Reenter / Reenter AI

AI that explains risk, not just scores it

Reenter uses AI to verify applicants, set where each tenant’s premium starts, suggest a realistic budget and area, and match tenants with homes. Landlords take less risk and good tenants pay less. Every score comes with its reasons.

A woman checking an application on her phone
Score 86 · low risk
Reenter Cover from 6.5%: ₺2,730 a month

How it works

Four layers, each one explainable

Verify

  • ID through official checks
  • Income from payroll or bank data, with consent
  • Title deed for landlords
  • Listing through EİDS
Output: verified profile

Score

  • Rent-to-income ratio
  • Income stability and tenure
  • Rental payment history
  • Credit score band, with consent
Output: score 0–100, its reasons and the starting premium step

Recommend

  • Rent budget from income and debts
  • Home size from household and work
  • Areas ranked by budget, commute and lifestyle
  • Fair rent range for landlords from comparable homes
Output: budget, home type, areas

Match

  • Homes ranked for each tenant
  • Applicants ranked for each landlord
  • Fair rent suggestions at renewal
Output: rankings both sides can read

Under the hood

From your data to a decision, in six steps

Click a step to see what goes in, what happens, what comes out and what protects you.

Step 1 / 6

You choose what to share

Goes in
ID through e-Devlet, income from payroll or open banking, your rent history on Reenter and, only if you agree, a credit score band (Findeks).
What happens
Each source is asked for separately, with what it’s used for. Documents are checked and then deleted; only the verified facts are kept.
Comes out
Verified facts, not raw documents
Safeguard
KVKK explicit consent. Health, religion, ethnicity, politics and criminal records are never collected.

Model cards

Six models, what each one does and doesn’t do

Not everything is machine learning. Where a published rule is fairer and clearer, we use the rule.

Tenant score

Sets where Reenter Cover starts and helps landlords compare applicants.

Inputs
Rent-to-income, income stability, job tenure, rent history, consented credit band
Method
Monotonic scorecard
Output
0–100, band, reasons
Human check
Landlord decides; review on request

Premium steps

Moves the premium between 4.5% and 12% as the record builds.

Inputs
Starting band, claims and late payments every six months
Method
Published rules, not machine learning
Output
Step 1–7
Human check
Insurer partner underwrites

Rent estimate

A fair asking range for landlords and a sensible renewal suggestion.

Inputs
m², rooms, district, building age, floor, amenities, minutes to transit, recent agreed rents
Method
Hedonic price model with comparables
Output
Range and 5 comparable homes
Human check
Landlord sets the price

Budget and area finder

Tells a tenant what they can comfortably pay and where.

Inputs
Income, debts, household, workplace, priorities
Method
Transparent rules and a 40/30/30 ranking
Output
Rent range, home size, areas
Human check
Suggestions only; runs in your browser

Search and matching

Understands plain sentences and ranks homes and applicants.

Inputs
Your words and filters, saved homes, published ranking weights
Method
Language parsing and weighted ranking
Output
Filters you can remove, ranked lists
Human check
You see and undo every filter

Listing safety

Stops fake listings and payment scams before they reach anyone.

Inputs
EİDS authorisation, title-deed match, reused photos, prices far below market, “pay outside the app” messages
Method
Rules and anomaly detection
Output
Verified badge or held for review
Human check
A person checks every flag before removal

How it learns

Careful, scheduled, and reported

Models that price risk can drift or pick up bias. These are the checkpoints before any version reaches a real tenant.

What each release note shows

Model version, in plain words

  • What changed and why
  • Calibration: predicted vs. actual late rent by band
  • Fairness: approval and pricing gaps on test groups
  • How many results people reviewed or changed
  1. NowRules first

    Before launch, scores come from published rules and the insurer partner’s tables. The demo on this page runs that simplified version.

  2. BetaShadow mode

    In the İstanbul beta the model scores in the background while people make every decision, so we can compare the two.

  3. LaunchCalibrated with the insurer

    Each band’s price is checked against what actually happened in the beta, then the bias test runs before go-live.

  4. Every quarterRetrain and report

    New version, compared with the old one on accuracy and fairness, with a short public change note.

Budget and area finder

How the budget and area model works

A free tool for tenants. It uses only what you type, runs in your browser and shows its reasoning.

  1. Budget

    Income after loan and card payments × 30% (comfortable) to 38% (stretch), a little lower with children. Reenter Cover, the service fee and a typical building fee are taken out first.

  2. Home size

    One bedroom for a couple or for each adult flatmate, one for every two children and a study if someone works from home, turned into a room count (e.g. 2+1) and an area.

  3. Area fit

    Each area gets 40% for affordability, 30% for the commute to where you work or study and 30% for the things you said matter, such as schools, transport or the sea.

  4. Data

    For now, illustrative area rents and typical commute times. After launch, agreed rents and listings on Reenter, updated monthly.

Interactive demo

See how a score is built

A simplified version of the model. Move the inputs and watch the factors.

0/ 100

Starting premium
After 12 claim-free months

Why this score

Demo only. The production model will be trained and validated with our insurer partner and audited for bias before launch.

Ranking weights

What each ranking looks at

Published so landlords and tenants know how they’re compared.

Property ranking

Location
40%
Condition
30%
Amenities
20%
Tenant feedback
10%

Tenant score

Financial stability
70%
Rental behaviour
20%
Profile completeness
10%

Landlord score

Tenant feedback
50%
Responsiveness
50%

Plain-language search

Describe the home, we build the filters

Reenter assistant

Answers with the law and your lease

My landlord wants a 45% increase. Is that allowed?
Your lease renews in November, so the legal cap is 31.49% (12-month average CPI). Above that isn’t enforceable. Want me to draft a polite reply with the numbers?
Yes please
Done. I’ve also added the renewal date and the maximum new rent to your calendar.

Guardrails

Fair by design, and checkable

Discrimination in renting is a crime in Türkiye (TCK 122), and automated decisions can be challenged under KVKK art. 11. We build for both.

What we do

  • Show the reasons behind every score
  • Ask explicit consent for income and credit checks
  • Let anyone ask for a human review
  • Test the model for bias before and after launch

What we never use

  • Gender, age, marital status or pregnancy
  • Nationality, ethnicity, religion or language
  • Health, disability or political views
  • Criminal records or social media

Questions

About Reenter AI

Will an algorithm decide whether I get a home?

No. The score sets where your Reenter Cover premium starts and helps landlords compare applicants. The landlord decides, and you can ask for a person to review any automated result (KVKK art. 11).

What data is used?

Only what you agree to share: verified income, job tenure, rental payment history on Reenter and, with explicit consent, a credit score band. Never protected characteristics or criminal records.

Can I improve my score?

Yes. Verify your income, add a previous landlord’s reference, or pay rent on time through Reenter. The app shows which factor would help most.