Two property valuations, two completely different methods.
A CMA — comparative market analysis — is what a real estate agent prepares. They pull recent sales, hand-pick comparable properties, apply judgment-based adjustments for differences, and produce a defensible price range. Time: 60 to 90 minutes. Cost: free at the listing appointment.
An AVM — automated valuation model — is what a software algorithm produces. It pulls public records, MLS data, and property characteristics, runs them through a regression or machine-learning model, and outputs an instant estimate. Time: under a second. Cost: free from sites like Zillow, Redfin, and RealAnalytica.
Both estimate property value. They are not interchangeable. This guide is the practical breakdown of when each one is the right tool.
For the broader landscape (CMA, BPO, appraisal), see CMA vs BPO vs appraisal. For the conceptual background on what a CMA is, see the complete CMA guide.
What Is an AVM?
An AVM is a statistical model that estimates the market value of a property using available data. The most familiar AVMs are:
- Zillow Zestimate — public-facing AVM, available for almost any property in the US
- Redfin Estimate — similar, with their own model
- RPR's AVM and RVM (REALTOR Valuation Model) — agent-facing, MLS-augmented
- RealAnalytica's AVM — agent-facing, MLS-augmented, integrated with the CMA workflow
All of them follow roughly the same pattern: pull recent sales of comparable properties in the area, pull characteristics of the subject (beds, baths, square footage, lot size, year built, sometimes condition rating), feed both into a model, and output an estimated value with a confidence interval.
What an AVM does not do: walk the property, account for interior condition, evaluate curb appeal, look at the kitchen, check the roof, ask the seller about deferred maintenance, or apply local market judgment.
What Is a CMA?
A CMA is the same general idea — estimate value based on comparable sales — but done by a human agent who can walk the property, hand-pick the best comps, apply judgment-based adjustments, and present the result in a format the seller can defend.
The two key differences from an AVM:
- Comp selection is curated, not algorithmic. An agent might use 3 well-chosen comps that match the subject's neighborhood, finish level, and recency. An AVM uses every nearby sale that matches a few quantitative filters, including ones the agent would reject as poor matches.
- Adjustments are explicit and reasoned. An agent assigns a dollar value to a finished basement, a renovated kitchen, an extra bedroom — using calibration from local sold pairs. An AVM either ignores those factors or treats them as model features with implicit (and often miscalibrated) weights.
Where AVMs Beat CMAs
Speed
An AVM produces a value in milliseconds. A CMA takes 60 to 90 minutes. For a seller in the "what is my home worth?" curiosity phase — six months before they would even consider listing — an AVM is the right tool. Asking an agent to do a full CMA for casual interest is a poor use of everyone's time.
Buyer first-pass screening
Buyers scrolling listings need a rough sanity check on whether the ask price is in the right ballpark. An AVM does that in a second. Once they shortlist a property they are serious about, a CMA from their agent gives them the actual offer strategy.
Portfolio scale
Investors evaluating dozens of properties cannot wait for dozens of CMAs. AVMs let them filter the universe down to the 5 to 10 properties worth a real CMA.
Standard properties in well-documented markets
In a typical suburban neighborhood with regular sales activity, similar housing stock, and modest variation in condition, AVMs land within 5% of fair value most of the time. For a 3-bed, 2-bath, 1,800 sqft home built in 1995 in a subdivision where comparable homes sell every two months, the AVM has plenty of data to work with.
Where CMAs Beat AVMs
Unique properties
Custom builds, historic homes, properties with unusual lot configurations, luxury homes, and homes with extensive renovations all confuse AVMs. The model assumes the property is similar to its inputs; when it is not, the estimate drifts. Error rates of 15–20%+ are normal on these properties.
Listing pricing
Pricing a listing is a judgment call, not a calculation. Should you price at the top of the range to capture an aspirational buyer? Slightly below to spark a bidding war? An AVM cannot weigh these tradeoffs because they are about strategy, not data. An agent CMA gets the price right because it is built around the strategic intent, not just the math.
Condition-dependent value
A renovated kitchen can add $30,000 to a property's value. A leaky basement and a failing HVAC can subtract $40,000. AVMs see neither. CMAs see both, because the agent walks the property and applies the adjustments.
Rural and thin-data markets
AVMs need comparable sales to triangulate value. In rural areas where the nearest comp is two miles away and sold three years ago, AVMs degrade fast. An agent who knows the local market can make defensible comparisons that the algorithm cannot.
Adjusting for off-market context
Was the comp a quick sale because of a divorce? Did a comp close at a premium because the buyer was waiving inspections in a heated week? Agents see this in the MLS remarks and adjust. AVMs treat all sold prices as equally informative.
Typical Accuracy
Approximate error ranges in well-documented suburban markets:
- AVM (Zestimate, Redfin Estimate, similar): 4–7% median error for standard homes; 10–20%+ for unique or recently renovated homes
- Agent CMA: 2–5% from final sale price when comps are well-chosen and adjustments are honest
- Formal appraisal: 2–3% from final sale price — but only matters for mortgage underwriting
The numbers vary by market, property type, and the quality of the underlying data. For a primer on the appraisal side of the picture, see CMA vs BPO vs appraisal.
When to Use Each
Use an AVM when…
- A seller is exploring whether their home is worth thinking about listing
- A buyer is screening dozens of listings for first-pass affordability
- An investor is filtering a portfolio
- You need an instant ballpark and the property is standard
- You are getting a refinance estimate before committing to a formal appraisal
Use a CMA when…
- You are listing a property and need a recommended list price
- You are buying a unique, renovated, or condition-dependent property
- The AVM error rate is unacceptable for the decision you are making
- You need a defensible price you can show a seller and explain comp by comp
- You need to factor in property condition, local market judgment, or strategic timing
The Honest Take: They Are Complementary
Most modern real estate platforms ship both, because the two valuations answer different questions.
An AVM is the answer to "is this property worth thinking about?" — instant, free, good enough for screening.
A CMA is the answer to "what should we list this at?" — judgment-driven, defensible, calibrated to the strategy.
How RealAnalytica Handles Both
RealAnalytica ships an AVM and a full CMA workflow inside the same AI-native, end-to-end platform that runs from first lead contact to signed close. That means:
- Instant AVM — for any property, any time, with confidence intervals so agents and clients know the model's certainty
- Full CMA presentations built from MLS comps with AI-assisted adjustments and branded output
- Listing presentations generated from the same contact and property data — no rebuilding in a separate tool
- Branded flyers and social posts for the listing once the seller signs
- Real-estate-specific sequencing tuned to listing-appointment, seller-nurture, and lead-conversion workflows
- Built-in CRM and eSignature — when the seller signs the listing agreement, it happens inside the same session
The result is that an agent uses the AVM for the casual "what is my home worth?" inbound question, then upgrades to the CMA workflow when the conversation gets serious — without leaving the platform, without switching tools, without rebuilding any data.
For the full comparison of CMA tooling options, see the real estate CMA software comparison. For the practical step-by-step on building a CMA, see how to do a CMA in real estate.
Frequently Asked Questions
What is more accurate, a CMA or an AVM?
A well-done CMA is consistently more accurate, especially for unique, renovated, or condition-dependent properties. AVMs typically run 4–7% median error in well-documented suburban markets, while agent CMAs run 2–5% when comps are well-chosen and adjustments are honest. AVM error rates can exceed 15–20% on luxury, custom, or rural properties where comp data is sparse.
Can an AVM replace a CMA for listing a home?
No. Listing pricing is a strategic judgment, not a calculation. AVMs cannot account for property condition, local market timing, or the seller's tradeoffs between price and time-on-market. Use an AVM for the first-look question of whether the home is worth listing; use a CMA to actually price the listing.
Why do Zestimates and Redfin estimates disagree?
Because they use different models, different data, and different weights. Zillow and Redfin each have their own AVM trained on their own dataset, so the estimates often diverge by several percentage points on the same property. Treat both as approximate signals, not authoritative numbers. For a defensible value, you need a CMA or an appraisal.
Do agents use AVMs or only CMAs?
Both. Agents use AVMs for first-pass property screening, quick "what is this worth" questions from clients, and as a sanity check against their own CMA work. They use CMAs whenever the decision matters — pricing a listing, advising on an offer, settling a price dispute. The two tools complement each other.
Is RealAnalytica's AVM more accurate than Zillow's?
It is calibrated specifically to the markets RealAnalytica serves (Rhode Island, Massachusetts, Connecticut) and is augmented with MLS data Zillow does not have access to, which improves accuracy in those markets. For nationwide property estimates, Zillow has more training data; for hyperlocal Northeast properties, RealAnalytica's AVM tends to be tighter. Either way, both should be paired with a proper CMA before any major decision.
