Real Estate Brokerage Commission Splits: What to Offer

Real estate brokerage commission splits are published by nearly every brokerage that will recruit against you, and they cluster in a narrow band: 80/20 to a $16,000 annual cap, 85/15 to a $12,000 cap, graduated schedules starting near 50/50 and improving with production. You can look all of that up in an afternoon.

The number on your side of the same trade has never been published by anyone. What it costs your brokerage to carry one agent for a year does not exist as a figure. Not from NAR, not from a regulator, not from an analyst, not from a single brokerage. We went looking twice, in two separate research passes, and confirmed the same absence both times.

So the split is the largest financial decision a new brokerage makes, and it is the only one you will make with no cost basis. Here is what is actually measured, who paid to measure it, and what you are left to decide alone.

We build websites for brokerages. We have no stake in which split you pick, and no recruiting model to sell you.

What the big brokerages publish, and who is publishing it

Two brokerages publish a company-wide number you could hold them to. The rest publish nothing, or publish a formula whose real cost is set somewhere else.

Read the right-hand column before the middle one. Every number in this table with real specificity was published by a party whose business is signing agents, and the two most valuable brokerages on the list publish no split at all. That tells you what a split is treated as in this industry: a recruiting instrument, disclosed when it wins and withheld when it does not.

What agents actually have, last measured in 2022

The only distribution anybody publishes counts agents, not brokerages, and the most recent breakdown we could source covers 2022.

NAR's Member Profile found REALTORS reporting a fixed split under 100 percent (42 percent of them), a graduated split (19 percent), a capped split (15 percent), 100 percent commission (14 percent), and salary only (2 percent). The prior year's edition read 36, 20, 18 and 15 across the same first four categories.

Three things about that. It is self-reported by dues-paying members, which is not the same population as all licensees. It measures what an agent says they have, not what a brokerage offers, and one brokerage can run three of these at once. And a six-point swing in the fixed category between adjacent years, published without a confidence interval, means you should read the ordering and ignore the decimals.

It also predates August 2024 entirely, which matters for the section below.

What does not exist, at all: any count of US brokerages by model. What share run a graduated schedule versus a cap versus a flat monthly charge is unmeasured. We have now asked for it in two research passes.

Company dollar, or what is left after you pay everyone

The brokerage's retained share of commission revenue roughly halved in a decade, from 22.4 percent in 2012 to 11.0 percent in 2021.

That series comes from RealTrends brokerage benchmarking, and it is the only multi-year measurement of the brokerage side that exists. Label it accordingly: RealTrends sits inside HW Media, which sells marketing and recruiting products to brokerages, so it has a commercial interest in brokerages reading brokerage content. It also does not disclose its sample size, its response rate, or whether participating brokerages self-select, which would tilt the sample toward larger and more sophisticated firms. It is the best number available and it is not an audited one.

The second finding in the same data is the one nobody quotes. Over that decade, profit measured against gross commission revenue barely moved, from 4.5 percent to 3.7 percent, while profit measured against company dollar rose from 20.1 percent to 33.3 percent. Brokerages lost half their gross margin and stayed roughly as profitable by cutting everything else. You will open with nothing to cut.

The line you will hear at every brokerage event, that company dollar fell from about 30 percent to about 15 percent, traces to RealTrends' own 2018 post. That post put 2017 at 14.8 percent and described roughly double that "a couple decades" earlier without naming a year, a sample, or a source. The 14.8 is documented. The 30 is an aside that got repeated until it sounded like data.

One more from the same body of work: teams retained an average gross margin of 61.8 percent against 13.8 percent for brokerage companies. The team model captures the economics the brokerage used to.

The number nobody publishes: what an agent costs you

There is no published figure, from any source, for what it costs a brokerage to carry one agent for a year. The pieces are priceable. The total has never been assembled by anyone with no stake in the answer.

• Errors and omissions coverage. Insurance marketplaces put the average around $59 to $68 a month, roughly $708 to $815 a year. Both sources sell the coverage, and both quote the policyholder's cost, which is yours or the agent's depending entirely on how you set it up. Practitioners report everything from zero to $1,600 a year, mostly because brokerages absorb it at wildly different rates. Nobody measures which.
• Transaction management software. Agent-facing seats run about $25 to $60 a month. Brokerage suites start near $340 a month and scale with headcount. All of that pricing comes from the vendors' own pages.
• Association and MLS dues. Roughly $500 to $1,000 a year per agent for national, state, local and MLS access combined, described almost everywhere as the agent's cost. No source quantifies how often a brokerage absorbs it as a recruiting concession, or what that adds up to.
• Your supervision and compliance time. No dollar figure exists anywhere. Not an estimate, not a range, not a survey. It is the cost that scales fastest with headcount and it has never been priced.

You will be told that agents are nearly free because they are independent contractors who pay their own way. Nobody has tested that claim. And the cost most likely to break a new brokerage, your own attention, is the one that has never appeared in a spreadsheet anybody published. That is the same arithmetic behind recruiting for production rather than headcount, and it belongs in your startup budget before it belongs in your split.

What changed in 2024, and what it did to the pot

The August 17, 2024 practice changes altered how buyer-side compensation is offered and agreed. The measured effect on the rate itself has been small, and largely reversed.

Read the direction, not the decimals. Rates dipped after the changes and recovered to at or above where they started by mid-2025. Every specific number available comes from a company whose business model depends on the answer, and the closest thing to a disinterested source cites other people's estimates for the part that matters. There is no large-sample, record-based measurement of the before and after by anyone without a position.

For your purposes: a split is a percentage of a percentage, and the first percentage is now negotiated in writing with a buyer before a tour rather than advertised through the MLS. Agent gross got less predictable. A cap did not.

Three models, and what each does to your cash

Nobody has measured which model performs better for a brokerage. What follows is arithmetic and cash-flow shape, labelled as reasoning rather than evidence, because no evidence exists.

Graduated. You keep most of the early production, and your best agents pay the most. Which is exactly why your best agents leave for a cap.

Cap. Predictable for both sides and front-loaded for you, since the agent owes the same total whether they close it in March or in December. Do the arithmetic before you copy the number: a $16,000 cap is 16 percent of a $100,000 producer's gross commission and 1.6 percent of a $1,000,000 producer's. A cap is a volume discount for your largest agents, funded by everyone below them. It is also worth knowing how rarely it binds. eXp's own income disclosure reports a median of $13,463 paid to active agents in 2025 across 63,410 of them, which means the overwhelming majority never come near capping. For most of your roster, the split simply keeps coming out all year.

Flat monthly, or 100 percent. Published examples run from $98 a month with nothing per closing, to $39 a month plus $495 a closing, to $125 a month plus $125 a closing. This converts you into a headcount business: your revenue stops tracking the market and starts tracking your roster. There is no upside in a hot year and no cushion in a cold one, because your costs are per agent and now your revenue is too. Whether brokerages on this model survive at better or worse rates than the others has never been measured. The only skeptical analysis we found was uncited commentary, and everything supportive was written by brokerages recruiting into it.

What the law lets you pay, and how

In every state we checked, compensation for licensed work flows through the broker. The live question is whether you may pay an agent's company instead of the agent.

One question we could not answer in any of the four: whether and how you may pay a commission to an agent who has already left. No statute or commission rule we found addresses it. If your first agent departs with a deal in escrow, that is a question for your commission and your attorney before it happens rather than after. Advertising rules for the site itself are a separate matter, and we cover fourteen states individually, starting with California and Texas.

Two more pieces of law, both stable:

Independent contractor status has not moved. The federal safe harbor at 26 U.S.C. 3508 still treats a licensed agent as a non-employee for federal tax purposes when three things hold: a valid license, substantially all pay tied to output rather than hours, and a written agreement stating they will not be treated as an employee. NAR's federal policy tracking reported no imminent change as of April 2025 and the statute is unchanged. California exempts licensed agents from the AB5 test and routes classification through Business and Professions Code 10032(b) or the older Borello factors instead. Note what the second condition does to your split: pay tied to hours rather than output is the thing that breaks the safe harbor.

Be careful what you take from settlement coverage on antitrust. What the 2024 settlement and the DOJ's February 2024 statement of interest actually address is buyer-side compensation, whether it can be advertised through the MLS, and how buyer agreements are written. We found no primary source and no enforcement action addressing whether brokerage owners may compare agent splits with each other. That is not a finding that it is fine. It means the question is open, and it is worth an hour of your attorney's time rather than a confident answer from a blog.

Does a better split recruit better agents?

No disinterested source has ever shown that it does.

This is the third time this cluster has gone looking for evidence that a specific brokerage input drives recruiting, and the third time it has come back empty. Every claim linking a generous split to recruiting or retention outcomes came from a party recruiting into a particular split. No peer-reviewed work, no record-based analysis, no survey from anyone without a model to sell.

The nearest thing to evidence points the other way. The only study in this area with fully disclosed methodology, ten interviews with experienced agents who had recently changed brokerage, found money and technology treated as table stakes while leadership access and belonging decided it. Ten interviews cannot carry statistical weight, and we said so when we wrote about recruiting. It is still more than anyone selling a split has produced.

Set the number you can still afford at thirty agents

The split you give your first three agents is the split every agent after them will hear about, and the one you will have to justify when you try to change it.

Set it against your own costs rather than against eXp's cap. eXp is running an 85,000-agent business on a revenue model you do not have, and a cap that works at that scale is a number, not a strategy. Work out what you will absorb per agent, coverage, software, dues and your own hours, decide how many agents that supports before you need help, then set the split so the answer is still true at thirty. If the arithmetic only works when everybody produces, it does not work. Production concentrates hard, and the agents most available to a brand new brokerage are the least likely to produce.

Then publish it. Two of the largest brokerages in the country publish no split at all, which makes plain terms a thing you can beat them on for free.

When you have the number, it needs somewhere a recruit can actually read it. That part is ours: live before you open, priced in the open.