"ZOMBIE TRAFFIC" Separating fact from fiction & emotion

[FishingDad]

This recent discussion about "ZOMBIE TRAFFIC" is just utter nonsense. What are people saying, anything worth while or just a communal <snip> because sales are down on the norm? The talk is firmly in the tin foil hat area.

Are you talking about SERPs, if so why, if your positions are dropping then that's that. If positions not dropping are you seriously saying Google is sending you people they know will not buy from you !? REALLY?!

Are you talking about PAY PER CLICK? if so then your talking possible click fraud then, aren’t you?

Giving any constant period on the internet, people buy or they don't buy and there's many many factors why they will one day and might not the next day.

[edited by: goodroi at 5:55 pm (utc) on Nov 10, 2015]
[edit reason] Let's be careful to keep the discussion on a professional level [/edit]

[Robert Charlton]

Simon_H, as I review my last two posts above, they are very rough, with some obvious key points omitted... so it's not surprising you may be taking these comments further than I intended. I apologize for the confusion.

It's late enough that I'm not going to straighten these out tonight... but let me mention for now some thoughts I think are important, and take it more slowly from here. Let's start with just an overveiw of how seed sets might have been used organically...

Essentially, Panda is a trial and error algorithm, "heuristic" is the word used, where initially gut-level types of quality assessments, initiating in a rigorous, scorable set of questions were increasingly refined to produce an algorithm which returned results that filtered out the initial set of bad sites, and which could then be more easily refined from a smaller group that remained.

Initially, separating out a proliferation of shallow, pointless machine-generated content was the big problem that Google needed to deal with. Content Farms, which were annoying everybody (except perhaps to Jason Calacanis) were also targeted early on in the algo's evolution.

Google initially evaluated those seed sites, I'm guessing, on a test bed offline, refined the algorithm until good quality sites (at least by Google's definitions) remained and bad quality sites were removed... and when results were satisfactory enough, Google tested the algorithm, as opposed to the list of sites, on a limited area of the web (eg, on a single data center). I'm assuming the initial refinements were query neutral. In all these tests, the idea was to retain the good sites, get rid of the bad, with no false positives. The algorithm is what evolved... the seed sets were used for calibration, and for guiding initial algorithm choices, using an area of AI call Decision Tree learning.

I'm guessing that some of the zombie results appear at this stage, to sites that are in a limited but public test area in Google's index. Because I'm not a statistician, I can only conjecture that traffic may have been throttled or added to these sites, to allow comparison among some test sites, perhaps with filters on and filters off. With each algo revision, there might be another set of tests. Thus, my guess... the seeming correlation between updates and zombie effects, but not a huge percentage of sites affected. I'm not sure why the same sites get hit frequently, except that they fall in the same grey area, and that Google might choose to follow them over time.

Beyond separating sites with one attribute from sites with an opposite attribute, there must have been evaluation of edge cases that might have kept some sites in the group.

Eventually, though, the algorithm would go wide, to the entire web, then perhaps be further refined, or split into numerous branches. In general, though, I think the emphasis overall would have been to make sure that the algorithmic refinements would scale over a wide range of sites... and allow the next stage of refinements.

Over time, the algo has considered factors of above-the-fold ad density, user intent, user engagement, and factors involving personalization, etc. There is a huge amount of "recursion" in the algorithm... more on that to come.

Also, while there's much that I'd suggest as required reading, the following for me was an important interview, written by Steven Levy, whose book "The Plex" I would also say is required reading for anyone seriously interested in Google. Here's the interview between Levy, Amit Singhal and Matt Cutts.

TED 2011: The 'Panda' That Hates Farms: A Q&A With Google's Top Search Engineers
Steven Levy - 03/03/2011
http://www.wired.com/2011/03/the-panda-that-hates-farms/ [wired.com]

This interval is just scratching the surface, but it's beautifully written, and provides a more orderly introduction than trying to get everything into one post, using too many big words because there's no time for description.


Simon_H wrote...

Regarding the cyclic AI correction factor, yes, data arrays would presumably be shared across organic and, say, shopping results so each would end up learning from the other and there would be in indirect dependency.

I think that we may be looking at the "dependency" between organic and paid differently, in that I wouldn't have said that there's a dependency between the two at all.

I'm very much doubting that there's any correction factor after clicks that are shared, and I have no idea of how close the correspondence is that everybody is seeing between organic and PPC. That correlation, I think, belongs in the AdWords forum, and I haven't seen anyone post on it yet.

I think both organic and paid may be receiving the same RankBrain query feed, of extremely odd queries, but that is a big guess. I'd love to hear what kinds of queries members think aren't working any more. My guess is that just as keyword-stuffed titles are being rewritten, long keyword heavy queries, targeting say exact phrases, might not work any more. Feedback would be appreciated.

The other possibility seems to me would be random clicks by human spammers, trying to hide obvious click patterns by introducing lots of distraction. There's a video about something similar being observed on Facebook ads a while back. I'll try to dig up.

Beyond that, there are too many unknowns about the affected queries, and about RankBrain, let alone about Zombies, for me to try to connect the two...

More to come, though slowly. Again, I'd appreciate input from those who've seen and measured the results first hand.


Note: Edited above to add attribution to Simon_H's quote for clarity, and to correct typo.

[Simon_H]

@Robert_Charlton Wow - thanks for that. Really looking forward to the more that's to come! I'm sure we all appreciate any conclusions by anyone are very rough. But still incredibly useful to discuss thoughts and not hold back.

One immediate comment... You mention that PPC-related observations/correlation belong in the Adwords forum, but I don't think that's the best approach. The zombie phenomenon potentially falls across organic, paid, RankBrain, maybe Panda, etc, especially if the cause sits very early in query processing such that it affects the whole serp. This may be rubbish, but I'd like the opportunity to discuss this without having to start a new thread. So I believe we need a single thread to discuss this. I appreciate there's a risk the thread loses focus, but I'm sure that can be managed.

[glakes]

You mention that PPC-related observations/correlation belong in the Adwords forum, but I don't think that's the best approach. The zombie phenomenon potentially falls across organic, paid, RankBrain, maybe Panda, etc, especially if the cause sits very early in query processing such that it affects the whole serp.

The two are related with conversions turning off and then on for both free and paid results. Splitting the discussion about zombies off into two threads, after originally being split off the main Google SEO thread, which is now closed, is just about as confusing as to why we are seeing zombie patterns for months. Anyway, contributing to such discussions is futile when threads are needlessly closed, after being moved, only to be restarted again in now two active threads and references to zombies patterns in the main Google SEO forum without using the word "zombie" for fear of having the post deleted. I said it before, and I'll say it again - zombies are the biggest news and most active issue facing a good number of ecommerce websites (mine included).

Anyone with the least bit of business sense, who loses $x,xxx.xx in a day to zombie traffic is not going to have active campaigns for very long. A lot of people feel violated by Google, with some wanting a portion of their money back for what is clearly invalid clicks. Some caught the problem right away while others lost even more hoping that Google would fix what is perceived to be a problem (our problem of zombie clicks is profit for Google). Months later zombies are still coming from free and paid search results. Campaigns go off with zombies, get restarted a week later for testing and more money is lost confirming that zombies are still clicking paid ads. The only constant, without losing money, is to analyze free visitors. Once the zombie pattern changes in the free visitors, then it may be worthwhile to run paid ads again without having to monitor zombie clicks every hour on the hour.

So yes, I agree with Simon. The subject matter is zombies and they come and go on both free and paid Google search results at the same time. Make one thread and keep it in one thread. And for the people that get confused after visiting such a thread, then read the posts instead of jumping in on page 5 of the thread expecting to know it all.

[Simon_H]

Thanks @glakes. Just seen the other zombie thread.

@Robert_Charlton Happy to continue our discussion here if you want as it's not noisy. Or perhaps you can move the (sensible) posts from here to the other thread and shut this down.

To add some meat to your test set theory... We experienced the 'sinister surge' (as Glenn Gabe calls it) before Panda 4.1. We saw a big increase at Panda 4.0 in May 2014 (we'd been hit in the past), but then a further significant prolonged and growing surge from the beginning of August and then killed with Panda 4.1 in Sept. So it's fair to say that we seem to fall in the Panda grey area, although these increases were good traffic, not zombie traffic. I've always assumed that test surges are site specific, i.e. Google is measuring the response to the surges and will apply any findings to the site itself only. But that makes little sense, as it would mean Panda is being applied inconsistently. More likely is that sites experiencing these prolonged surges are part of a test set where results are used to shape the Panda filter, and that filter is then applied to all sites when deployed, including the ones being tested.

So our particular case would potentially correlate with a test set theory. But really need to hear from others who can similarly analyse their traffic.

[wgchris]

Long time, no post. Looks like my broken site has climbed out of the bad bucket. Conversions are normal and holding steady, like they've always been. Maybe RankBrain had me sitting in the corner until it knew what to do with me. *shrugs*. Here's to hoping things keep rolling.
<snip>

[edited by: goodroi at 6:17 pm (utc) on Nov 19, 2015]
[edit reason] Please discuss the topic and avoid personal comments :) [/edit]

[Awarn]

You wonder how Google can control traffic. For a few years it just seemed like we had a penalty and were not being indexed properly. Oh we tried all kinds of things to correct it. Seemed like an algorithmic penalty. Never had a manual penalty. The one very strange thing I saw was that structured data never seemed to get recognized to the level it should. Oh sure about 20% showed but not anywhere close to what it should. Basically a flat line. Then late October overnight it was like a switch flipped. Sales shot up, structured data more than doubled. Jumped to position 6. Now the interesting part. On 11/04 a partial penalty appears. We have consistently been cleaning links and using the disavow. Structured data is again a flat line (we are about 50% of where it should be). It just looks very obvious now that we are being controlled.

[Simon_H]

@Awarn When you say that structured data didn't seem to get recognised but then doubled, do you mean this is what Google Search Console showed? Or was there another way you assessed this?

[Awarn]

Yes search console shows this but when I test the data I know it is accurate. I have used the G tool and Yandex tool. I made no changes in the timeframe that would affect the data. Another thing is it is a product page that basically all products use the same page, You know ......-i1234.html so if one page works all product pages should work.

[goodoldweb]

Looks like someone flipped the switch back ON. Strong sales over the past 48 hours or so. Left right and center. A number of different websites, including eBay.

Anyone else noticing this?

[ecommerceprofit]

for last 3 days been good - looks like switch flipped back on as you say

[Simon_H]

@goodoldweb @ecommerceprofit Nice! (No change here though). There's been some serious fluctuations in Google algo weather forecasts since Wednesday, so that may correlate with what you're seeing. Despite the magnitude of the fluctuations, Google haven't yet confirmed that this is anything other than a standard core algo update.

This may be very revealing because if Google does eventually give some idea of what the update relates to or anyone else analyses the effects of the the algo change, it may give us some big clues as to what is causing the zombie phenomenon.

The only caveat is that @goodoldweb mentioned that this is affecting eBay too. So maybe this is purely seasonal. Have you seen overall traffic volume increases or are you seeing identical traffic volume but much better conversions?

[ecommerceprofit]

No traffic volume increases...but excellent conversions that we have enjoyed for years during every month of the year...especially during November... have not stopped liked they used to...up...down...up...down...

We are crossing fingers that our site was some sort "lab rat" in Google's experiment and they let us go back into the wild...or better yet...perhaps our web site was abducted by Google's new alien algorithm for testing and released back to Earth...of course, no one believes us abductees and treats us like crazy people...

[Simon_H]

Not sure why we didn't spot this before (or did someone?) but check out the Metrics in Mozcast. Around 17th Sept, there was a huge drop in Domain Diversity and PMD. Then a few days ago, they've jumped right back up to pre-Sept 17th levels. These dates correspond with when the zombie phenomenon was reported to have kicked in and when (some) people are reporting 'recoveries' from it.

I've mentioned EMD/PMD before and there were at least 3 zombie sufferers that had partial match domains, including us. Any thoughts?

[blend27]

Old, almost forgotten page 2(in goog serp) Ecom site, 24 Orders TODAY on a site that has an overage of 2-3 a month at best for the past 11 month. I was going to shut the site down at the end of the year.

All orders within 4 hours period(ALL from GOOG). Slight Spike in traffic + Orders.

Checked, SERP is the same, just organic traffic.

Great Burst..

Now, several hours later as dead as it was.

[3zero]

Maybe the market is just saturated....

[ecommerceprofit]

Yesterday and today down big time...oh well.

[Simon_H]

@ecommerceprofit Go to Metrics on Mozcast and change to 90 days. Check out Domain Diversity and PMD Influence. On Sept 17th they dropped massively = these zombie threads started. Around 17th/18th/19th Nov they returned to normal, i.e. when you said sales were good. 20th Nov onwards, they bombed again. This seems to correlate perfectly with your dates...?

[ecommerceprofit]

I'm not so sure we can ever explain any of this...I looked and see your point (and enjoy your theories and everyone else's theories) and will keep following the discussion and reporting what happens to our site.

[goodoldweb]

Traffic is spiking but back to square one here as far as conversions go. The switch has been flipped to OFF again.

[ecommerceprofit]

and to clarify...our traffic is not spiking but at normal levels as usual...just conversions in the tank for 2 days.

[WhiteHatTryHard]

The problem is that most people posting to these Zombie Traffic threads are running websites with maybe 5 to 12 sales per day or even less and at such low numbers, you are always going to have huge fluctuations. - This makes the issue look very conspiratorial and tinfoilishly stupid.

In my personal opinion all that we are really seeing is that people have less and less disposable income to spend on useless online shopping, as all the debt that our western governments have amassed is slowly, but surely coming to bite us in the ass.

If there is going to be an economic collapse in the west, its going to hit internet marketers very hard, since we generally do not sell goods that are considered "essential".

[goodoldweb]

^ nonsense

There is so much evidance that Google are actively controling/filtering/throttling/capping traffic to websites, ecom in particular. Hence, your comments simply indicate inexperience when it comes to selling goods online.

[Simon_H]

@WhiteHatTryHard I think you're generalising. There is certainly a contingent of commenters with one or more low quality/low revenue ecommerce websites who preach endlessly how evil Google is for not ranking them at #1 and appear to have jumped on the zombie bandwagon when all they're seeing are natural fluctuations or a gradual decline in sales due to competition or simply don't know how to properly manage paid campaigns.

But some of us have popular high revenue sites and we do see this phenomenon. We certainly see periods of hours/days with continual basket adds and sales and this will suddenly stop dead for hours, with no connection to time of day or other factors that may affect traffic. And this phenomenon isn't new; it's been reported in the past, especially during these fluid periods when Google is in the middle of testing/deploying major updates.

Also bear in mind that zombie traffic is just a sensationalised way of describing traffic shaping/quotas, something else that has been regularly reported for many years. This shaping may be a clever way that Google applies a penalty/filter to a site perceived as low quality, or it may be how Google controls traffic to a site that is part of a test set, or it may be Google playing with paid widgets appearing in different serps, or it may just be the algo going wrong. Also, we know that Google traffic shapes with paid; that's the whole principle of how it spreads budget throughout the day taking quality score into account. So it's naive to conclude that this phenomenon doesn't exist just because you don't see it on your own website.

[tangor]

Chuckles. There are real world reasons why things go bump on the user/economic side. Then again, there are reasons the algo goes bump. I don't see these as being mutually exclusive. What is more important, for the ecommerce site, is what actual loss is accrued during these periods? That is expressed on the bottom line. IE. site generally sells 1 item an hour and misses an hour compared to a site that sells 1000 items per hour. First site has valid 200% markup per item, the latter has 4-7% markup per item --- or any of the many variations thereof.

At present, while I can see it on some sites, I don't let this eat my life... rather, it is a cost of doing business and I have other things to occupy my mind to grow the site(s) in value and income production.

[goodoldweb]

There is certainly a contingent of commenters with one or more low quality/low revenue ecommerce websites who preach endlessly how evil Google is for not ranking them at #1 and appear to have jumped on the zombie bandwagon when all they're seeing are natural fluctuations or a gradual decline in sales due to competition or simply don't know how to properly manage paid campaigns.

And how do you know that?... how would you possibly know your business is bigger then theirs or that you have better skills at managing paid ad campaigns. Simply doesn't make any sense...

[glakes]

Also bear in mind that zombie traffic is just a sensationalised way of describing traffic shaping/quotas, something else that has been regularly reported for many years. This shaping may be a clever way that Google applies a penalty/filter to a site perceived as low quality, or it may be how Google controls traffic to a site that is part of a test set, or it may be Google playing with paid widgets appearing in different serps, or it may just be the algo going wrong.

Or it may be Google trying to earn more money. Why is it that this is always omitted from any possibility? Zombies, shaping, quotas, etc. are all forms of MANIPULATION. If it were not manipulation, then the best sites would get the best traffic and sales Google has to offer. Not once every week or two.

[goodoldweb]

^ exactly!

[Simon_H]

@tangor I totally agree that these on/off periods can be perceived as a natural cost of doing business just like any other algo or seasonal changes. But you typically want to try to reduce the cost of doing business. The point of discussing it is to try to identify the cause. Nothing wrong with that. We earn good money from Google Shopping despite working on very low mark-up. Nevertheless, I'd still like to understand why things turn on and off on both paid and organic despite traffic volume remaining unchanged. Just a shame that all theories to date don't have any easy solutions.

@glakes Yes, it is manipulation. But every part of a serp is manipulation. And, yes, Google wants to earn more money. I don't see the issue. Some sites do very well. Others don't. You need to get your site into the first group. May I ask a question... On your paid campaigns, did your actual daily spend come close to (or reach) your daily budget? Or were you always way below your daily budget?

[aristotle]

Here's my theory in a nutshell:

Zombie traffic = ordinary normal traffic with the likely buyers removed

In other words, Google identifies people who know what they want and are ready to buy, then uses various methods to entice (divert) them to favored sites. The rest of the traffic stays the same, but since it doesn't produce many sales, people call it zombie traffic.

[EditorialGuy]

If it were not manipulation, then the best sites would get the best traffic and sales Google has to offer. Not once every week or two.

And the best sites are....?