Does a listing agent's CMA carry any more weight than Zillow if both are wrong by the same 8 percent

The structural difference matters more than the dollar gap. A CMA is a human judgment call that can be interrogated, which you already proved by pulling DOM and price cut history on those two stale comps. An AVM like Zillow is a regression model trained on recorded sales and public data, and it cannot tell that one of its inputs sat 61 days with a cut because it weights the closed price, not the story behind it. So when both are wrong by a similar margin, they are wrong in different ways: the agent's number bends toward what she thinks will get the listing signed, and Zillow's number bends toward whatever the model learned from a neighborhood cohort that may not match your specific two-unit profile.

The one question worth pressing on is how Zillow classified the property. If it indexed your duplex against single-family sales rather than small multifamily, the $191k figure is comparing the wrong population entirely, and the agent's four comps, even the weak ones, are at least pointed at the right asset class. That is where the CMA earns its keep over an AVM, not in the agent's skill, but in her ability to filter by property type in a way the algorithm often does not.

What is the income picture on the occupied unit, and do you know the current market rent for the vacant one?

The CMA is more transparent, which is different from being more accurate. Zillow's model is a black box trained on transaction volume; it degrades fast in low-turnover markets because it has nothing recent to anchor to. The agent's CMA is auditable, which is exactly what you did when you pulled those addresses. The fact that you found a 61-day sit and a $9,000 price cut in two of four comps is the relevant finding here.

The assumption doing the most work in her $198,000 figure is "stable market equals clean comps." A stable market means prices are not moving sharply, but it does not mean every sale was arms-length, well-priced, or comparable in condition. Days on market and price reductions are condition signals, not just noise. A comp that required a cut to close and took 61 days is telling you something about where buyer resistance sits. She is using it as a price anchor when it is closer to a ceiling marker.

The risk you have not named: you are evaluating two pricing tools when the more important variable is your hold calculus. The gap between $191,000 and $198,000 is $7,000, and that is inside normal valuation noise for a two-unit in a secondary market. The question of whether either number is close enough to make a sell decision is separate from which methodology you trust. If your occupied unit is producing rent, the income approach would give you a third data point, and a property that pencils on income in Canton at $147,000 basis may look different than a straight comp analysis suggests.

If you want a number you can actually act on, a licensed appraiser using a mixed approach, sales comps plus income, is the appropriate professional here. That costs money upfront but it is the only figure with methodological accountability behind it.

What is the current rent on the occupied unit, and do you have a sense of what market rent on the vacant side would be?

What was the agent's proposed list price, because that number tells you whether the $198k is an honest read or an audition.

The 61-day sit and the $9,000 cut are exactly what I would have flagged. Those are not stable-market comps, those are properties that were mispriced and had to chase the buyer down. "Stable market" means values aren't swinging, it does not mean every sale was clean. An agent leaning on those two to defend $198k is either not paying attention or is working backward from a number she wants to show you.

On your actual question: the CMA has one advantage over Zillow and it is not accuracy, it is that you can interrogate it. You pushed back on the comps and she had to give you a reason. Zillow gives you nothing to push on. But that advantage disappears fast when the agent has a listing incentive and picks comps that flatter the price, which is what appears to have happened here. What I'd want to know is whether Canton has had any meaningful absorption shift in the last six months, meaning are days-on-market trending up and are list-to-sale ratios compressing. If they are, her "stable market" framing is already stale. What is the current DOM average in that zip right now?