How to calculate ARV without guessing

What ARV actually means

After Repair Value is the price a property will sell for to a retail buyer once your renovation is finished. It is not the as-is value today, and not what you paid plus what you spent. Nobody buying your finished house cares what your rehab cost.

Every acquisition formula hangs off this one number. The 70% rule sets your maximum offer at 0.70 times ARV minus repairs, so a 5% error on a $320,000 house moves your offer by $11,200 and your exit by $16,000. ATTOM's Q1 2026 home flipping report put median gross profit at $66,000 on a 25.4% gross return across 64,348 flips. Gross means before rehab, holding and selling costs. A $16,000 ARV miss is a quarter of a margin that has not paid for the drywall yet.

ARV vs appraised value vs the Zestimate

Three numbers get confused constantly, and only one of them is yours to produce.

An appraised value is one appraiser's opinion for one lender on one date, and by default it prices the house as-is. On a rehab loan the lender orders it subject to completion per plans and specifications with your scope attached, and the result is an as-completed value: the appraiser's ARV. A Form 1004D completion report confirms the work afterward.

Automated valuation models publish their own error rates, which is more than most of the industry does. Redfin reports a median error rate of 1.85% for on-market homes and 7.27% for off-market homes across 92 million properties. Zillow publishes the same shape for the Zestimate, near 2% on-market and around 7% off. On-market models can see the list price, which is why they look so good.

Every house you are trying to buy is off-market, so the 7% column is yours, and median means half of all estimates are worse than that. On a $320,000 house the typical miss is about $22,400, with a long tail past it. Use an AVM to check the number you built by hand, never to replace it. If both AVMs disagree with you by more than 10%, one of you is wrong.

How to pick comps the way an appraiser does

Fannie Mae's Selling Guide is the closest thing to a rulebook, because it governs the appraisal that decides whether your retail buyer's loan closes. It requires a minimum of three closed comparable sales, prefers sales closed within the last 12 months, and makes the appraiser state each comp's distance in miles and direction, measured in a straight line. There is no fixed radius rule. There never was.

Tighten what the guide leaves to judgment. Closed and recorded only. Active listings tell you what sellers want; pendings give direction without a price. Start at 90 days and widen only when you run out, because every extra month drags a market-conditions adjustment behind it.

On distance: half a mile in a dense grid, a mile in ordinary suburbia, wider only where nothing closer exists. Never cross a line that changes buyer behavior: school attendance boundaries, city limits with different tax rates, flood zones, the far side of a four-lane arterial. Two houses 900 feet apart can sit in different markets.

Then similarity. Within roughly 20% on gross living area, so a 1,640 square foot subject pulls comps between about 1,310 and 1,970. Same style and story count, because the ranch buyer and the split-level buyer are not the same person. Within about ten years on age. And the line people skip: the comp has to be finished to the level you are finishing to.

Where to pull comps, free and paid

Redfin's sold filter is the best free comp tool an investor has: one screen gives sold price, sold date, a photo gallery and filters for beds, baths and square footage. The photos are the reason. You are grading condition, and condition is the largest adjustment you will make all day. Zillow's recently-sold layer and Realtor.com cover the same sales with weaker filters. County assessor sites are authoritative on ownership and sale date, but useless for price in the dozen or so non-disclosure states such as Texas and Utah, where sale prices never reach public record.

Paid tools such as PropStream, BatchLeads and Privy bundle comps with owner data and let you isolate cash sales. Their pricing moves, so check current rates. The real upgrade is MLS access through a licensed agent or an investor-friendly broker: days on market, price history, concessions, plus the agent remarks that quietly tell you a sale was an estate liquidation. Same data the appraiser reads.

Volume changes nothing here. If you mail a ranked list of 500 owners through Farmrix and 18 call back, you are comping 18 houses this month, not 500. Comping is on-demand work, which is why a free tool plus one agent relationship carries most operators a long way. The property data you need for pricing is a different job from the data you need for finding.

Price per square foot is not a shortcut, it is a wrong answer

Not imprecise. Wrong. It is a different method that produces a number-shaped object, and one minute of arithmetic exposes it. Take the three comps from the grid below:

• Comp A: $312,000 ÷ 1,580 sq ft = $197.47
• Comp B: $341,000 ÷ 1,910 sq ft = $178.53
• Comp C: $289,000 ÷ 1,600 sq ft = $180.63

Average those and you get $185.54. Multiply by the subject's 1,640 square feet and your ARV is $304,300. The grid below, from the same three sales, lands on $322,000. The shortcut is $17,700 light, about 5.5%, and at a 70% offer that is $12,400 you never offered on a deal somebody else just won.

The spread is worse than the average lets on. Comp A alone gives $323,800. Comp B alone gives $292,800. Same three sales, same subject, $31,000 of range depending on which listing you opened first. A method whose answer depends on click order is not a method.

The second version of this error does more damage. Investors take average price per foot and use it as the square-footage adjustment: Comp B is 270 feet bigger, so subtract 270 times $185.54, call it $50,100, and Comp B drops to $290,900. Marginal floor space is not worth the average price of all floor space, because the lot, the garage and the kitchen all got divided by square footage to make that $185.54, and none of them grow when the house does. Contributory value for extra square feet usually runs a fraction of the headline rate, often a quarter to 60% of it, varying by market.

A measurement problem sits underneath it. Since April 1, 2022, Fannie Mae has required the ANSI Z765-2021 standard for gross living area, and ANSI excludes any space partially or completely below grade. A finished walkout basement contributes zero GLA. Your tax card and the MLS count it anyway, so your denominator may not describe what the comp's describes.

Building the adjustment grid

Fannie Mae states plainly that it does not have specific limitations or guidelines associated with net or gross adjustments. What it demands is that each adjustment reflect the market's reaction, and it tells appraisers to override convention: if the rule of thumb is $20 per square foot and paired sales support $100, use $100.

Derive your own rates from paired sales. Pull 8 to 12 recent sales in one submarket, find two that match on everything except one feature, and the price gap is your starting adjustment. Repeat three or four times per feature and take the middle. An afternoon per farm area, not per deal, and the numbers hold six to twelve months.

Work the lines in order of money at stake:

1. Market conditions. Convert 12 months of median sold price in your submarket into a monthly rate. A market moving 4.2% a year is about 0.35% a month. Apply it to every comp by sale date.
2. Condition and finish level. The biggest line and the one you will get wrong. Grade each comp's photos 1 to 5 on kitchen, baths and flooring before you look at the price, so the price does not steer the grade.
3. Gross living area, at your paired-sales rate. Never at the average price per foot.
4. Bathrooms, then garage stalls, then lot size, measured against the neighborhood-typical lot rather than the comp.
5. Bedroom count last, and small, because you already paid for the space the bedroom sits in. Double-counting here is the most common grid error after condition.
6. Concessions. Fannie is explicit that dollar-for-dollar deductions are not appropriate. An $8,000 seller credit rarely moved the price $8,000.

Impose the discipline the guide declines to. If a comp needs more than about 15% gross adjustment, it is not a comp, and no arithmetic converts it into one.

A worked comp adjustment, line by line

Subject: a 1978 ranch, 1,640 square feet, 3 bedrooms, 2 full baths, 2-car attached garage, 7,400 square foot lot, in a suburban grid where the median sold price has risen 4.2% over the past twelve months. Monthly market rate: 0.35%. Your scope is a full cosmetic renovation at a $58,000 budget, taking it to the finish level of a typical renovated resale on that street. Today is July 2026.

Paired sales in this submarket support $55 per square foot of GLA difference, $5,000 for a half bath to full bath upgrade, $6,000 per garage stall, $1.25 per square foot of lot difference, and $3,000 for a bedroom once GLA has been adjusted. Those rates are illustrative. Derive yours, or the grid is decoration.

Comp A took a 1.9% gross adjustment, closed two months ago, sits 0.3 miles away and matches the subject on every physical line. It carries the most weight. Comps B and C needed 7.9% and 12.6% and landed at $326,150 and $325,550 from opposite directions, one pulled down for size, the other pushed up for condition and garage. Two comps within $600 of each other after completely different adjustments is the strongest signal in the grid.

The reconciled range is $318,000 to $326,150, a spread of 2.6%. Call the ARV $322,000 and underwrite $320,000, since the low comp is also the cleanest one. Run the offer: 0.70 times $320,000 is $224,000, minus $58,000 of rehab, leaves $166,000 as the most you can pay. If you are wholesaling and want a $12,000 assignment fee, your contract price is $154,000.

How lenders actually use your ARV

A fix-and-flip lender runs two tests and lends the lesser. One caps the loan at a share of ARV, commonly 65% to 75%. The other caps it at a share of cost, often 85% to 90% of purchase, with rehab released in draws. One lender's published 2026 fix-and-flip terms sit at 75% of ARV and 90% of cost, whichever is lower, with rates of 9% to 12%, 1 to 3 points, and 12-month interest-only payments. Terms vary by lender, by state and by your track record.

The ARV they use is not the one you built. They order their own valuation, usually an as-completed appraisal with your scope attached, sometimes a broker price opinion on smaller loans. If theirs lands under yours, the loan shrinks and the gap comes out of your pocket at closing. Underwrite so the deal survives their number coming in 5% to 10% below yours.

Retail rehab financing works the same way. FHA's 203(k) program sizes the mortgage from the after-improved value, requires at least $5,000 of rehab on the Standard version, and still has to fit inside the county FHA loan limit. A 203(k) buyer is a retail exit for the houses you decided not to renovate.

Where ARV estimates go wrong

Check these before you sign anything.

• Comping to the ceiling. That one $360,000 sale on the street had a quartz waterfall island and a finished basement bar. A $58,000 cosmetic scope does not produce that house.
• Ignoring days on market. A comp that closed in 6 days and one that took 118 days and two price cuts are not the same data point at the same price. The slow one is your ceiling.
• Seasonality. A February closing in Minneapolis and a June closing are different markets wearing the same ZIP code. Adjust or drop the comp.
• Lot differences no data field records. Backs to a park versus backs to a six-lane road. Street view every comp before it enters the grid.
• Flip-to-flip comping. Two investors using each other's exits to justify their own prices is how a submarket talks itself into a loss.
• Non-arm's-length sales. Estate transfers, intrafamily deeds and REO dispositions show up in public records at prices that say nothing about market value. Read the deed type.
• Never re-pulling. On a five-month project in a market moving 0.35% a month, the ARV you wrote at contract is 1.75% stale by listing day. Occasionally that works in your favor.

What to do on your next deal

1. Before you offer, pull three closed and renovated comps inside half a mile and 90 days, within 20% on square footage, same style. Build the grid. Write the number and the date down.
2. Check it against the Zestimate and the Redfin Estimate. Agreement proves nothing. A gap over 10% means go look again.
3. Underwrite the bottom of your adjusted range, not the middle.
4. Re-pull comps the week you list, and again if the house sits past 30 days.
5. Log predicted ARV against actual sale price on every exit. After eight deals you will know your personal bias, and for almost everyone it runs high.

None of this pays anything without deals to run it on. Farmrix handles the front of that funnel: it scores every owner in your market on how likely they are to sell in the next 6 to 12 months, ranks them, and mails postcards to the top few hundred instead of blanketing a bought list. Farmrix packages start at 500 ranked owners with 500 postcards mailed for $1,195. You still price the houses yourself. See motivated seller leads or pricing.