Posted by Moberstein
The Discover Feed from Google completely fascinates me. Aside from the fact that it provides highly relevant content, it also appears to be out of the reach of a game. In a way, it seems almost out of the reach of pure SEO (which makes it downright tempting to me).
It all got me to understand what was making the feed tick.
So I did what any sane person would do. I've spent almost two months doing all sorts of queries in a variety of ways to see how this affects my Discover feed.
Here are my forays.
My approach to analyzing Google's Discover feed
Let me explain what I did and how I did it to give you a better understanding of this analysis and to point out its gaping limits.
I performed all kinds of user behavior five days a week over a two-month period to influence my Discover feed.
I ran queries on specific topics on mobile devices, searched for other topics on the desktop … clicked results … didn't click results … went straight to websites and clicked … went straight to websites and didn't click anything, etc.
In other words, I wanted to see how Google would react to my various behaviors. I wanted to see if one behavior affects what is shown in my Discover Feed more than other behaviors.
To do this, I looked for things that I would normally never look for, went to websites I would normally never visit, and at times limited my normal search behavior so as not to affect the feed.
For example, I hate celebrity news and passionate gossip, so every day (outside of the weekends) I went to people.com and scrolled the website without clicking anything. I then recorded if related material (i.e., celebrity nonsense) ended up on my Discover feed the next day.
I recorded all of my various "web behaviors" in the same way. I would perform a certain behavior (like looking for things on the phone for things related to a particular topic but without clicking results) and recording what happened in my Discover feed over time.
Here is a breakdown of the various behaviors I have performed, as well as the issues associated with each behavior. (For the purposes of recording, each behavior corresponds to a single topic or website, so I can determine the impact of that behavior on my Discover feed.)
Let me briefly cover the behaviors mentioned above:
- These are all topics / websites that I am not interested in or involved in in any way (especially self-help car repair).
- When I clicked on a YouTube video, I would watch the whole thing (I mean, I haven't seen half the time, but Google doesn't know … or does it?)
- Whenever I visited a site I would scroll through the content and stay on the page a bit.
- A search for a "segment of a topic that's already in Discover Feed" means that the whole topic has been showing up regularly in my feed (in this case, NFL football and MLB baseball). However, the subtopics, in this case the cowboys and marlins, were topics that I never specifically looked for and that weren't included in my feed. Also, the data for these two categories only reflect one month of experimentation.
All of this indicates various limitations.
Is it possible that Google sees a topic like entertainment news as “more worth discovering” than sewing? It is.
Is it possible that visiting a website like Fandango during a pandemic (when many theaters were closed) influenced Google's decision to include or exclude things related to the topic the site was covering? It is.
What if I didn't skip the weekends and do the above tasks every day? Would that have made a difference? I dont know.
I am not trying to present the data I am presenting as overly conclusive. This is just what I did, what I found and everything that got me thinking.
Then let's do it.
How user behavior affected my Discover Feed
Before I get into the "data," I want to point out that the heart of my observations is not in the data itself, but in some of the things that caught my eye on my Discover Feed along the way.
In addition, this data is anything but conclusive or concrete and in most cases speaks for my unique user profile. Let's look at the data as there may be some general insight.
As mentioned earlier, I wanted to see the impact of the various online behaviors on my Discover feed. How often has Google added content to my feed on the topics associated with each behavior?
For all the times I went to japantimes.co.jp, how many times did my feed contain content related to Japanese news? In all the times I've searched for and watched lawn care YouTube videos, how many times has Google shown me such content in Discover?
Here are some of the most noticeable highlights, reflected in the graphic above:
- Watching YouTube videos on mobile did not affect my feed in any way (although it certainly did affect the YouTube ads that were shown to me).
- Watching YouTube videos on the desktop has little effect (in fact, adding “sew” to my feed was just a topic card that didn't have a URL).
- Searching Google alone without clicking on a result was ineffective.
- Visiting a desktop site and clicking was very effective in filling my feed with "cook" content. However, this did not apply to mobile devices.
Watching YouTube videos (desktop) about sewing was only successful to get Google to add the topic to their "Discover More" cards.
I want to emphasize that when I say things like "YouTube Mobile Clocks had no effect" I am not saying that as a general statement. Rather, such a statement only coincides with the way I looked at YouTube (one video playback per day). Of course, if you're watching a large number of YouTube videos on a topic in a short amount of time, Discover will pick it up.
I actually tested that.
I gave free hands to my kids at different times to grab my phone and watch a large number of videos on specific topics (surprisingly, they were happy to sign up and watch huge amounts of YouTube).
I have two 9 year old boys. You watched an obscene number of YouTube videos and performed an insane number of searches related to airplanes and flight simulators. I'm still waiting for the day when my feed will stop showing cards on this topic. Here is my search history to prove it:
The other little guy spent a few hours watching videos about the weather and animal behavior that resulted from it (hey, it was during the height of the quarantine). On the same day, I saw the following on my feed:
Needless to say, if Google thinks you are dealing with a particular topic, it will posthaste that topic on your Discover Feed.
My goal in all of this wasn't to figure out the fastest way to get Google to update the topics displayed in your Discover feed. My methodology was about checking if there was one type of behavior that Google seemed to take more seriously than another when adding new topics to my Discover feed.
Google then reacted differently to my different behaviors.
That doesn't mean I can draw many conclusions based on the above data. For example, Google saw that I go to foodnetwork.com every day and click on an article, a strong signal that “Cooking” deserves to be included in my Discover Feed.
Google was inclined to think of my behavior when visiting foodnetwork.com and clicking an article each day to confirm that I wanted to "cook" content on my Discover feed.
At the same time, Google completely ignored this behavior on mobile devices. Every day I went to japantimes.co.jp and flipped through an article. Still, never once did Google add anything to my feed that was remotely related to Japanese news.
I suspect that the topic here was too far removed from general search behavior. While it was reasonable for Google to assume that I wanted cooking-related material on my feed, the same did not apply to topics related to Japan.
I think this is the same reason why the topic associated with visiting a website on the desktop without clicking anything was added to my feed. The topic here was celebrity news, and I imagine Google rated this topic as one of the most relevant to Discover. So much so that Google tested adding it to my feed in various places.
Although Google never clicked an article when visiting people.com every day, Google still flirted with showing celebrity news content on my Discover feed.
However, there is reason to believe that desktop behavior has a greater impact than the behavior of mobile users.
The case for the dominance of Discover Feed on the desktop
About a month after my little experiment, I wondered what would happen if I started looking for and clicking on things that were segments of topics that had already appeared on my feed.
The decision for these segments was quite easy. My feed is constantly filled with material about baseball and American football. So I decided to look for and click on two teams that I'm not interested in. That way I was able to see the effects of my behavior even though the topic was already in my feed as a whole.
Specifically, I've been looking for things to do with the Dallas Cowboys on the desktop and clicking a search result every time. Likewise, I did the same for the Miami Marlins baseball team on my cell phone.
In both cases, the content specific to those teams didn't have to be shown in my feed yet.
Here are the results of this activity:
Over a 30 day period, I found 10 instances of Dallas Cowboys content and 6 instances of Miami Marlins content in my feed.
Just like the first dataset I presented, there is a difference between mobile and desktop.
Is that a general rule? Is this based on my particular profile? I dont know. It's just an interesting point that needs further investigation.
I'll say that I doubt the content itself played a role. If anything, there should have been more results on the marlins on the phone as I was very much involved in the World Series that was going on at the time of my activity.
What does this data actually mean?
There are so many factors at play that the above data is a little "difficult" to use. Yes, I think there are some trends or indicators. However, I don't want you to take all of this away. (Is it also such a crime to just consume data because it's interesting to see what's going on?)
Then what should I take away from you?
As part of my data analysis (if you call it that at all), I looked at how long it took for a behavior to result in the ingestion of Discover feeds. Surprisingly, the numbers were pretty consistent:
Google discounted the 31 behavioral instances around my "browse desktop without a click" activity (e.g., finding anything related to "fishing" but clicking nothing) that affected my feed and found what I did did pretty quickly.
In general, it took Google fewer than 10 behaviors to believe that the topics displayed on my feed should be updated.
That really is the point. Despite the normal things that I regularly and intensely search for and deal with (e.g. SEO), Google took this "lighter" yet consistent behavior as a signal to update my feed.
Google was very aware of what I was doing and was pretty quick to act when you factor it all out. In the case of "food / cooking" content, as previously shown, Google took my behavior very seriously and consistently displayed such content in my feed.
Forget which behavior generated more cards in my feed on which device. The fact that it was different at all is telling. It shows that Google is researching the type of engagement and location related to your overall profile.
Personally, I think if you (yes, you, the person reading this) did this experiment, you would get different results. Maybe some of the trends match, but I'd imagine it is.
And now for the really interesting part of all of this.
Dive into what was and wasn't in my Discover feed
As mentioned earlier, the data is interesting for some of the possible trends it alludes to and shows how closely Google is monitoring your behavior. The most interesting facets of this little project of mine, however, were seeing what Google was showing and what was not showing day in and day out.
Are Google Profiling Users Using the Same Account?
The first month of this study coincided with a lockdown due to COVID-19. That meant my children were home all day for a month. It also meant they watched a lot of YouTube. From Wild Kratts to Woody Woodpecker, my kids consumed a bunch of cartoons and did so through my Google account (so I could see what they saw).
Wouldn't you know a funny thing happened? There was no "cartoon" content in my Discover feed. That month I checked my feed religiously and never once saw a card about a cartoon.
Isn't that strange?
Not if Google is profiling my account based on the devices used or even the content used. All indications are that Google is aware that the content my kids saw was not consumed by the Discover (me) user.
This is not a route at all. The same thing happens all the time on my YouTube feed. While my desktop feed is filled to the brim with Fireman Sam, the YouTube app on my phone is a mix of news and sports (I don't do “SEO” on YouTube) as my kids generally don't watch their “programs” my phone.
The URLs I was visiting were missing from Discover
Something was missing from my Discover Feed, and this is having a huge impact.
Virtually none of the URLs I visited during my two month experiment showed up in my Discover Feed!
I've visited the Food Network website about 40 times, clicking and reading an article or recipe each time (pretending it's fair). As I neared the end of my experiment, Discover showed me some types of food / cooking-themed content every day.
Not once did Google show me a URL from the Food Network! Do you like apples? Well how do you like these apples? (Cooked slowly with a hint of cinnamon.)
This was the general trend for any type of behavior that brought up new topics on my feed. I visited a couple of websites about car repairs, Google threw some cards on the subject in my feed … none of them were a website I visited.
The only time I saw the same website that I visited and that showed up in my Discover Feed was ESPN for some of the sports queries I did and people.com that I visited every day. However, I think this was completely random as both of these sites are top sources of content in their rooms.
Yes, some websites that I visit regularly will generally show up on my feed. For example, there were some local news sites that I visited several times a day for almost a month following COVID-19 in my area. I freely admit that it was a compulsion. One that Google picked up.
In other words, it took Google a ton of time to believe that I wanted that particular website or sites on my feed. Additionally, Google doesn't seem to want to just display content from the URLs you visit unless the signal is otherwise immense.
This leads me to my next question …
Is Discover Really an SEO Problem?
What can you do to optimize for Google Discover? It's almost an absurd question. I visited the same website every day and Google still didn't include the url in my feed. (I am again aware that certain behaviors trigger a certain URL. My point is that Google is not as capable of doing this as you might think.) Everything points to a certain lack of control. Everything indicates that Google does not want to pigeonhole the content displayed in Discover.
In other words, you can't create content specifically for Discover. There is no such concept. There is no such control. There are no standardized “ranking signals” for which you can try to optimize.
Optimizing your images to make sure they are high quality or at least 1,200 pixels wide is not really "optimized" for Discover. It just entitles you to get into the ballpark. There is no standardized way to actually get into the field.
The whole idea of Discover is to offer content that is specifically relevant to a user and all of their various specifics. The term “optimize” for something like that is almost impossible to calculate.
As with optimizing your images for Discover, all you can do is position yourself.
And how do you position yourself for inclusion in the Discover Feed?
One of the sites that kept popping up on my feed was dallascowboys.com. This makes sense as I've been looking for things to do with the Dallas Cowboys and have clicked on all possible outcomes as a result. However, on my “travels” I did not visit dallascowboys.com. However, when Google saw that I was interested in the Cowboys, this was one of the first websites I was served.
You don't have to be a rocket scientist to understand why. What other website is more relevant and authoritative than the franchise's official website?
If you want your website to be included in Discover, you have to be incredibly relevant and authoritative whatever your website is about.
That means investing time and resources into creating unique and substantial content. It means developing a complete strategy for creating a current identity. After all, you want Google to understand that your website is about, digging into, and dealing with a certain topic a lot (i.e. the topic is closely related to who you are as a website).
That sounds a lot more like "content marketing" than pure SEO, at least to me.
A cross-disciplinary marketing spirit merges
For me, Discover is the figurehead for the fusion of pure content creation and SEO. It speaks in favor of the idea of needing a more abstract understanding of the strategy for in-depth content in order to be effective in the "Google verse".
It is perhaps a different type of movement than what you normally find in the world of pure search engine optimization. As opposed to delving into the smallest details (be it a specific technical problem or some aspect of content optimization), Discover urges us to take a more holistic approach and take a step back.
The way Discover is built advocates a broader approach based on meta-analysis of how a website is perceived by Google and what can be done to build a stronger profile. It's almost the perfect mix of content, marketing, and an understanding of how Google (SEO) works.
Sign up for The Moz Top 10, a bi-monthly mailer that keeps you informed of the ten hottest SEO news, tips, and radical links discovered by the Moz team. Think of this as your exclusive roundup of things you don't have time for but want to read!