Correcting the Weighted Average in the Redfin Advantage

Redfin reported on the Redfin Advantage last month, analyzing 65,242 records from the Multiple Listing Services that brokers use to share listing data in the Bay Area and Seattle. Our goal was to understand whether Redfin customers fared as well as customers of traditional brokers in their negotiations, so we compared the price our customers paid to the list price, and then compared what customers of other brokers paid too.

And we found that our customers fared better than the average, though we were careful this time to acknowledge that this may have been because our customers were more engaged in the negotiations, or because they were focused on properties more likely to sell at a discount to the list price. We reported that in San Francisco County, our customers’ negotiating advantage compared to the average was 1.635%; in Santa Clara County it was 1.079%; in Seattle it was .498%. The average difference was 1.07%.

But most of our deals came in Seattle where the difference was smaller. In table 3 of the white paper report, we calculated the weighted average of the negotiating advantage. We measured our buyers’ average final price as a percentage of list price, weighting the average according to the number of deals we had completed. We also measured the average final price as a percentage of the list price paid by other brokers’ customers, weighting the average according to the number of deals completed by those brokers. As the table in the report makes plain, we then subtracted one number from the other for a weighted average.

On a call earlier this week, a Windermere broker asked if we should calculate the weighted average differently. He was right to ask. The accurate way to calculate the weighted average is to calculate the difference between Redfin and other brokers, and weight it based on the number of deals Redfin completed in each market. We have adopted this approach, and are re-publishing the report, and changing our site. The new weighted average is lower than it was: .603%. The average for each market was always correct and remains the same.

Our goal was not to overstate the Redfin Advantage; we published the data for others to perform their own calculations, and we are issuing this correction though the broker had already said his question would remain private. We just notified every blogger we briefed or emailed about the Redfin Advantage, explaining the error.

Those of us who are computer scientists, statisticians and brokers deal in numbers every day, and we should know better. We checked and re-checked all the numbers for each market, but our statistics team did not validate the weighted average. The marketing team that prepared the report calculated and re-calculated the weighted average, but just did so by calculating two numbers — one for Redfin, one for the rest of the market — and then subtracting the two.

The Redfin Advantage is still an advantage, and it’s the same advantage that we reported for each market, but the weighted average was miscalculated. We are sorry for our mistake.

I know it has been a ragged week at Redfin. We hold ourselves to a higher standard than this, and apologize to everyone to whom we gave the wrong weighted average. Thanks to the Windermere broker for bringing this to our attention.


  • Pingback: The Redfin Advantage: Bigger, Broader, Higher Statistical Confidence | Redfin Corporate Blog

  • Jeremy Miles

    I meant to ask this in the previous post about this, but I forgot. When calculating the p-value associated with the difference, did you take into account the clustering (non-independence) of individual sales within agents? If you did, did that make a difference?

  • Lane Bailey

    Still wondering about seller paids… I average a about 3% in seller paids on my buyer transactions. (closing costs, etc.)

  • Frank Borges LL0SA= Broker

    Hey Lane, I found that in the DC area the seller subsidy is .75% worse. This is because ONLY DC discloses seller subsidy on the MLS.

    The Reason Redfin can’t have as high seller subsidy is because of their rebate. Lenders combine the seller subsidy and rebates and don’t allow that number to get too high.

    HENCE having to adjust the offers accordingly to LOOK lower, when in fact the NET is NOT lower