Key takeaways
- Likely-to-sell models rank homes by signals such as ownership tenure, equity and life events; they estimate odds, not intentions.
- Many decisions to move start with a life event: 77% of sellers cite one, most often a change in household size[1].
- Even a good model flags far more homes than will list, so judge it by lift over a simple baseline on your own farm.
- Outreach to model-flagged owners is cold outreach: Do Not Call registrations must be honored indefinitely[2].
What likely-to-sell models do
Predictive seller analytics, often sold as likely-to-sell scores, rank homeowners in an area by how probable it is that they will list within a set period, often six or twelve months. The vendor combines public records, such as deeds, mortgages and permits, with commercial data and the history of homes that did sell, and trains a model to spot patterns that came before past listings. Each home gets a score, and agents focus their prospecting on the top slice.
The idea is sound: sellers do not appear from nowhere, and some signals really do precede a sale. The practical value depends on three things: how much better the model is than a simple rule, how fresh its data is, and what you do with the list. This guide covers each.
The signals most models use
| Signal | Why it can predict a sale | Caveat |
|---|---|---|
| Length of ownership | Owners move more often after certain tenure milestones | National tenure is at a record high, so old rules of thumb may mislead |
| Estimated equity | High equity makes a move financially easier | Estimates can be far from reality for updated or unusual homes |
| Mortgage rate | A low locked rate discourages moving | Life events override rate lock-in |
| Life-event records | Probate, divorce filings and similar events often lead to sales | Sensitive; outreach needs great care |
| Permits and improvements | Pre-sale repairs and updates can signal a listing | Many owners renovate to stay |
| Absentee or investor ownership | Investors sell on financial triggers | Data on ownership type is patchy |
| Past listing or rental activity | Withdrawn or expired listings and rental listings show intent | Often already known to every agent in the area |
Avoid models that use or infer characteristics such as age, family status, religion, national origin or disability, directly or through close proxies. A score that ranks owners by demographic traits can lead to unequal treatment in who receives outreach, and the Fair Housing Act prohibits discrimination based on race, color, national origin, religion, sex, familial status and disability[3]. Ask any vendor for the list of inputs.
Tenure, rates and why owners stay put
Ownership length is the backbone of most models, and it has shifted. NAR's 2026 Profile found sellers had owned their homes for a median of 12 years before selling, an all-time high[4]. Rates are part of the reason: by early 2026 buyers had faced nearly 3.5 years of mortgage rates above 6%[5], and owners with much lower locked rates have a financial reason to wait.
For a model, that means tenure thresholds learned from the 2010s may now flag too many homes too early. Ask vendors whether their model was retrained on recent sales and how it treats rate lock-in.
Life events drive many decisions
Zillow's 2026 seller research found 77% of sellers cite a life event as influencing their move, most often a change in household size (47%), a new job (30%) or remote work[1]. Few of those events show up in public records in time to be useful. A new baby, a job offer in another city or a parent moving in are visible to friends and past clients long before they are visible to a data vendor.
Decisions also take time. The median seller thought seriously about selling for three to four months before listing, and 27% thought about it for at least six months[6]. Sixty-six percent at least thought about renting out their home instead of selling, and 24% seriously considered it[7]. A model can help you find owners in that thinking period, but the conversation that converts them is about their situation, not about a score.
Where predictive models fall short
- Low base rates: in any year most homes do not sell, so even the top-scored slice is mostly owners who will stay.
- Stale data: public records lag, and a model refreshed quarterly may flag homes already listed or recently sold.
- Shared lists: if every agent in your market buys the same scores, the top homes receive the same letters and calls from everyone.
- Missing signals: the life events that matter most are often not in the data at all.
- Black boxes: vendors may not explain why a home scored high, which makes it hard to have a relevant conversation or to check for bias.
- Privacy discomfort: outreach that reveals what you know, such as a recent divorce filing, damages trust.
Read the numbers like a skeptic
Example (fictional): your farm has 2,000 homes and, in a typical year, about 80 list, a base rate of 4%. A vendor says its top 10% of scores captures three times the base rate. That means the top 200 homes would contain about 24 listings in a year: well worth focusing on, but 176 of those 200 owners will not list. If you treat the list as a list of sellers, you will waste most of your effort and annoy most of the recipients.
Example, continued: a simple rule, such as owners with more than 15 years of tenure, might capture twice the base rate on its own. The model's real value is the gap between three times and two times, and that is what you are paying for. Ask for that comparison on your own area.
How to test a vendor
Ask for a backtest on your farm
Have the vendor score your area as of a date twelve or more months ago, then check how many top-scored homes actually listed since.
Compare with a simple baseline
Measure how the top slice performed against a tenure-only or equity-only rule. The difference is the model's lift.
Check the inputs
Get the list of signals and confirm none is a protected characteristic or a close proxy.
Check freshness
Ask how often records and scores update and how listed or sold homes are removed.
Run a holdout
For six months, market to half of the top-scored homes and hold back the other half, then compare listing appointments.
Turning a score into a useful contact
A score tells you whom to contact, not what to say. Never mention the score or the data behind it; owners do not want to learn that a vendor has rated their likelihood of moving. Lead with something useful about their home or street instead, and let them tell you their plans.
- Mail or a door hanger with a short note on recent sales on their street and an offer of a free price range, with a way to reply.
- A neighborhood market update by email for owners who have opted in, with a link to book a valuation call.
- For past clients and sphere contacts with high scores, a personal call or note: checking in, any changes at home this year?
- An invitation to a nearby open house, which gives curious owners a low-pressure way to meet you.
Track what each approach produces in your CRM: replies, valuation requests and listing appointments, by score band. After six months you will know whether the top band truly responds better than the rest, which is the only test of a model that matters to your business.
Outreach rules still apply
Model-flagged owners did not ask to hear from you, so contacting them is cold outreach. Registrations on the National Do Not Call Registry must be honored indefinitely or until cancelled[2], and there were 259 million active registrations as of September 30, 2025[8]. Robocalls that are sales calls are illegal without written permission whether or not the number is registered[9], and telephone solicitations are barred before 8 a.m. or after 9 p.m. at the recipient's location[10]. Mail, door-knocking where permitted and useful local content are often better fits for scored lists.
Pair scores with relationships
The strongest seller signals are often already in your CRM. Zillow found 81% of seller-buyers who used an agent hired the same agent for both sides of their move[11], and 30% of sellers found their agent by referral[12]. Use predictive scores to decide which past clients and sphere contacts to call first this quarter, and to choose which neighborhoods to farm, rather than as a list of strangers to cold call.
That combination plays to each source's strength: the model notices patterns across thousands of homes, and you know which of your past clients just mentioned a new job.
Frequently asked questions
Do likely-to-sell predictions work?
They can concentrate your effort on homes somewhat more likely to list, but most flagged owners still will not sell. Judge a model by its lift over a simple rule on your own area.
What data do predictive seller models use?
Usually ownership tenure, estimated equity, mortgage details, permits, past listing or rental activity and some life-event records, combined with the history of homes that sold.
Can I call homeowners on a likely-to-sell list?
Only within telemarketing rules: scrub against the Do Not Call Registry, respect calling hours and never use automated or prerecorded calls without written consent.
Are predictive seller scores a fair housing risk?
They can be if inputs include or closely track protected characteristics. Ask vendors for their inputs and avoid any that use age, family status or similar data.
References
Zillow Research (2026). Sellers: Results from the Zillow Consumer Housing Trends Report 2026 (opens in a new tab) ↑
Legal Information Institute, Cornell Law School (eCFR text) (2026). 47 CFR § 64.1200 - Delivery restrictions (TCPA rules) (opens in a new tab) ↑
U.S. Department of Housing and Urban Development (2026). Housing Discrimination Under the Fair Housing Act (opens in a new tab) ↑
National Association of REALTORS® (2026). Top 10 Takeaways from NAR's 2026 Profile of Home Buyers and Sellers (opens in a new tab) ↑
Realtor.com Economic Research (2026). The Best Time To List a Home: The Week of April 12-18 (2026 analysis) (opens in a new tab) ↑
Zillow Research (2026). Sellers: Results from the Zillow Consumer Housing Trends Report 2026 (opens in a new tab) ↑
Zillow Research (2026). Sellers: Results from the Zillow Consumer Housing Trends Report 2026 (opens in a new tab) ↑
Federal Trade Commission (2025). National Do Not Call Registry Data Book, Fiscal Year 2025 (opens in a new tab) ↑
Federal Trade Commission (2025). National Do Not Call Registry Data Book, Fiscal Year 2025 (opens in a new tab) ↑
Legal Information Institute, Cornell Law School (eCFR text) (2026). 47 CFR § 64.1200 - Delivery restrictions (TCPA rules) (opens in a new tab) ↑
Zillow Research (2026). Sellers: Results from the Zillow Consumer Housing Trends Report 2026 (opens in a new tab) ↑
Zillow Research (2026). Sellers: Results from the Zillow Consumer Housing Trends Report 2026 (opens in a new tab) ↑