We have two completed homes sales to compare against their estimates! That’s the end of the good news!
88 Garden Street (Let’s Tour an Underpriced 2 bed, 2 bath) was the first home that I saw and it sold for $1,205,000 as compared to my estimate of $1,061,100 (-12%). Even as I made the estimate I noted that it felt a little low and that I could see the home going up to $1,150,000.
Though the estimate is not great, one thing I am happy about is that I didn’t update my coefficients after I saw the final number in order to produce something that “looked better.” I liked the $ / square foot base, the neighborhood score, and the usable square footage score and I thought those three factors were enough to understand the value of this property. I think that the error here comes more from my ability to estimate those latter two values, which I can improve, than from the model itself.
My instinct is that I need to rethink the usable square footage score, and how I generate this coefficient, because it produced a $100k haircut that I am not seeing in the result. I made an observation about 88 Garden Street that the living room was narrow and that I would like it better if I could move some square footage from the dining room to that living room. I converted that observation into a 0.9 usable square footage score and I think that this score is far too punitive. I probably need to recalibrate the usable square footage score to contribute no more than a +/- 5% adjustment to the base square footage.
Next up – 2511 Baker Street (A Case Where Auction Mechanics Make Sense), which was a bloodbath. 2511 Baker Street sold for $6,250,000 as compared to my estimate of $7,500,000 (+20%). Ouch.
As I went back to look into this one and see where the difference could come from, I noticed one glaring error in my calculation which is that I credited the house with 3,700 square feet, where both Redfin and the SF property information map say it has 3,000 square feet. This was before I was aware of the property information map resource, but still, I got that number from somewhere. I overheard ~3,500 at the open house and I would have found 3,700 somewhere but I see it neither on Zillow nor Redfin right now.
As it flows through my formula that would have revised my estimate down $910,000 (700 square feet * $1000 / square foot * 1.3 neighborhood score) to between $6,400,000 (low) and $7,900,000 (high). And it would have revised my $7.5m final estimate (picking one number in the range), to $6.6m which is a more respectable 5% over the estimate.
Moral of that story – I’m going to be more careful with my square footage numbers and think about how to better deal with the case where the exact square footage is in question (even to the seller/broker) such as this past weekend with 1662 Union Street.
- Number of home sales: 2
- Sum of final home sale prices: $7,455,000
- Cumulative error (absolute value): $1,393,900
- Percent error: 18.6%
Yikes. I might need to borrow from the NBA and amnesty that 2511 Baker estimate at some point.