Spike of Death - Does traffic spike before negative impact of an update?

[NickMNS]

In the past I've heard people speculating that before being penalized or negatively impacted by an algo update your traffic spikes. I've now been impacted by two recent updates, March and most recently last week. In both cases I clearly see the traffic spike pattern. In March (17th) the spike lasted a full day (24 hours, 8am Saturday to 8am Sunday) and resulted in >30% increase in traffic as compared to the same day (Saturday) for the 3 previous weeks. By the following day traffic had dropped 15% despite continuing to have strong traffic in the early morning hours. Traffic dropped to about 50% for the week over week traffic comparison for the following Saturday (March 24) and remained at that level.

Fast forward to last week. Again I experienced another traffic spike (statistically significant), however this time it was short lived, it only lasted a few minutes, 30 minutes tops. But during that short period my traffic spiked roughly 10X it's normal levels. The attached image shows the spike but doesn't really do it justice. There were other spikes that look similar in the days leading up to Oct 8th. Those other spikes were not really spikes just periods of sustained traffic at the high-end of the range, where traffic ramped up to some level and then ramped down. This can be seen by a growth in traffic in the previous and following hours. But on October 8th the spike shows only for that one hour, 2.5X the previous and next hour's traffic. Traffic in the following days dropped roughly 15% as compared to the respective same day in the previous week.

Is this a coincidence or a repeatable pattern?
Does traffic spike before being hit by an algo update?
If it is a pattern, what can be gleaned from it, if anything?
What about a positive impact?

[imgur.com...]

[aristotle]

Spike of Death

The world will end, not with a bang, but with a whimper

[keyplyr]

Why would it? If there is such a thing, what's the rationale?

What do your raw server logs say?

[NickMNS]

Why would it? If there is such a thing, what's the rationale?

That is the thing. I don't really see any real reason for it. It was a rumor I heard from back in the Panda days, I had never been negatively impacted then and now that I have I am seeing this pattern emerge.

What do your raw server logs say?

I only have my logs for this month, and I have not analyzed them in any meaningful way, but just looking over them manually everything appears to jive. There are no glaring patterns, like being hit many times by the same IP or UA. All the requests appear legitimate.

[keyplyr]

Every once in a while I see a pretty big spike that has no apparent reason.

With TLS and no-referrer headers, there isn't as much log info to glean, and as I often post, I don't like any of the analytics software.

So nowadays I usually assume the traffic spike is from SM. On a normal day, even without me dropping links, FB can send me 10k uniques.

A spike for me would be an additional 10k in one hour. That would likely be someone posted a link to one of my pages and it took off. I would usually notice that, especially if it was at a lower traffic time of day.

I never heard the conjecture about a traffic spike prior to a penalty. I usually stay out of those discussions. Seldom does any real infomation get identified.

[justpassing]

IF:
- traffic spike => bad
- traffic loss => bad
- traffic unchanged => bad, because no evolution
- full moon
...

Keep cool :)

[NickMNS]

So nowadays I usually assume the traffic spike is from SM. On a normal day, even without me dropping links, FB can send me 10k uniques.

I get often get spikes from social media not in order of magnitude that you described and generally much smaller than the ones shown in my graphs. But this is not that. My site gets it's traffic from long tail keywords (that keywords that appear infrequently in search, not long keywords) as such traffic rarely lands on the same page twice. So during normal traffic almost every user is on a different page. This is not the case when a user shares a link on SM, then I see a sudden spike in traffic to single page. This can also occur with breaking news events where there is a sudden interest in an entity for which my site provides stats and information.

Whenever I get a traffic spike, my first check is to see if the traffic is all landing on a single page. But in both these instances this was not that. During the two peaks the mix of pages appeared random as is the case normally. Also, bounce rate, pages per session and average time on page all remained within normal ranges. The only way to get this traffic pattern is to increase the flow of user from Google.

Bots couldn't reproduce this pattern without considerable effort, as the bot would need to know all the potential url's and then select randomly from the list. Not a very likely pattern for a bot and to what end. This pattern appears nonsensical coming from Google, well it would be even more nonsensical for a bot.

I never heard the conjecture about a traffic spike prior to a penalty.

I did a quick search and it is not prevalent theory but I did find reference to it, here is link to a screamingfrog blog post that dates back to 2012, see the heading 'Negative Seo'
[screamingfrog.co.uk...]

Seldom does any real info[r]mation get identified.

I fully agree. I just found it strange, in both instances when I have been impacted by an algo update, I have also seen this pattern emerge. A sample size of 2 is not likely representative.