Redfin/Internet RE Pricing algorithms suck?

I was looking at Redfin’s pricing algorithm(They show which other houses they have considered to make a prediction on a house price).

So I was checking on a house(all 3 good school districts in Fremont) and they have arrived at that house’s current pricing estimate by looking at 2 houses in Union City(bad schools) @1.5 miles away. In addition, among the remaining 3 houses, 2 that they have considered is a house sold located right across an elementary school(took much longer to sell) and another which had a real bad inside compared to “normal” houses. So, in this case 4/5 houses aren’t really comparable to normal in terms of city+location+condition.

So does Redfin and maybe by logical inference does pricing estimates of other internet RE companies(zillow) suck? & are they facilitating up incorrect pricing of houses in neighborhoods?(at least in the Bay Area)

Yes they all suck.

Redfin could have done better. It knows the school district boundary at least. I am surprised they didn’t bother to improve it. So much for being a “tech” company.


I remember there being a feature (maybe it was Zillow?) where you could claim ownership of a house on their site and then pick good comparable reference points and even update square footage etc if necessary. I think they then updated the price estimate accordingly.

One thing I’ve noticed on Zillow especially is that the price that the house last sold at gets factored in. So the comps they show are not the whole story. I think they are doing some sort of scale up/scale down relative to previous sale price. Both of my properties were relatively underpriced when I bought them. The zestimate and Redfin estimate for them are still much lower than the neighboring homes - in some cases the neighboring ones are basically identical. In one case, an apartment that was “oversold” a few years ago currently has a price estimate that is not achievable, while its neighbors have prices in range.

1 Like

Also I’m talking about condos here so they really are almost identical.

This is junk !!!

Not sure if you are talking about 4288 Warbler Loop Fremont, but same issue there. Stopped by at that house and layout is quite nice.

1 Like

I think algorithms make sense for non-bay area market - they’d be of by 10-30%, which would still not be a huge amount elsewhere.

Lenders and valuation companies insist the appraised value must not be lower than current listing prices. Most of the local original asking price has been bogus and deliberately low to attract buyers before the price hike. For this reason most appraised value for lenders are often lower than contract selling price for same home. As for RF algorithm they never made a big deal out of it and it looks reasonable. I have not looked at Z for sometime as I have 4 different AVMs that banks use as first cut at the mortgage. About 25-30% of my clients have to pad the difference often with >> 20% down if there is a shortage. With 30-40% down banks do care much about appraised value.

1 Like

Around 5 years ago (2012-13) when prices where moving up significantly after the downturn/flat market, Zillow estimate was way behind. Often, the estimates were 25% off from market price. But then they updated their algorithm.

Today, I find both Redfin and Zillow are quite accurate, at least for my properties’ neighborhoods.

I assume their algorithm uses only the general neighborhoods comps and market trend. I wonder if they use school info. In OP’s example, looks like school zone is not used. And I doubt they are smart enough to know immediate neighborhood, like good (quiet) vs bad (busy) street.


Another house which was sold some time back.