Flat fee models are insulated at the transaction level and more exposed at the roster level, which is the trade windlass named. Work the arithmetic on both sides.
Split shop: 100 agents, 4 deals a year each, 400 sides. At a 9,000 average commission and 30 percent, that's 1.08m. Drop the average to 7,200 and you're at 864k, down 216k, with the roster intact.
Fee shop: same 100 agents at 500 a month plus 250 a side. That's 600k in desk fees and 100k in transaction fees, 700k total. Commission compression takes nothing off the desk line. But a 500 dollar monthly fee against a shrinking commission check is a much sharper decision for a two-deal-a-year agent, and those agents are the majority of most rosters. Lose 20 percent of the roster, weighted toward low producers, and you lose 120k of desk fee immediately plus their transaction fees, and you lose it in a month rather than over a year.
The part that decides it is your low producer concentration. If half your desk fee revenue comes from agents doing one or two deals a year, your revenue is a subscription business sold to people whose ability to pay just got worse. Cheap to run, fast to unravel. The split shop's revenue is concentrated in producers who are the last to leave.
What neither model handles well is the compliance load. Written buyer agreements and compensation disclosure mean file review per transaction went up, and on a fee model you cannot bill that back without raising the fee that's already the reason people leave. Price your supervision cost per side before you pick a structure.