Case Study: Repair Estimate vs Actual Cost on a Detroit Rehab
One of the most common objections to software-generated repair estimates is accuracy. "How can an algorithm know what repairs a house needs?" This case study compares a Deal Run AI repair estimate with the actual renovation cost on a completed Detroit rehab to answer that question with real numbers.
The property
3/1 bungalow in the Grandmont Rosedale neighborhood. 1,100 sqft, built 1948. Purchased at $45,000 for a fix-and-flip. The property had an outdated kitchen, one bathroom needing full renovation, original hardwood floors under carpet, a 15-year-old furnace, and cosmetic wear throughout. The roof was 8 years old and in good condition.
AI estimate vs actual cost
| Category | AI Estimate | Actual Cost | Variance |
|---|---|---|---|
| Kitchen | $8,500 | $9,200 | +$700 (8.2%) |
| Bathroom | $6,000 | $5,800 | -$200 (3.3%) |
| Flooring | $4,500 | $3,200 | -$1,300 (28.9%) |
| Paint (interior) | $3,000 | $2,800 | -$200 (6.7%) |
| HVAC | $5,500 | $6,800 | +$1,300 (23.6%) |
| Electrical | $2,000 | $1,500 | -$500 (25%) |
| Plumbing | $1,500 | $2,100 | +$600 (40%) |
| Exterior / paint | $3,000 | $3,400 | +$400 (13.3%) |
| Miscellaneous | $2,000 | $2,700 | +$700 (35%) |
| Total | $36,000 | $37,500 | +$1,500 (4.2%) |
Analysis
The total estimate was within 4.2% of actual cost — well within the range needed for deal evaluation. Individual line items varied more, with the largest variances in flooring (hardwood refinish was cheaper than expected), HVAC (furnace replacement required ductwork modifications), and plumbing (discovered corroded supply lines during renovation).
The flooring variance was the most instructive: the AI estimated new LVP installation, but the contractor recommended refinishing the original hardwoods under the carpet, which was cheaper and added more value. This is a judgment call that AI cannot make from photos alone but that experienced buyers will make during their walkthrough.
For wholesalers, a repair estimate within 5-10% of actual is more than sufficient for marketing purposes. Buyers expect to adjust estimates based on their own walkthroughs and contractor relationships. The estimate's job is to get them in the door, and a 4.2% variance does that with credibility.