Human -> Machine -> Human: Ruth Burr Reedy MozCon 2019

Editor's Notes

1. The discourse.
2. SEOs talk a lot about following algo updates...
3. ...and whether paying attention to the QR Guidelines is worthwhile
4. I don’t find this conversation very interesting TBH! But I get it – SEOs are busy, and it’s really frustrating to see advice being put out there that you know isn’t going to help people do SEO. It hurts to watch busy people waste their time. I think what both these conversations are missing is where these activities might fit in to a larger framework or approach to SEO. How do we know where to spend our time?
5. People in the “ignore that” camp often focus on Real Company Shit
6. I don’t disagree with this at all - but it’s entirely possible to do these things in a way that does not create SEO success, which is a wasted opportunity. RCS is not an excuse not to learn SEO fundamentals or technical SEO.
7. It’s both! We can’t optimize for just humans or just Google. It has to be both.
8. My personal framework for approaching changes in our industry is: we are humans, creating things for humans, with a machine serving as the intermediary between the two.
9. How do you “just create good content”? You figure out what information people need to complete their task and provide it. Where do you get that information? User testing is useful, but expensive – and you don’t always have access to your clients’ users.
10. Google has more data than anyone on how people use search engines to find information
11. They also have more data than anyone on how people talk about things, i.e., entities and topics, using written language.
12. Google wants humans to find what they're looking for when they search, which means Google spends a lot of time trying to understand what people are looking for when they search.
13. There is Google the COMPANY: Humans making and communicating decisions about what’s important to the company
14. And there is Google the MACHINE: algorithms on top of algorithms
15. SEO tactics often come down to optimizing algorithmic inputs and tracking algorithmic outputs - hence the constant discussion over whether or not something is a “ranking factor.”
16. But Google is focused on results - do people find what they want? The SERP is only a means to that end. We should be looking at Google’s outputs not as an end in and of themselves, but as information of how to achieve those results.
17. One way to understand Google is to leverage Google’s huge body of user data to understand the human-readable quality signals that Google (as both a company and an algorithm) have deduced are important, and then to create that quality, while optimizing machine-readable equivalents or proxies of those signals.
18. The question then becomes: how can I demonstrate quality in a machine-readable way?
19. A human might ask, “What is this page about?”
20. How does a robot answer “What is this page about?” -> Keyword use; synonyms, stems, and close variants; topic modeling, TF-IDF (need to add a graphic of a robot here, looking at the factors, or otherwise convey the concept)
21. “Is this page a good example of a page about this thing?” (need to add a graphic of a human person here, having the thought, or otherwise convey the concept)
22. “Is this page a good example of a page about this thing?” (need to add a graphic of a robot here, looking at the factors, or otherwise convey the concept)
23. Natural language processing and neural matching
24. This is great because humans are terrible at creating content that’s about what they think it’s about!
25. Fortunately, optimizing for NLP has a lot in common with optimizing for readability.
26. Natural language processing
27. Natural language processing
28. “Can this site be trusted?” (need to add a graphic of a human person here, having the thought, or otherwise convey the concept)
29. “Can this site be trusted?” - Trust factors, HTTPS, author entity authority (need to add a graphic of a robot here, looking at the factors, or otherwise convey the concept)
30. Google’s evaluation of link signals is already very similar to how humans decide if information is credible
31. “Is this site easy to use?” (need to add a graphic of a human person here, having the thought, or otherwise convey the concept)
32. “Is this site easy to use?” - Page speed, crawlability, UX factors (need to add a graphic of a robot here, looking at the factors, or otherwise convey the concept)
33. What’s interesting here is that we can see Google trying to evolve its understanding of page load time to more closely mimic the way a human would experience it - this is why metrics like time to interactive have become important alongside total PLT
34. So now that we’ve reminded ourselves of what we’re trying to do, let’s revisit those two SEO battlegrounds we talked about earlier. The QR Guidelines: Do not give insight into algorithmic inputs - DO give insight into the quality signals Google is intending to send - useful through that lens
35. Used for two things: to TEST and to TRAIN
36. E-A-T
37. Expertise, authoritativeness, and trust are not quantifiable qualities that an entity either does or does not have (need to add graphics here of a human and a robot having this same conversation)
38. These are human-readable ways of thinking about the question “how do we tell if someone knows what they’re talking about?”
39. Marie is going to talk more about this later and you should listen to her
40. What, of those signals, is machine-readable?
41. How might a machine find the same information in another way?
42. What, of those signals, can we boost?
43. What, of those signals, is easiest to game/fake, and therefore likely to be weighted more lightly?
44. People are constantly trying to “game the system” - the challenge for search engines is to create a sophisticated enough set of machine-readable inputs that will separate the wheat from the chaff
45. The easier a signal is to game, the less value it’s going to pass on its own - it’s likely that certain signals around e.g. keyword relevance are only valuable in context with quality signals.
46. So take another look through the QR guidelines and ask –what would a machine do?
47. If you follow one person w/r/t QRGs make it Jen Slegg
48. What does it mean to “chase the algorithm”?
49. The end result should not be optimizing for inputs (keywords in title tags are now 0.82% more important). If you do that, you’re focusing on the wrong information. But that doesn’t mean that there’s nothing to be learned from algo updates.
50. You shouldn’t be solely reactive to algorithmic changes, but when they happen, it’s worth trying to understand them to validate that you’re on the right track
51. With that in mind, here’s what I look at in the wake of an algorithmic update:
52. Information is usually mostly useful: At scale Within your niche Talk about recent update re: health sites
53. How, if at all, did this impact the sites I work on?
54. Across the board? Check SEO Twitter, WebmasterWorld Forums. What do other impacted sites have in common with mine?
55. Most of the time it’s not that you lost rankings - it’s that someone else gained them. What does the SERP say? It is not always possible to regain rankings or traffic for a keyword you’ve lost Sometimes it’s not a matter of a penalty or of something that was working, no longer working
56. Sometimes it’s that the kind of page, or the kind of information, that Google has determined users are looking for has changed E.g. “best whatever” almost always returns aggregators over individual whatever providers
57. Understanding the end goal (not what, but why?) of a given algorithmic pattern or change can give you insight into how your target audiences are using search to make decisions
58. Ask yourself: What do these things say about what my customers are looking for?
59. How can I provide those things… ...in ways that Google can see and understand?
60. There's a compelling argument to be made that when an algorithm is focused on improving Google's results quality in a certain respect or fixing an existing flaw, it's almost certainly not the last time they'll try to improve there. An algo update around content quality probably means they're going to continue iterating on that.
61. Like with health sites, it would have behooved anyone to start really ramping up the quality signals on their sites after the March update
62. Does Google look at (X Thing) as a ranking factor? If it’s a machine-readable proxy to a human-readable quality signal, they’ve probably at least tried it If it’s easy to game, they may still be working out how to make it less gameable You should want high organic CTR and good dwell time because they’re a signal that you’re providing humans with what they want
63. That’s the future.
64. So now that we’ve reminded ourselves of what we’re trying to do, let’s revisit those two SEO battlegrounds we talked about earlier.
65. Does Google look at (X Thing) as a ranking factor? If it’s a machine-readable proxy to a human-readable quality signal, they’ve probably at least tried it If it’s easy to game, they may still be working out how to make it less gameable You should want high organic CTR and good dwell time because they’re a signal that you’re providing humans with what they want
66. Does Google look at (X Thing) as a ranking factor? If it’s a machine-readable proxy to a human-readable quality signal, they’ve probably at least tried it If it’s easy to game, they may still be working out how to make it less gameable You should want high organic CTR and good dwell time because they’re a signal that you’re providing humans with what they want
67. Does Google look at (X Thing) as a ranking factor? If it’s a machine-readable proxy to a human-readable quality signal, they’ve probably at least tried it If it’s easy to game, they may still be working out how to make it less gameable You should want high organic CTR and good dwell time because they’re a signal that you’re providing humans with what they want
68. Does Google look at (X Thing) as a ranking factor? If it’s a machine-readable proxy to a human-readable quality signal, they’ve probably at least tried it If it’s easy to game, they may still be working out how to make it less gameable You should want high organic CTR and good dwell time because they’re a signal that you’re providing humans with what they want
69. Does Google look at (X Thing) as a ranking factor? If it’s a machine-readable proxy to a human-readable quality signal, they’ve probably at least tried it If it’s easy to game, they may still be working out how to make it less gameable You should want high organic CTR and good dwell time because they’re a signal that you’re providing humans with what they want
70. Does Google look at (X Thing) as a ranking factor? If it’s a machine-readable proxy to a human-readable quality signal, they’ve probably at least tried it If it’s easy to game, they may still be working out how to make it less gameable You should want high organic CTR and good dwell time because they’re a signal that you’re providing humans with what they want
71. Does Google look at (X Thing) as a ranking factor? If it’s a machine-readable proxy to a human-readable quality signal, they’ve probably at least tried it If it’s easy to game, they may still be working out how to make it less gameable You should want high organic CTR and good dwell time because they’re a signal that you’re providing humans with what they want
72. Does Google look at (X Thing) as a ranking factor? If it’s a machine-readable proxy to a human-readable quality signal, they’ve probably at least tried it If it’s easy to game, they may still be working out how to make it less gameable You should want high organic CTR and good dwell time because they’re a signal that you’re providing humans with what they want