Update: In the months after this post was published, Google updated their algorithm so that 301 redirects no longer have any authority penalty. Many tests on Wayfair sites have shown that we can now 301-redirect pages without any ranking loss (assuming the page contents are the same, of course). In fact, this also means there is no longer any hit for migrating to HTTPS. It's worth noting that this is not true for 302 redirects, however. We still see rankings loss when 302 redirecting pages to a new URL.

At Wayfair.com, we conduct a lot of SEO tests. We're constantly measuring and evaluating our strategies, some of which were shared in our last post for YouMoz, Accidental SEO Tests: When On-Page Optimization Ceases to Matter. Sometimes, however, we stumble across what we call "accidental SEO tests." This typically happens when a bad code deploy unintentionally hurts our SEO, and we end up learning something useful from our mistake.

Tens of thousands of 301 redirects

One of our accidental tests involved regularly 301-redirecting large batches (i.e., tens of thousands) of product pages. On average, we found a consistent (and essentially permanent) traffic loss of about 15% for 301-redirected URLs.

In the past, Google has said a small amount of PageRank is lost through a 301 redirect, which is the same as through a link. Now, for the first time, we can put a hard number to how much that loss is.

Structure of an accidental SEO test

Like any good SEO team, our product pages were set to use the name of the product in the URL. Furthermore, if for any reason a product URL was changed, the old URL was set to automatically redirect to the new one.

What we didn’t realize, though, is that our merchandising teams were also busy being good at their jobs, part of which involved changing the naming standards of products on a regular basis. Every change they made was good for the customer. But when the the naming standards changed, it caused thousands of products to change names. This, in turn, updated the URLs of those product pages, triggering a 301 redirect on every page.

For example, when updating for the purpose of having a consistent style, the merchandising team changed “barstools” to the more accurate two-word version of the product name, “bar stools.” Wayfair had over 8,000 bar stools, all of which 301-redirected to a new URL following the name change. Then, a couple of months later, the merchandising team found that they were getting better results by including the height of the bar stool in the product name, so they updated the product names again, which resulted in the product pages 301-updating once more to a brand new set of URLs.

This process of updating product names was being implemented across dozens of different product classes, with multiple updates per month. It quickly added up to a lot of 301 redirects.

Measuring the impact

After reshaping our URL logic to prevent the constant redirects, we realized that we had a great opportunity to find out exactly how 301 redirects affect organic traffic. Nailing down data was easy. We had the exact dates of the changes; groups of thousands to tens of thousands of pages, with tens of thousands of organic visits; and could compare those classes against others that we knew didn’t change to exclude the impact of Wayfair’s overall increase in organic traffic.

We found with surprising consistency that we had a drop very close to 15% of organic traffic for any product class that changed URLs. In our bar stools example, we lost just under 15% of organic traffic at the first change. When the URL changed again over a month later, we lost another 15%.

Every product class we looked at showed the same drop within one to two weeks of the change. Sometimes the drop was almost immediate (like with bar stools); other times, however, it was spread out over a couple weeks (e.g., area rugs, with over 30,000 products).

We did not see any evidence of recovery from the impact of the 301 redirects, even after many months. There was the appearance of recovery — class traffic levels eventually returned to where they started — but that was because our overall organic traffic was increasing across the entire site. We were still 15% below where we would have been without the redirects.

What’s particularly fascinating about this number, 15%, is that it is exactly the amount of PageRank loss Google described in the original PageRank paper. So our measured results matched theory with surprising precision. Perhaps the broader authority signals Google now measures follow the same logic for flowing through pages as they did in 1998? Or perhaps it’s just a happy coincidence.

What it means

We’ve always known there was a “small” cost to implementing a 301 redirect, but our accidental SEO test showed us that the cost is quite significant, and it becomes much greater with every hop in a redirect chain.

It’s worth stressing, however, that we are not saying that 301-redirecting any particular page is going to cost you 15% of your organic traffic. If you rank in position #1 for competitive terms, a redirect could drop you to position #2 or #4. That would cost you far more than 15% of your organic traffic. On the other hand, your page could be so strong that you may not not see any loss in rankings after redirecting it.

What our data suggests is that, on average, there’s a 15% traffic loss following a 301 redirect; but any individual redirect could be much better, or much worse.

While 301-redirecting a dead or changed page to the new location is still good practice, the best practice of all is not to change your URL in the first place.