Zombie Traffic from Google and Traffic Shaping/Throttling - Analysis

[tedster]

We need a dedicated thread to look at this odd phenomenon being reported by a subset of our members. It really isn't about any particular "update" because the apparent signs have been reported since 2008.

I have personally seen just a few examples of traffic shaping and nothing I could really call zombie traffic, but I think it's time for all of us to take the reports seriously and at least give advice on how to analyze what these webmasters are seeing.

To truly make sense of this, we'll need to pull in many areas of Google that we rarely talk about. This ain't your daddy's SEO! Here's a pretty good overview, from 2010: [webmasterworld.com...]

[tedster]

Here are some questions I'd like to ask to begin fine-tuning the discussion:

Zombie Traffic
1. Are the keywords providing the zombie search traffic the same as those that provide converting search traffic?
2. During the time periods when zombie traffic occurs, are the rankings changed in any way?
3. What is the on-site behavior of this traffic? Do they show more pages per visit? Fewer?
4. Do the sombie periods happen at the same time of day or day of the week?
5. Are there any notable geographic differences?
6. Is the total traffic high enough for these changes to be statistically significant?

Traffic Shaping
1. How much variation is there from high to low traffic levels?
2. Is the traffic capped on a daily basis? Weekly?
3. Does weekend traffic cap the same as week day traffic?
4. Any clue HOW Google is accomplishing this effect?
5. Do you think it's intentional on Google's part, or possibly an unintended artifact?

I think those affected need to study the bejeezus out of their data. Fire up those spreadsheets!

[scottsonline]

Tedster I need to lay out some history first. Up until about 2/10 our site was busy. Died slowly through February and march. Zombies. Last few days of march the bounce rate plummeted. Last week traffic referrals and conversions at record levels.

Flip the coin. A close friend in another niche went stone cold quiet saturday. traffic levels time on site and pages are the same as pre saturday. Much like our site in march he is seeing equal statistical traffic but not the same converting traffic.

My thoughts are a threshold is tripped and undefined/uncertain traffic is sent. The tire kickers that google has not been able to get a handle on. The people that dont buy. Maybe google bundles them into buckets thinking "maybe this is the type of site they will like" and that's the push we see as the buckets are moved from site to site. Through the algo theyve been grouped as nomads and are being pushed in the hopes they find that site that'll make them happy.

Well think about it. Which group shops but doesn't "buy?". Competitors, people thinking about a purchase but that are months away from buying, people researching for others etc. When we saw the weird traffic in march there were many 3-5 week time to purchase in that month. A statistical anomaly for us. We also saw much traffic from competitors that thought we were the manufacturer. That all stopped a week ago and conversions went straight up.

What changed? Beats me! But I half think the zombies my friends are seeing are the bucket of malcontents and competitors that google has not yet classified and is testing against sets of sites.

[petehall]

It's 8:30am here, and already the site has taken 70% of the entire sales for yesterday. Traffic levels are completely normal.

Why? Zombies?

In the UK its a National holiday today. People are likely to be in their homes with access to a laptop or PC. I think this is why the traffic today is converting better.

According to my own data, an iPhone search referral just looks like a standard safari referral. There's no way of telling it's from a mobile device (unless someone can help me with this).

The site works perfectly in Safari on an iPhone, Mac or PC... however a sale in this instance involves entering quite lot of information. Something which would be a real pain on the iPhone.

I think the zombies are smart phone users. We get much more weekend traffic than we ever used to - I think this is the zombies returning on their home PC or laptop on a weekend.

@scottsonline I'd be interested to know if your sales process was really simple or not, and whether you think this could be a valid reason.

[scooterdude]

Some analytics packages have a drill down facility to check the browser and screen resolution of the visitor, helps to see if it where a mobile surfer.

Besides, there was a thread here on ww, where it was suggested that mobile surfers convert the best, perhaps Tedster knows which thread it was :)

[aristotle]

It appears to me that some people here mistakenly interpret random statistical fluctuations as real changes in the amount or quality of Google traffic. I noticed on another thread that someone gives almost hourly updates on these random statistical fluctuations.

[tedster]

I have a similar suspicion, aristotle. At the same time, depending on the details there might be something here (where there'e smoke, there may be fire.) I'm hoping we can use this thread to get into more solid analysis of what people are seeing.

For example, a few years back someone reported that his traffic went to zero every day once it hit a certain number. He shared his analytics with me in private, and that was most certainly the pattern - no matter what time of day, once Google Search traffic hit that number, it fell to zero for the rest of the day (until midnight local time.)

That was the strongest evidence I've seen for traffic shaping. Would love to see something that solid relating to zombie traffic, too/

[Rasputin]

How to know if a site has traffic throttling? I suspect our main site does, and that it operates 'hour by hour' rather than day by day (ie if one hour is high the next seems to 'correct' it), but I don't know how to calculate whether it is 'normal' or not.

For example, the same day of the week each week for the last 5 weeks, google source only, unique visitors:
week 1: 3468
week 2: 3492
week 3: 3433
week 4: 3538
week 5: 3544

Other sites we have don't show the same effect - but have less traffic so perhaps would naturally appear more random.
Traffic does vary by day of the week - sundays better than saturdays for example - but each saturday, sunday etc is consistent with the previous one.

Over the last year because of panda the whole site has been transformed - much improved articles, better template and layout, attempts to improve user and social engagement etc - but the numbers usually follow exactly the same pattern.
In analytics an 'hour by hour' comparison chart with the previous week shows they are exceptionally similar.

So question is, how to actually conclude that traffic resticting is taking place or whether this is normal?

Note: in all posts referring to 'wildly changing results' it would be useful to know approximately how many visitors the sites actually get each day and what levels of variance are being referred to etc

[bluntforce]

I don't know about zombie traffic, but I've seen traffic that converts flow along nicely for part of a day, then it shuts off, only to resume some number of hours later. It used to be weekends were totally dead, but that changed over a year ago.

From what I've seen, it's keyword related, not page related as a given page will apparently retain it's position in the serps for one keyword but will drop significantly for another keyword.

With that said, a shorter term way to deal with it is to promote alternate keywords/phrases. If a "main" keyword/phrase is reduced, the alternates keep things going.

My personal impression is it's taxonomy/architecture based. The algorithm/classifier can't quite figure out where a page should fit given the information it has.

I've recently gone the "noindex" route with some pages that received authoritative inbound links in the last year or so, but they aren't pages that generate income or thematically support income producing pages. I'll give it a month or two to see how that goes.

I know traffic shaping/zombie traffic is real because I've seen it, but I think it's more related to individual site clarity/focus rather than a search engine agenda. Like an electrical breaker, power goes through until it trips, then nothing until it's reset.

[Shaddows]

We're an ecom, and our value proposition is NOT focussed on price (needless to say, that is still important).

We have a lot of information, we have unboxing videos, UGC reviews (purchaser and Pro reviewer only). We have product selectors.

We have a broad product range, but a small % makes up the vast majority of sales.

Now, we have funnels. Funnels are subject to statistical variance but suprisingly stable net outcomes. To be clear, the number of visitors starting each funnel, and the rate of completion for that funnel, varies enormously. But the total number of transactions (the sum over tunnels) is much more stable.

Occasionally, we get odd referral patterns, where a higher number of users come in on certain products than expected. There is a corresponding drop in expected volume against other products. Multiple products occur in both the plu and minus camp, all gradually but similtanueously. We call this a referral shift. It is NOT zombie traffic, because it follows expected funnel behaviour.

Zombie traffic is non-conforming traffic. It shows up on products, often counter-cyclically to the referral shift. So, if the Group A products are down on normal traffic, there will be some degree of Zombie traffic filling in.

But zombies do not confine themselves to voids in normal traffic, they roam around on the rest of our products.

They are quite often foreign, always appear real from User-Agent perspective, and just refuse to engage with the site. Bounce is high (single page request, no exit click), pageviews are low.

Re: Traffic Shaping- we took a revenue hit when a low value product got national press. All the extra referrals canabalised existing referrals. We couldn't pin down the mechanism, but then we've focussed on "What is ranking for [term]" rather than "Where are we ranking for [term]"

----------
Added

The reason I mentioned our value prop and other stuff is because we recognise we have different types of visitors.

We have bouncers that are price checking, and we're not it. We have info-seekers, but they engage with a non-transactional element at a given rate. Neither of these are what I mean by zombies, and I just wanted to be clear that we already filter them out.

[netmeg]

First off, I don't really believe in zombie (human) traffic or traffic shaping; I've never seen it across 250+ sites, some with high traffic, and I can't think of a motive behind it. I can't speak to traffic shaping.

However, for those of you reporting high levels of traffic that does not engage with your site (particularly foreign traffic) does this traffic actually show as being referred from Google?

The reason I ask is that one of my sites has been zombie-BOTTED for going on two months now. A growing number of others of us have also noticed this (it's outlined here in the Analytics forum [webmasterworld.com...] and this is not like most zombie bot attacks, in that it absolutely looks like human browser traffic. But it's being logged as direct traffic - there is no referrer. And it all shows as IE on some form of Windows.

So you might want to look closely at your log files and analytics. Just in case. I've seen at least three cases in the past ten days where someone didn't even realize they were getting this kind of traffic till they read my blog post on it, and went and looked.

Just puttin' that out there.

(petehall - you can absolutely discern your mobile traffic with your log files or your analytics or stats programs - iPhones show their operating system and their screen resolution (generally 320 x 480 or something close to that)

[edited by: Robert_Charlton at 2:00 am (utc) on Apr 11, 2012]
[edit reason] fixed link [/edit]

[petehall]

@netmeg thanks for your help but I was actually meaning form the referring Google URL. I know stats can pick out this sort of thing, but I like to have a look through my own system at live data and all I record is the referral.

Suppose I'll have to start recording some more data....

[netmeg]

Well no, you're not going to get that from the Google referral (unless it comes from Google mobile) because that's a different kind of metric.

[Shaddows]

Google referrals only- we are currently in a busy zombie period, and about 20% of referrals are zombie. "Normal" is more like 5%.

We're lucky in that our zombie traffic just does not impact our normal transaction rate. It often ramps up before and during some kind of major shake-up. We normally sail through those changes with little impact, but the pain is reflected in the general outrage seen from some quarters in other threads.

To be fair, I didn't beleive in traffic shaping until one product started stealing other referrals- I've never seen a buzz-cut line that is the classic (if rare) report. And I didn't believe in zombies until we modelled it.

In a way, "Zombie" is an unnecessary personification. Until the last year or so, we just left it unanalysed, as temporary unfocussed traffic during an major or minor update. We assumed it was the algo honing its query/user intent matching by website. So we just accepted the additional traffic, higher bounce rate and lower conversion rate. It was temporary, uninteresting and non-impactful.

Its only since we started trying to extract more revenue from the blips that the low-level persistance, the evolution (it slowly ramps up, and usually rapidly falls at the macroscopic level) and the total failure to engage made it more interesting.

[Seb7]

I had a thought last night, has anyone checked to see if the zombies are generated using iframes.
ie. is the top.location the same as self.location ?

[backdraft7]

In my niche, for the time being, the Zombie's have been held at bay for the past 3 days.

When they do arrive, sales conversions drop off the charts for periods of 6 or more hours, morning, noon or night. During those times visitors are coming through the "side doors", or sub pages. Many are foreign.

As I explained long ago when I first mentioned Zombie Traffic, I have door bells on key pages throughout my site so I can monitor my site activity without even looking. When these periods of poor conversions start, the doorbells go silent, traffic levels remain about the same, serps don't change at all and I have not made any recent changes to the site. It just appears that people just stop buying for no real reason. I attribute this to non-targeted traffic aka Zombies.

I know my market well, I know my traffic patterns well (by sound) and all of this started after the May Day update. Prior to that time, I enjoyed years of mostly trouble free traffic. When things are running "normal" I can go for weeks or months with the same consistent conversions. But at update time I can hear the change. I then come back to WebmasterWorld to see what's up and sure enough there is another algo update announced.

We all panic when these updates are announced and many see the same outages I see, some do not. However, after all of the recent updates (BTW - Panda had very little negative effect on my site) the pattern seems to remain the same....serps remain basically unchanged, sudden loss of traffic quality, then slowly things come back to normal. I believe it's more of a system loading issue (as mentioned in my last post on the April SERP topic).

The bottom line? I really don't see a point in trying to track down specific reasons because in the end it all remains a big black box and we'll never know if we are right or wrong (which is what infuriates many webmasters), so we're just chasing our tails.

As far as traffic throttling: This seems to be a separate issue that occurs during "normal" traffic quality periods. I've examined sales numbers from years back and they showed "natural" sinusoidal variations over weeks and months, and were increasing ever so slightly ever year.

Again, after May Day, patterns became almost "square wave". It's rather odd when you get consecutive weeks of exactly 15 sales per day, every day. Not sure how they know I made a sale because I do not set conversion target pages. No matter, I'm sure they know.
Now days, sales are either the same or lower than previous years. Oh, and my content is usually more popular when the economy is slow and gas, food and beverage prices are high, so the economy angle makes little sense. My competition has not changed, in fact we lost a few competitors.

Is traffic shaping intentional? Only Google knows. One thing is sure, there is no shortage of conspiracy theories for throttling. I'm not sure the two are the same. Or is shaping just the PC version? ;^)

The good thing is that once the smoke clears from these major updates, we start clicking along fine again. I'm happy to say that none of the major algo updates have had any lasting negative effect (other than the apparent throttling) on the site. At least we're still in business.

I think I need to look into nutmeg's zombie bot topic...
Looked into it, it does not appear apply to my situation.

[edited by: backdraft7 at 2:19 am (utc) on Apr 11, 2012]

[backdraft7]

One more comment about "Zombie Traffic" I believe the first time I myself mentioned it was June 5 2010. Maybe someone else mentioned it in 2008...but I'm not sure who or why.

backdraft7 - 7:13 pm on Jun 5, 2010 (utc -6)

About 2 weeks into this Mayday change and still pounded down to half or less of the sales level of last year. While I'm tempted to make some site changes to improve conversion, I am hesitant to do so because sales we going along great right up to midnight of May 17th. Traffic looks to be the same as last year, but this is some kind of weird "zombie" traffic.

[edited by: backdraft7 at 2:22 am (utc) on Apr 11, 2012]

[Robert Charlton]

Mod's note.... just fixed link in netmeg's post above. For reference, the thread she's talking about is...

Logs Show Surge, but Not Human?
http://www.webmasterworld.com/analytics/4420174.htm [webmasterworld.com]

[bluntforce]

@backdraft7

Understanding user behavior requires analysis of user behavior. If keyword A brings in 1,000 users but only 5 enter a transactional funnel, perhaps keyword B that only brings in 200 users but also has 5 entering a transactional funnel should be looked at a little closer.

Helpful hint: Don't call her nutmeg.

[backdraft7]

Helpful hint: Don't call her nutmeg.

LOL! I honestly saw it as nutmeg...sorry netmeg! I'll put my glasses on from now on.

Understanding user behavior requires analysis of user behavior.

And here lies the paradox...if it works great for weeks / month on end, then suddenly shuts down, what good is that analysis? Then we're assuming the site changed the user's reaction, when in reality it's more likely a matter of traffic quality (non-targeted or otherwise).

Once you find a design that funnels people in and works, why change the funnel? That's always been my approach - build a site that converts well, then throw traffic at it (might sound unscientific, but it always worked) ....only problem is, if that traffic is non-targeted (or Zombie), then that approach becomes useless...as do most other approaches. Poorly targeted traffic might as well be no traffic at all.

As Larry the Cable Guy would say - "It's like checkin on yer burgers when they're already burnt!"

(or sumpthin like that) ;^)

Ok, so the next challenge is to adapt...but what I've found so far is that adaptation is not necessary, patience is. Trying to adapt to Google's upsets can be foolhardy and throw your conversions when things return to "normal". I've studied my users behavior to the point of nearly attaining the Holy Grail of 2% conversion (when traffic is of normal quality).

Keep in mind, I'm not speaking for everyone out there. Myself and a small group of sites are apparently the only ones to see this phenomenon. I'm not sure why I share so much of this info, because many look down on us like we've seen a UFO, are idiots or are off our rockers....
...or a combination of all three.

The more I look at the Zombie problem, the more I am convinced that it's more related to system loading during major algo updates. Traffic shaping? That's another post MayDay problem to be investigated.

[IanTurner]

My few words on Zombie traffic - if Google sees bounce rate as a quality signal for a site - a simple negative SEO tactic would be to worsen your competitors bounce rate by throwing them zombie traffic.

If Google is also doing traffic shaping then another good negative SEO tactic would be to send them zombie traffic at the high conversion times during the day.

@netmeg it is relatively simple to create a bot that looks like a regular visitor and that sends referrers - you can also create a bot that will follow links on a page.

In order to employ the negative SEO tactics outlined above - of course you would need a bot net and delivery method to the zombie PCs so that would make those tactics the prerogative of sophisticated hackers.

[bluntforce]

@Ian
While I'm still digesting what has been reported, I don't see Netmeg's issue as a defined negative SEO tactic. Could turn into that, but for now, it appears to be a waste of resources.

I just limited user agents, while I certainly block a few valid users a day, hundreds of bots also get blocked. As long as the majority of users have a good experience, I've done my job.

[backdraft7]

Just another observation - The past 6 days has been a "normal" stretch of decent quality traffic and regularly spaced sales conversions. Now, Saturday, which is usually pretty busy and converts well is seeing lots of foreign and non targeted, poor converting traffic. The Zombies are back.

[realmaverick]

If the phenomenon of zombie traffic exists, there would have to be either a) an increase in traffic or b) a change in SERPS. Logically at least.

One site I manage for an electrician, his traffic levels are consistant. He gets several calls a day but then randomly, he won't get a single call for several days. Traffic levels are normal, as are SERPs and keywords. I have no explanation for why this happens. It's not related to days of the week, weather nor anything else I can pinpoint.

Those who are experiencing the zombie traffic, would traffic for that day not be higher? and while conversion rates down, are overall conversions not the same?

It's hard to make sense of, if there are no visual changes in the SERPs.

[backdraft7]

mav - now you're starting to understand. No changes in serps, no changes in traffic volume, just sudden drops in conversions, as if the visitors are Zombies, not buying humans. Of course I am not implying there are actual Zombies surfing the web, but for some reason the traffic seems extremely non targeted.

I was talking to a group of people at a supper club last night and they asked what I do for a living, I tried explaining it to them, which led to my trying to explain Google update upsets. These non experts instantly knew about periods when Google delivers some real garbage results and they are unable to find what they were looking for or useful sites suddenly disappear. It's during these periods that we WM's experience Zombie traffic.

Again, analysis is great, and you can crunch numbers until you're blue in the face trying to figure out the cause, but we will never know for sure. If Google is to earn any "user love" they need to soften the ill effects of their "quality updates". WebmasterWorld doesn't get much feedback from normal users, so last night was rather enlightening for me and backs up my suspicion that updates must unleash a certain level of chaos on their system.

The updates seem to be more of a brute force slash & burn for many mid level sites. The top 100,000 may not be affected (so you SEO pro's aren't seeing them), but beyond that all heck breaks loose.

Google wouldn't be much of a search engine with just 100,000 sites. That's why I believe they owe it to webmasters and uses alike to soften the blow in their continual efforts to clean house. Perhaps that would prompt everyone to start loving Google again. Everyone but the black hatters, which is fine.

[realmaverick]

Odd you should say that backdraft, I had a very similar conversation last night with my mum. I told her about the above the fold update and she commented "that it didn't bloody well work then did it". haha

Whether it exists or not, I doubt the top 100,000 would be exempt. Although it's likely they wouldn't notice it, as their volumes are much higher. But ultimately, who knows.

[backdraft7]

another update - after a decent Saturday and a fantastic Sunday, Monday brings another round of Zombies. Not one conversion between 6 am and 4 pm. Yesterday it was sales every hour and multiple sales per hour.
Today, it's foreign traffic galore. Looks like normal apache volume. Mondays are usually rockin'

It's also happening to most of my other webmaster contacts today.

[petehall]

This may be old news to everyone else, but I just ran a couple of tests and an image preview referral URL from Google is exactly the same as if I click the link.

So could it be possible the zombies never make it to your website... they simply look at the preview and move on down the list.

Foreign / junk traffic referrals could be image preview on the likes of Google HK, which we all know ranks some UK site quite highly (often higher than on Google UK!).

Apologies if this has already been covered, but I could not see any mention of it.

[backdraft7]

update - another big drop in traffic quality (but not volume) on Sunday 4/22 and Monday 4/23. Record low conversions Sunday. No sales all day Monday then at 11pm , 2am and 5 am consecutive sales from Australia. My site is set to target US traffic only. Funny how a 20M market is producing more than a 330M market.
SERP positions look fine, but once the switch flips, it doesn't matter. For sale sign will be going up in the front yard soon.

[Robert Charlton]

No sales all day Monday then at 11pm , 2am and 5 am consecutive sales from Australia. No sales all day Monday then at 11pm , 2am and 5 am consecutive sales from Australia. My site is set to target US traffic only. Funny how a 20M market is producing more than a 330M market.

backdraft7 - As soon as I saw news of the update, I kind of expected to see you here. Thanks for the report. I'm sorry it keeps hitting you. I've been wanting to jump into this thread, just as I'd been wanting to return to a discussion I started almost a year ago, but was unable to follow up....

Are we entering another Panda user-engagement calibration phase?
http://www.webmasterworld.com/google/4314423.htm [webmasterworld.com]

Among other things, I've wanted to thank Shaddows for his key posts in that thread as well for his 2010 Traffic Shaping thread, which tedster linked to above.

I've since come to think that traffic shaping isn't just about Panda, and have been thinking for a while that it had to do with all Google testing, much of which I've assumed is about localization.

Please forgive this very hasty post, but I'm seeing something tonight that bears on your post above, and I wanted to hook up various threads. Note this discussion about geo-confusion (or is it testing?) between the US and Australia in the May SERPs thread...

Google Updates and SERP Changes - May 2012
http://www.webmasterworld.com/google/4447978.htm [webmasterworld.com]