43% of buyers found their agent by referral — so the model treats the past-client list as the primary channel

The published buyer data says two things at once. 52% of buyers found the home they purchased on the internet, and 43% found their agent by referral. The property is a search problem; the agent is a relationship problem. 88% of buyers used an agent at all, which means the competition is other agents rather than disintermediation. And the first-time buyer share has fallen to 21%, the lowest recorded since 1981, so a larger proportion of transactions now come from people who have bought before and therefore have an agent, or had one. Against that, real-estate websites convert at 2.8%, the lowest of thirteen measured industries.

Half of buyers find the house online. Not quite half find the agent through somebody they know. Most real-estate budgets are aimed at the first number and judged against the second.

1. 01

   Referral is a channel with a cadence, not a hope

   It has a list, a trigger, a frequency, a message and a measurable response, which is the definition of a channel. The model gives it an owner, a contact schedule and a line in the monthly report beside paid and organic, rather than treating it as the thing that happens when the other channels are working.

2. 02

   The lowest first-time buyer share since 1981 changes who the list should contain

   A shrinking first-time cohort means a growing share of transactions come from repeat buyers, who are by definition people with a prior agent relationship. That makes the past-client database more valuable each year the share falls, and makes prospecting to first-time buyers a smaller pool than it has been in four decades.

3. 03

   88% used an agent, so the threat is the agent next door

   The published data does not support a disintermediation story. It supports a competition story. That distinction matters because the two imply opposite investments: one says defend against a platform, the other says be the name that gets mentioned.

4. 04

   A 2.8% website conversion rate makes the site a poor place to ask and a good place to prove

   At the lowest conversion rate of thirteen measured industries, the site is not where the relationship gets started. The model moves the ask into the referral motion, where the request is made by a person who already transacted, and leaves the site to carry the proof that makes the referral easy to accept.

**43% of buyers found their agent by referral** (49% of first-timers) · **88% purchased through an agent** · ~46% started their search online, ~half found the home online.

The model builds the past-client and sphere database as an owned asset first: every prior transaction in the CRM, tagged by close date, property type and neighbourhood, with a named owner for each record. On top of it runs a fixed cadence rather than an occasional one — a review request at close, an anniversary contact, and listing alerts scoped to the neighbourhoods that record already cares about, which is the piece that gives the list a reason to open anything. Paid media supports the motion rather than leading it, aimed at property discovery where the published data says buyers are actually searching. Reporting is on referrals received and on transactions attributed to source, which requires asking every new client how they arrived and recording the answer.

1. 1

   Weeks 1–3

   Database audit

   Every prior transaction into the CRM, tagged and assigned an owner

   Owner: OmniFlow + brokerage

2. 2

   Weeks 3–5

   Cadence design

   Close, anniversary and listing-alert contacts scheduled as a standing calendar

   Owner: OmniFlow

3. 3

   Week 5 onward

   Review and referral asks

   Request at close and at anniversary, with the wording fixed rather than improvised

   Owner: Agents + OmniFlow

4. 4

   Week 5 onward

   Source capture

   Every new client asked how they arrived, and the answer recorded in the record

   Owner: Brokerage

5. 5

   Month 3 onward

   Media in support

   Paid aimed at property discovery, measured separately from the referral motion

   Owner: OmniFlow

6. 6

   Month 12

   Review

   Referrals received and transactions by source, reported against the cadence

   Owner: OmniFlow + brokerage

The model targets referrals received per quarter and transactions attributed to source, and would expect the referral line to be slow, compounding and largely unaffected by short-term budget changes, which is exactly why it is usually the first thing cut and the last thing measured. Attribution here depends on asking every new client how they arrived, so the model treats source capture as a delivery task with an owner rather than as a reporting afterthought. The published shares quoted here frame the first quarter only, and are replaced by the brokerage's own source mix as soon as enough records carry an answer to read one.

Modelled. Inputs: National Association of Realtors, 2025 Profile of Home Buyers and Sellers; Ruler Analytics, Conversion Rate Benchmarks 2026 (110M+ sessions, 5M+ conversions); WordStream by LocaliQ, Google Ads Benchmarks 2026 (13,474 US search campaigns, April 2025 – March 2026). Modelled outputs are not a forecast or a guarantee of results. Cited: NAR *2025 Profile of Home Buyers and Sellers*.

Every figure below is published research, not a client result. They are the inputs to the arithmetic above, listed so it can be checked rather than taken on trust. The bracketed number points to the full citation at the end of this page.

These are the surfaces this engagement is run and measured from, shown with representative figures built around the benchmarks cited on this page. Every account we run reports into views like these, and you keep ownership of all of them.

Each figure on this page, the system it is read from, and the definition and window it is measured over.

Not applicable — this is a modelled engagement. Its weakest input is survey scope: the published profile surveys buyers who completed a purchase, so it says nothing about the people who searched, engaged an agent and did not transact, and the referral share among that group is unknown. The model also assumes referral behaviour is reasonably stable across market cycles, which a single year of survey data cannot establish, and the database audit exists to replace the published shares with the brokerage's own source mix as soon as source capture produces enough records to read.

A modelled capability, not a client account. Figures illustrate what the model produces and are labelled as modelled wherever they appear. See all our work

Full citations for everything cited on this page, with the sample and period each study covers, so you can go and read the original.