently look better than genuine payroll records, which are messy, inconsistent and full of odd deductions. Fake references answer promptly and speak warmly, because that is their entire job. Synthetic identities blend real and fabricated data specifically to pass superficial checks. And fraud rings are excellent storytellers, because they have told the story many times and refined it.
Meanwhile the legitimate newcomer has documents in another language. The gig worker has irregular deposits. The recently separated applicant has a gap and a complicated explanation. Real life is untidy.
In an informal process, a polished fraud application will outperform a legitimate newcomer, student, or variable-income renter — because the fake file is optimized for appearance, while the real file reflects real life.
Read that twice if you are someone who trusts your instincts. Your instincts are being farmed. Every element that impresses you in an application is the element a professional has learned to manufacture, and the honest applicants are the ones who cannot polish their way past you.
Five Bands, Not Pass or Fail
Most screening treats the job as binary: approved or rejected, green or red. Real files are not shaped like that. A strong file and a dangerous file should never be handled the same way as a thin file, a newcomer file, or a borderline file where a stabilizer is available.
So the output of assessment is not a verdict. It is a diagnosis — and, like any diagnosis worth the name, it comes with a prescription.
Band 1 — Strong Match
ProceedThe file is verified, the published criteria are met, and no material red flag requires delay.
- Identity and key evidence confirmed
- Capacity, payment history and fit align with published criteria
- Move forward to lease
Band 2 — Match With Conditions
Proceed carefullyGenerally sound, but one or more neutral conditions should be completed or documented before move-in.
- For example: a guarantor, proof of reserves, insurance confirmation, or a document corrected at source
- Conditions must be neutral, published in advance, and applied consistently
- This is not a hidden rejection. It is a managed path
Band 3 — Review Required
Hold & verifyThe file may become acceptable, but material questions remain unanswered.
- Verification is incomplete or contradictory
- Context matters and more evidence is needed
- The decision is paused, not guessed
Band 4 — Not Yet Ready
Build the fileThe applicant may become a future match, but the current file is too thin or too unstable to proceed now.
- Common with newcomers, students, and people in transition
- Offer a documented pathway: stronger records, reserves, verified history, rent reporting, or time
- This creates a trust-building road instead of an unexplained dead end
Band 5 — Do Not Proceed
StopThe file does not support trust. The risk is too high, the facts do not line up, or fraud indicators remain unresolved.
- Identity mismatch, fabricated references, falsified income, non-cooperation, unresolvable contradictions
- The reason must still be documented by criterion and evidence
- Even a "no" requires a real process
Band 4 is the one I would ask you to look at hardest, because it is where the Toronto nurse from Chapter 1 was lost nine times.
Nine landlords had only two boxes. Her file did not fit the "yes" box, so it went in the "no" box, and nine times a landlord who wanted a reliable tenant threw one away. With five bands she lands in Band 4 with a documented pathway — show me twenty-four months of verified payment history from any country, through a source I can confirm — and she clears it in about twenty minutes, because she had the evidence the whole time.
The band is not the end of the process. It is the beginning of the next correct action.
The Rubric: Five Criteria, Published in Advance
A band needs something to come from. That something is a written rubric — published before you look at a single file, versioned, and applied in the same order to everyone.
Here is a working example. The weights are illustrative and yours may differ; the discipline is not optional.
Income against rent obligations. The ratio is a signal, not the whole story — a 34 percent ratio with twelve months of reserves is a different situation than 34 percent with none.
Payment behaviour, reference quality, stability — all confirmed at source. Note the word verified: unverified history scores nothing, in either direction.
How complete, coherent and verifiable the file actually is. Contradictions matter more than gaps. A gap has an explanation; a contradiction has a reason.
Guarantor, reserves, rent insurance, strong employment verification, portable history. This is the criterion that lets you say yes safely to files you would otherwise refuse.
Neutral realities tied to the unit, the building and the lease terms. Occupancy standards, parking, lease duration. Never a proxy for who you would prefer living there.
One rule governs whether something belongs on this list at all, and it is the sharpest test in the chapter:
If a criterion cannot be described in neutral language and applied consistently to every applicant, it must not decide the outcome.
Apply that test honestly to your current process. "Stable employment" passes — you can define it and measure it. "Someone who'll take care of the place" fails, because there is no neutral definition and in practice it collapses into whether the applicant reminds you of yourself.
The Decision Note
Every assessment ends by writing one short note. This is the single highest-value habit in the chapter, and it takes ninety seconds.
The note is written as a process explanation, not a personality judgement. Compare these two records of the same decision:
Here is the full shape of a real note:
Notice what this note does for you eleven months later, when something has gone wrong and you are asked to account for your decision. It shows the criteria existed before the applicant did. It shows what was verified and how. It shows the condition was applied to everyone, not invented for one person.
It also protects you from yourself. Writing the note forces you to name the actual reason — and if you cannot write a neutral sentence explaining a decision, that is the system telling you the decision was not made on neutral grounds.
Thin File Does Not Mean Bad Tenant
Modern applicants include newcomers, students, gig workers, remote workers, contract professionals, separated households and people rebuilding after disruption. Many are excellent tenants with limited local history.
The old process has one response to all of them: reject. That response is expensive, and increasingly it is also a legal exposure.
Auto-reject
Assume instability
Ignore or dismiss
Approve on narrative
Black-box rejection
The rule underneath all five rows is the one I would put on the wall:
Every risk signal must become one of four things: a verification step, a published condition, a diagnosis band, or a documented stop. If it becomes a gut feeling, the system has failed.
Stabilizers, not punishment
A guarantor, proof of reserves, employer verification, insurance confirmation or documented rent history all work as stabilizers — they convert uncertainty into a safer, explainable path.
Two cautions, and both matter. Stabilizers must be published in advance and available to everyone, or they become a quiet way to impose extra burdens on particular applicants — which is precisely the pattern human rights tribunals look for. And insurance is not a substitute for verification: it does not fix a fake identity, a fabricated document, or a broken intake process.
Real Case: The Rejection That Became a Complaint
He was right about the applicant. He still lost.
A landlord declined an applicant who was a single mother of two receiving a disability benefit as part of her household income. He would tell you, and I believe him, that his concern was the rent-to-income ratio.
The problem was everything around that decision.
He had no written criteria, so there was no published ratio to point at. He had approved a previous tenant at a similar ratio eight months earlier. He had asked this applicant for a co-signer and had not asked the earlier one. And when he declined, he told her the unit was "probably not suitable for young children," which he meant as a comment about the stairs.
At the hearing he could not produce a standard, could not show consistency, and had made a statement touching family status. His actual reason — the ratio — was never recorded anywhere, so it existed only as his later account of his own thinking.
He was not, in my reading, a discriminatory man. He was an undocumented one. And in a human rights process, those two things are very difficult to tell apart, because the tribunal is looking at pattern and record, not at intent.
This is why I keep insisting that the rubric is written before the applicants arrive. A criterion set in advance is a business standard. The same criterion produced afterwards to explain a decision is an argument — and it will be treated as one.
"Consistency is both the legal protection and the moral foundation."
Jimmy Ng · IDEAL FrameworkThe Review Path
A fair process includes a way to be wrong. Applicants sometimes fail on missing evidence rather than on facts, and a short review window costs you almost nothing while recovering good tenants you would otherwise lose.
Explain the band or the barrier in neutral process language. "Verified payment history covers four months against a published standard of twenty-four" — not "your application was unsuccessful."
A defined, time-limited opportunity to submit missing or corrected evidence. Five business days is usually enough.
New proof goes into the same file, dated and source-linked, exactly as in Chapter 5.
The same rubric, applied again. No moving goalposts. If they clear the stated bar, the band changes.
The updated outcome is documented and retained alongside the original.
Step four is where most review paths quietly fail. If you invent a new standard during review, the process becomes arbitrary again — and worse than arbitrary, because now there is a record of you shifting the requirement after seeing who the applicant was.
Technology at the Kitchen Table
AI screening assistants — what they can and cannot be trusted to do
Software that promises to score applicants using artificial intelligence is arriving quickly in this market. Some of it is genuinely useful. Some of it will get landlords into serious trouble. The difference is easy to state.
What is it, in one sentence? Software that reads application material and produces a summary, a flag, or a score to help you decide faster.
Why does it exist? Because of the Sunday afternoon in Section 7.1. Fifty files genuinely is too many for careful human review, and the market responded with automation.
What can it safely do? Three things, all of them valuable. It can summarize a long file into a structured overview so you read the important parts. It can flag contradictions — an address that does not match between two documents, a date that cannot be right, an employer name spelled two ways. And it can check completeness, telling you what is missing before you waste time.
What must it never do? Make the decision, or produce a score you cannot explain. If a tool tells you an applicant is "high risk" and you cannot say which neutral criterion produced that and what evidence sits behind it, you have not automated your judgement — you have outsourced your liability. "The software said no" is not a defence at a human rights tribunal. You made the decision; you own the reasoning.
What does it cost? Increasingly bundled into screening products, often a few dollars per application. The real cost is the temptation to stop thinking.
The test to apply before buying: ask the vendor to show you the decision note their system produces. If it points at named criteria and specific evidence, it is helping. If it produces a number with no traceable path back to a fact, walk away — that is a black box, and you will be the one standing behind it.
Doing This Without a Budget
The Low-Cost Path
- One page of written criteria, dated before you advertise
- The same five headings scored the same way, in the same order
- A two-line decision note per applicant
- Five bands instead of yes/no
- A stated five-day review window
- Notes kept with the tenancy file from Chapter 5
- Cost: nothing but discipline
The Scaled Path
- Versioned rubric applied identically by every staff member
- Structured file summaries with verification flags before scoring
- Decision notes generated and stored automatically
- Band outcomes tracked so you can audit consistency over time
- Periodic review: are certain groups clustering in Bands 4 and 5?
- Review path built into the workflow, not handled ad hoc
That fifth item on the right deserves a sentence of its own. Reviewing your own band distribution is the cheapest fairness audit available. If newcomers consistently land in Band 4 while comparable local applicants reach Band 2, your criteria are doing something you did not intend — and you would rather discover that yourself than have it discovered for you.
Chapter Seven Checklist
- Publish your criteria before reviewing any file. Written, dated, versioned. This single act does more legal protection than everything else in the chapter combined.
- Collect only what the decision needs. Privacy discipline is part of assessment, not separate from it.
- Verify before you score. A claim is worth zero until it is confirmed at source — favourable claims included.
- Apply the same rubric in the same order to every applicant. No shortcuts for the one you like, no extra hurdles for the one you don't.
- Replace yes/no with five bands. Most of your lost good tenants are sitting in what used to be a single "no" pile.
- Run the neutrality test on every criterion. If it cannot be stated neutrally and applied to everyone, remove it.
- Write a decision note for every applicant, including the ones you approve. Ninety seconds. Band, criteria, evidence, next step.
- Apply the translation rule to every worry: verification step, published condition, band, or documented stop. Never a feeling.
- Offer stabilizers rather than rejection where uncertainty is the problem — and offer them to everyone equally.
- Provide a five-day review window with the same rubric on re-assessment. No moving goalposts.
- Audit your band distribution every six months. Look for patterns you did not intend.
- Never accept a score you cannot explain. If a tool cannot show which criterion and which evidence produced a flag, do not act on it.
A hospital does not triage a patient by vibes. It records identity, symptoms, observations, history, and changes over time. Housing needs the same discipline.
Landlords resist this pillar more than any other, and I understand exactly why. It feels like being told your experience is worthless — that twenty-five years of reading people counts for nothing next to a spreadsheet.
That is not what I am saying. I am saying something narrower and, I think, harder to argue with.
Your judgement is real. It is also unverifiable, unrepeatable and undefendable — and in a market where fraud is professional and human rights complaints are decided on pattern rather than intent, those three weaknesses now cost more than the judgement is worth.
There is also something I did not expect when I started this work. Landlords who adopt a written rubric describe the same relief, in almost the same words. The Sunday afternoon stops being exhausting. The decision stops being personal. You are no longer weighing human beings against each other in your kitchen — you are checking files against a standard you set calmly, in advance, when nobody's housing was in front of you.
That is a better job. It is also, not coincidentally, a fairer one.
And notice what the five bands have really done. The old process asked one question — is this person good enough? The new one asks a better question: what would have to be true for this to work? Sometimes the answer is nothing, and you proceed. Sometimes it is a stabilizer. Sometimes it is more verification. Sometimes it is honestly not yet.
But you will stop throwing away nurses.
The match is diagnosed. Now it has to be made durable — because the best screening decision in the country is still just a prediction until the lease turns it into a relationship that keeps working.