When you're put in charge of SEO during a new website launch, the last thing you want to see is traffic to the site drop after the launch. Especially when you were involved from the very beginning – even if you're thorough and cover your bases, anything can happen, but it's usually on us as SEOs to figure out the cause.

In fact, the site's traffic didn't drop at all. Instead, it turned out to be a very difficult to find tracking issue. Here's the process we took to diagnose the problem.

Reviewing analytics, we noticed a drop in traffic to the website that seemed to coincide with the launch of the new website.

As shown by the following chart, between the dates of Jul 1, 2014 – Aug 27, 2014, overall pageviews on the site were down by 35.62% (∆ 62,786) and entrances are down 37.61% (∆ 18,471).

analytics-screenshot-1

However, there was a similar increase in traffic that started during April 2014 and continued through the end of June 2014.

analytics-screenshot-2

Possible causes

1. Pages on the new site are not being crawled and/or indexed properly.

According to Google Webmaster Tools, between 2/16/2014 and 6/15/2015 there were 3,335 pages indexed. There are currently 2,168 pages indexed.

Conclusion: All of the important pages are still indexed – the only ones we lost were news pages that were redirected and will eventually be indexed again. Additionally, if this were the culprit it would have resulted in a drop in organic search traffic only – we saw a drop from all mediums.

analytics-screenshot-3

2. Old pages were not properly redirected and are no longer receiving traffic.

All old URLs crawled prior to the new site launched are properly redirected. There are sometimes URLs that are missed during that process, and we rely on Google Webmaster Tools to notify us of those URLs (noted as 404 errors).

We exported a list of all URLs in GWT reporting 404 errors and matched those up with traffic numbers from Google Analytics.

Conclusion: For a variety of reasons this process isn't perfect, but of the pages GWT could no longer find, only 171 entrances occurred, a tiny fraction of the 18k+ drop, which suggests we can rule this out as a cause.

3. Meta titles and/or page content was changed on the new site in a way that is substantial enough to result in fewer organic search impressions and fewer clicks.

It's possible meta titles & descriptions were changed enough that rankings were adversely affected. We took a baseline rank report prior to the new site launch and did not see any significant ranking changes. Still, though, we only took a baseline report of the keywords the previous SEO firm had selected.

Ideally we would do keyword research and map meta titles and descriptions to their appropriate pages prior to site launch, but the timeline for the new site launch was extremely tight and so we weren't able to do this and, instead, had to rely on the design firm to ensure titles didn't change too much – for the most part they did that.

Conclusion: This would have only affected organic search traffic and since we saw a drop for all traffic mediums, we can rule this out as a cause.

4. When we implemented redirects on the new site, not all URLs were redirected one-to-one to new URLs. Most were, but every page in the old news section was redirected to the new news homepage. It's possible those news pages were driving a significant amount of traffic in aggregate.

The following table shows traffic to all pages within the news section of the website from 7/22/2014 – 8/27/2014 compared with the period just prior to that (6/15 – 7/21). The new website was launched around Jul 27, 2014.

analytics-screenshot-4

Conclusion: Overall pageviews were down 70% (∆ 1,844), entrances were down 64% (∆ 206), however, this only represents 3% of the overall drop in pageviews and 1% of the overall drop in entrances, so we can rule this out as the cause.

5. The client was running an offline or online campaign that led to a substantial increase in traffic during the period of April 2014 through the end of June 2014.

We spoke with the client and confirmed there was no major advertising push that could have caused traffic to double during that period.

6. GA tracking issue

This seems like the most likely culprit because organic impressions and clicks in Google Webmaster Tools are actually up slightly over the period where we saw a drop in traffic according to analytics.

We pulled the GA tracking code from the old site from Archive.org:

<script type="text/javascript">
var gaJsHost = (("https:" == document.location.protocol) ? "/web/XXXXXXXXXXXXXXX/https://ssl." : "/web/XXXXXXXXXXXXXXX/http://www.");
document.write(unescape("%3Cscript src='" + gaJsHost + "google-analytics.com/ga.js' type='text/javascript'%3E%3C/script%3E"));
</script>
<script type="text/javascript">
try {
var pageTracker = _gat._getTracker("UA-XXXXXXXX-1");
pageTracker._setDomainName("none");
pageTracker._setAllowLinker(true);
pageTracker._trackPageview();
} catch (err) { }
</script>

This is an older, customized implementation of analytics that we're not familiar with, but nothing stood out as potentially causing a problem.

However, digging through analytics a bit more, we did find that, with a very few exceptions, all traffic prior to the new site launch was logged in multiples of 2.

In fact, a spot-check on June 10, 2014 shows that, out of the top-trafficked 249 pages on the site, only 2 pages had traffic counts that were not multiples of 2.

This is almost impossible and indicates the old analytics implementation was double-counting each visitor.

To verify this, we went back to Archive.org and reviewed a snapshot of the website during the period in which traffic spiked. There was an Archive.org snapshot available from May 17, 2014, which is inside the period we saw traffic spike.

Sure enough, we found 2 occurrences of the GA tracking code, which would cause all pageviews to be double-counted. You can see both circled in red in the following screenshot.

ga-double-tracking-1

Traffic increased around May 12, 2014 when a duplicate copy of the GA tracking code was added to the site. It then dropped after the new site was launched because the new site no longer has 2 copies of the tracking code in place and so is now correctly logging pageviews.

Diagnosing problems on a complex site like this one can be a challenge, especially if you only just started on the project (like we did with this client) and are unfamiliar with the site.

It takes time but if you're methodical in your approach and analyze all the possible causes, it's much easier to find the right answer.