Real estate · CMA preparation
A real estate CMA tool that drafts the work and leaves the price to you
Paste the comps you pulled from your MLS and ZeroTwo lays out the adjustment worksheet, flags the comps that do not belong and drafts the seller summary. You choose the comps, check every adjustment and set the range. Nothing here is an appraisal or a valuation guarantee.
You supply
The subject's facts, the closed sales and active listings from your MLS or public records, and the local dollar values behind each adjustment.
ZeroTwo drafts
The comp grid, the adjustment arithmetic, the list of comps that look wrong and why, and the seller-facing summary.
You decide
Which comps stay, whether each adjustment is right, the price range, and what you tell the client. It is a CMA, not an appraisal.
Where the time goes, and what a tool can honestly touch
Some of a CMA is clerical and some of it is judgment. A model can help with the first kind. It should not be asked to do the second.
Worth drafting with AI
- Laying out the comp grid from data you paste
- Arithmetic for each adjustment line, shown so you can check it
- Spotting comps that differ from the subject on obvious fields
- Writing the seller summary in plain language
Slow for a reason, and yours
- Choosing which sales are real comparables for this house
- Knowing what a feature is worth in this submarket
- Judging condition and the things the listing data leaves out
- Making the price call and defending it to the seller
No minutes-saved figure on this page. We have not timed an end-to-end run, so we do not quote one. To get yours, time preparation and review on your next three listings with and without the worksheet.
The CMA method in six steps, with the prompt for each
The defaults are starting points. Local data should override them, and your CMA should say when it did. Every prompt refers to data you paste, so the model has nothing to invent from.
Choose and pull the comps
YouStarting default: 4 to 6 closed sales near the subject, similar in type and size
Pull closed sales from your MLS or public records. The Massachusetts licensing reference describes recent sales as usually within the last six months and says comps in the same neighborhood help a seller relate. Tighten the window in a fast market, widen it in a thin one, and write down which you did.
Prompt
Here are the subject's facts and the closed sales I pulled [paste]. Tell me which sales fail a basic comparability test against the subject and why. Don't add comps I didn't give you.
Open in ZeroTwoExclude what does not belong
You, with a draftStarting default: No distressed or non-arm's-length sales in the set
The same reference lists sales to leave out: short sales, bankruptcies, foreclosures and transfers between relatives. Listing remarks often say so. Keep an excluded-comps list with the reason for each, so the seller can see what you left out.
Prompt
From the listing remarks I pasted, flag any comp that looks like a short sale, foreclosure, relative transfer or other non-arm's-length sale. List separately the comps you could not assess because the remarks were silent.
Open in ZeroTwoDraft the adjustments
ZeroTwo drafts, you set the valuesStarting default: Every adjustment is a dollar line with a reason and a local basis
Adjust for size, condition and features. The dollar value per square foot, per bathroom or per upgrade has to come from your local data. The model applies your values consistently across comps and shows the arithmetic. It should ask when a value is missing instead of supplying one.
Prompt
Using only the per-feature dollar values I list below, adjust each comp for size, condition and features. Show the unadjusted price, each adjustment line with its reason, and the adjusted price. If you need a value I haven't given, ask me instead of guessing.
Open in ZeroTwoRead the market-time context
ZeroTwo draftsStarting default: Days on market by price band, only where there are enough sales
Group the sales you pasted into price bands around the subject and report days on market for each. Skip a band with very few sales rather than quoting a median of two homes, and put the sample size beside every figure.
Prompt
From the sales I pasted, group days on market into price bands around the subject. Show the number of sales in each band, and skip any band with fewer than four.
Open in ZeroTwoCheck the active competition
You supply, ZeroTwo draftsStarting default: Current actives and pendings the subject will be compared with
Export the live competition from your MLS and note the date you pulled it. The model can write a one-line positioning note per listing from the facts you paste, but it cannot see listings you did not give it.
Prompt
For each active and pending listing I pasted, write one line on how it positions against the subject on price per square foot, condition and days on market. Use only the fields I provided.
Open in ZeroTwoWrite the range and the summary
You decide, ZeroTwo draftsStarting default: A range with a stated confidence, not a single magic number
You choose the range and the recommended list price. The model drafts a plain-language summary around your numbers, and it should include a short note on how thin or strong the comp set was.
Prompt
Write a one-page seller summary around this price range and recommendation [paste]. Include the DOM context, the active competition, which comps I excluded and why, and one sentence on how strong or thin the comp set was. Don't change my numbers.
Open in ZeroTwo
A defensibility checklist: defaults, exceptions and what to record
Fixed radius and age rules do not travel across markets. Each row gives a starting default, the situation where you should break it, and the note that makes the break defensible.
| Check | Starting default | Break it when | Write down |
|---|---|---|---|
| Comp count and recency | 4 to 6 closed sales from recent months. The Massachusetts reference says usually within the last six months. | The market is moving fast, or sales are thin. | The date range you used and why you widened or narrowed it. |
| Location | Same neighborhood or the boundary that matters to buyers, such as a school zone. | Rural, unusual or segmented areas where distance understates the difference. | The radius and any boundary you treated as a hard edge. |
| Size comparability | Gross living area close to the subject's. Pick a tolerance from local data; a figure like 15% is a convention some agents use, not a standard. | The only usable comps are noticeably larger or smaller. | The tolerance and each comp's difference from the subject. |
| Condition and features | Each adjustment is a dollar line with a reason and a local basis. | No local evidence exists for what a feature is worth. | Where each dollar value came from. |
| Market-time context | Days on market by price band when there are enough sales. | A band has too few sales to say anything. | The sample size beside each figure. |
| Active competition | The actives and pendings the subject will be compared with. | The export is stale or incomplete. | The date you pulled the list. |
When the market is thin
The state reference raises the question directly: what happens if there are no comps available? Widen the date range or the radius and say so, lean on fewer and larger adjustments only where you have local evidence, mark the range as lower confidence, and consider whether the client needs an appraiser instead of a CMA. A model can tell you which comps are weakest. It cannot create sales that did not happen.
What a CMA can and cannot prove
We make no accuracy claim for CMAs prepared with ZeroTwo. We have not benchmarked them, and published error rates for automated valuations do not measure them either.
7.20%
Zillow's published nationwide median error for off-market Zestimates.
Source: zillow.com/zestimate, checked 5 October 2026. That page's on-market figure showed as "NA" when checked.
A median is not a per-home band
Zillow describes the figure as a median: half of homes fall within that percentage of the final price, half do not. Applied to one house as a plus-or-minus dollar gap, it implies a certainty the statistic does not carry. It also says nothing about how a CMA performs.
Zillow's own page says the Zestimate is not an appraisal and encourages supplementing it with research such as a CMA from an agent. That supports using a CMA. It does not show that a given CMA is more accurate.
Test your own CMAs instead
- After each sale closes, compare your range with the closed price and keep the log.
- Count how many adjustment lines you changed after the model drafted them.
- Compare drafts from two models on the same comp set, then check the disagreements against your own data.
Running several models on the same comps compares drafts. It is not independent validation of a valuation.
CMA or appraisal: who prepares it and how you describe it
| Dimension | CMA | Appraisal |
|---|---|---|
| Who prepares it | A real estate licensee who is not a licensed or certified appraiser. | A state-licensed or state-certified appraiser. |
| Typical use | Showing a seller what a property might be priced at, and deciding whether to take the listing. | Formal valuations such as lending. Ask the lender what it requires. |
| Method | Valuation techniques similar to an appraiser's three approaches, most often sales comparison. | The appraiser's own valuation work under the licensing rules for the state. |
| How you may describe it | Licensees must never imply they are a licensed or certified appraiser. | Governed by the appraiser's license or certification. |
Source: Massachusetts RE62RC25, Comparative Market Analysis (CMA), checked 5 October 2026. It is a state continuing-education reference. Rules differ by state, so check your own licensing board.
What the finished CMA contains
Seven parts. The excluded-comps list and the confidence note are the two this page adds to the usual list.
Subject-property profile
Address, size, beds and baths, lot, year built and a condition note. Every other section reconciles against it.
Adjusted closed comps
Each comp shows its unadjusted price, each adjustment as a dollar line with a reason, and the adjusted price.
Excluded comps and why
The sales you left out, each with its reason. It shows the seller the set was chosen, not found.
Active and pending competition
What the subject will be compared with at launch, with the date the list was pulled.
Market-time context
Days on market by price band, with the number of sales behind each figure.
Range, recommendation and confidence note
Your low and high adjusted values, the recommended list price, and a plain sentence on how strong or thin the evidence was.
Seller-facing summary
A one-page narrative of the recommendation, written around your numbers.
Before you present it
Five checks that matter more than how the document looks.
- Every comp is one you chose, and the excluded list is attached.
- Every adjustment dollar value has a local basis you can state.
- The model's arithmetic was rechecked on at least two comps.
- The range reflects how thin or strong the comp set was.
- The document calls itself a CMA, and nothing in it implies an appraisal.
Frequently asked questions
What is a CMA tool in real estate?
A CMA tool helps an agent assemble a comparative market analysis: the subject property, comparable closed sales, adjustments, active competition and a price range. In this workflow ZeroTwo is a general AI workspace that drafts the worksheet and summary from the data you give it. It is not an MLS and it does not decide the price.
Does ZeroTwo supply the comps?
No. This workflow starts from comps you pull from your MLS or public records and paste or upload. Check your MLS's terms before sending its data to any third-party tool, and keep client details out of prompts unless your own policies allow them.
How accurate is a CMA prepared this way?
We have not measured it. We have not compared CMAs prepared with ZeroTwo against closed sales or against automated valuations, so this page makes no accuracy claim. The accuracy of a CMA depends mainly on the comps you choose and the adjustments you set. Back-test your own range against closed prices to learn how it performs in your market.
How long does it take?
We have not timed an end-to-end run, so there is no minutes-saved figure here. The drafting steps are quick to try. The review steps are not optional: choosing comps, checking adjustments and setting the range stay with you. Time your next three listings with and without the worksheet to get your own number.
Which model should I use?
Use whichever produces arithmetic and wording you can check. ZeroTwo lets you run the same comp set through different models. Agreement between models is not independent validation of a valuation, but a disagreement is a useful prompt to recheck a comp or an adjustment. See the models page for what is available.
What does it cost?
The Free plan is $0 with 100 free credits to start, 20 bonus credits every day you log in, and no card required. Plus is $14.99 per month and Pro is $29.99 per month ($26.99 per month billed annually). Check the pricing page for the limits on each plan.
See pricing and the model list. ZeroTwo publishes this page and sells the workspace it describes.
After the price is set
- Real estate contract review
After the price is set, the transaction starts. Summarize contracts and disclosures separately from valuation.
- AI for marketers
Promotion belongs here: listing copy and campaigns built after the pricing conversation.
- Chat with AI
Run the same comp set through different models and compare the drafts side by side.
- AI blog writer
Turn a pricing story from a closed listing into a market-update post.