Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

You, Google and Links: It's complicated - Paul Madden - SearchLove Boston 2017

3,073 views

Published on

At SearchLove Boston 2017 we discussed how links have become a complicated and compromised signal and what we can do to make sure we stay safe and get the most from our links.

Published in: Marketing
  • Transcript for THEIRNAME, Your transcript is expiring! (accept here) ❤❤❤ https://url.cn/5yLnA6L
       Reply 
    Are you sure you want to  Yes  No
    Your message goes here
  • Hello! Get Your Professional Job-Winning Resume Here - Check our website! https://vk.cc/818RFv
       Reply 
    Are you sure you want to  Yes  No
    Your message goes here

You, Google and Links: It's complicated - Paul Madden - SearchLove Boston 2017

  1. 1. You, Google and Links: It's complicated
  2. 2. Who dis ?
  3. 3. • 1997 made first website by typing in HTML 3.2 from a fully written out page of pencil planning • 2006-9 Swapped web design to become a professional Adsense spammer earning £30,000 a month for a while on 2 hours work a week • 2010 Reached banned account number 7 and gave up spam to start a link building agency • Grew the link building agency to 200 offshore staff • 2012 Link penalties arrived and repenting began (loses even more of my hair) • 2012 Founded Kerboo (as LinkRisk back then) • 2017 Now finds himself talking to a room of clever people about links
  4. 4. The debate
  5. 5. Today I would like to debate with you the answer to that persistent SEO question
  6. 6. Are links even worth thinking about anymore?
  7. 7. Yes
  8. 8. But as the talk title suggests …
  9. 9. It’s complicated
  10. 10. My position is that links still have a role to play in SEO success
  11. 11. That they now can get drowned out some terms by other signals and by Google
  12. 12. But that without paying them some attention you put your content at a severe disadvantage from the start
  13. 13. How come I feel well positioned to help with this debate?
  14. 14. I’m an SEO
  15. 15. SEO entry qualifications 101
  16. 16. Do you have a laptop?
  17. 17. Do you have an opinion?
  18. 18. You’re IN !
  19. 19. I come from a competitive SEO background* *read into that what you will
  20. 20. I had a large link building agency and competed in many many very competitive niches
  21. 21. Since 2012 I have been on the flipside of that world
  22. 22. Helping found, design and run Kerboo
  23. 23. Where we have since scored over 8 billion links for their LinkRisk
  24. 24. Helped thousands of sites recover from problematic link related issues
  25. 25. Yes in the early days some of those problems were the problems I’d sold them in the first place
  26. 26. As now I sit with this ocean of link data and client experience
  27. 27. And the biggest disavow database outside of Google
  28. 28. Fine but it’s 2017 Paul….
  29. 29. The point I am making is: -
  30. 30. Whilst I am no data scientist
  31. 31. I’m a doer With experience, perspective and enough test data to drown a hippo
  32. 32. Since the middle of last year I have been working on trying to wrestle with this topic and how to define or show the value of a link
  33. 33. The state of links today
  34. 34. It’s 2017
  35. 35. “When I get links sometimes they don’t seem to make much of a difference”
  36. 36. But the presence of links seems to be a persistent factor in ranking visibility
  37. 37. What we have to decide is:- “Is that just historical fact or do they still play a part in why pages perform?”
  38. 38. Why even care?
  39. 39. There was a time when links dominated the marketing world
  40. 40. How links became the primary signal
  41. 41. Teaching grandma to suck eggs
  42. 42. The original mechanics of links as a signal
  43. 43. We started with Backrub PR(A) = (1-d) + d (PR(T1)/C(T1) + ... + PR(Tn)/C(Tn))
  44. 44. “OMG this fat bloke just rolled out the PageRank algo at SearchLove “
  45. 45. We seem to have forgotten that when we think about links
  46. 46. The basic mechanics still apply “total equity into a page / total number of links out”
  47. 47. If we assume that equity still flows in the same basic way
  48. 48. Then raw link equity alone is too blunt a signal
  49. 49. Citation flow follows that core logic for the flow of equity across the link graph
  50. 50. Trust flow is a modified version of Citation flow to take into consideration the proximity of known trust
  51. 51. That’s an important point
  52. 52. Google quickly moved from the raw PR number to using that as a baseline and modifying their opinion based on other signals
  53. 53. For a long time those signals were drowned out by the reliance on the raw PageRank
  54. 54. Because PageRank took the place of human opinion
  55. 55. But we ruined the signal
  56. 56. Links and link schemes became the way we manipulated Google
  57. 57. We compromised the signal
  58. 58. Google couldn’t rely on it alone
  59. 59. It has taken them a long time to reduce their reliance on links as a signal
  60. 60. Because it was so important to them
  61. 61. In longer tail Google really only had our influence on the signals to make sense of the world
  62. 62. Nobody ever naturally linked to a page about ladders
  63. 63. And Google has always had some ways to make sense of the world without links
  64. 64. And there are whole niches that don’t rely on links
  65. 65. p0rn
  66. 66. They introduced LOTS of ways to stop and negate manipulation
  67. 67. Penguin Payday Pigeon Top Heavy Pirate Panda Mobile friendly
  68. 68. Google have looked to find cleaner signals to help
  69. 69. Signal diversification
  70. 70. Not only have they massively expanded and developed what they take into consideration
  71. 71. They have much more human opinion for real now
  72. 72. Search quality raters
  73. 73. And that human element *In combination with other signals Has helped Google reduce their reliance on links
  74. 74. And at the same time the way search is delivered has changed
  75. 75. The rise of the machines
  76. 76. Feb 2016
  77. 77. No ill informed opinions on AI were given by the fat bald bloke on stage I’m in marketing….
  78. 78. The persistence of the human layer
  79. 79. Prepare to learn how boring my life is…
  80. 80. Company accounts 2015
  81. 81. Who ?
  82. 82. The quality raters… • Lionbridge 2015 accounts • 100,000 outsourced professionals • Referred to as ‘geoworkerz’ • 12% of the total revenue was on the Google contract
  83. 83. An ongoing relationship • In 2014 there were an estimated 40,000 across 4 outsourcing companies doing search or ads quality review • No sign of changes in the volume of the contract
  84. 84. The role of raters is really two fold
  85. 85. They qualify at page level (PQ) that a page achieves what it says it does
  86. 86. They qualify the quality of the SERP delivered for a phrase
  87. 87. As the rating process starts with a SERP
  88. 88. Whilst the quality raters have no influence on links as a signal
  89. 89. They remain part of the way Google calibrate and train the machines
  90. 90. • Links are a compromised signal • Links are less influential when taken in isolation • Google have more reliable replacements for the human opinion of a page • The AI now rules how the actual SERPs are delivered
  91. 91. Links no longer dominate as they once did…
  92. 92. Given the link signal is unreliable what do we still believe?
  93. 93. First principles
  94. 94. It is clear that…
  95. 95. Links play a role in ranking
  96. 96. In longer tail they play a much larger role than competitive or brand terms
  97. 97. Google has moved away from a reliance on them as the most important signal
  98. 98. They provide relevance and qualification for ranking by the machines
  99. 99. And that the machines now take a much larger role in the actual delivery of results
  100. 100. Opinion based on observation
  101. 101. Links play a role in the judgment of your sites trust
  102. 102. Having legacy links in your profile that demonstrate intent to manipulate are your LinkDebt
  103. 103. Your LinkDebt contributes to the success and the failure of your organic visibility
  104. 104. Links play a role in determining relevance
  105. 105. Links qualify that a site is worthy of further consideration during the ranking journey
  106. 106. So if they can’t rely on links alone are they relying on an overall trust and authority metric?
  107. 107. “we don't have anything like a website authority score.”
  108. 108. If links don’t have a boost from an overall authority score for the domain…
  109. 109. Why doesn’t a raft of links from bloggers beat one great link from a trusted source?
  110. 110. Well in general it seems clear that equity alone is not the only factor
  111. 111. Its not about ‘natural’
  112. 112. Its about Google’s PropensityToTrust
  113. 113. Perceived intent is key
  114. 114. Its clear that there is trust involved but that trust comes from the overall mix of signals rather than just one
  115. 115. The types of links we should be looking for…
  116. 116. • We are looking for links that do not demonstrate an ‘intent to manipulate’ • From sites that are likely to be trusted • And which have actual equity to pass us • And which will maintain that equity and status in the future
  117. 117. What proof do we have that links still count?
  118. 118. We track newly discovered links DAILY for tens of thousands of sites
  119. 119. We crawl them and we score them
  120. 120. An increase in links does still have a positive impact on the visibility of lower terms
  121. 121. In head terms it doesn’t seem to have the same impact unless its in the presence of a more complete marketing effort
  122. 122. And …
  123. 123. Not caring about your legacy link issues still plays a role in suppressing success overall
  124. 124. But link penalties aren’t a thing now?
  125. 125. Sorry ‘bout dis
  126. 126. I ran some of the stats from Kerboo on manual actions (link penalties) applied
  127. 127. I tried to remove the actions against anything other than brand* in our penalty figures *I tried to remove the general spam
  128. 128. Its also true to say that many people doing cleanup even now are seeing a benefit above the insurance value alone from doing so
  129. 129. And there is evidence everywhere that links still count to some degree
  130. 130. Evidence that links count towards visibility
  131. 131. Reavow tests
  132. 132. There are very very few times we can turn signals on and off in SEO
  133. 133. But this is a unique opportunity
  134. 134. Kev Gibbons
  135. 135. https://www.blueglass.co.uk/blog/rem oving-disavow-file-seo-test/
  136. 136. The conclusion here is that removing the disavow turned links back on and rankings improved
  137. 137. Links in isolation improved visibility*
  138. 138. * Plus he increased the risk again And possibly introduced a bunch of red flags and things that could impact rankings in future
  139. 139. That’s just one test though… proves nothing Paul
  140. 140. Sigh… Let me share with you some examples of that AT SCALE
  141. 141. We have done >25 of these projects to date
  142. 142. They have all been positive for the site in question
  143. 143. Reavow as an SEO tactic • 45 million links + in the profile • 28,000 entries in a disavow file • Previously penalised domain mostly recovered now • We reviewed the disavow activity and chose to reavow some
  144. 144. What and how we reavowed • We worked in sections • We initially qualified based on link metrics • We then manually reviewed based on INTENT
  145. 145. We also tried to maintain a net gain in total disavow row count
  146. 146. For me that is clear evidence of the fact that links do contribute to improved visibility still
  147. 147. It wasn’t brand search changes It wasn’t other obvious marketing activity
  148. 148. It was links…
  149. 149. Reavow is an important thing to consider
  150. 150. But its massively dangerous if managed badly
  151. 151. And yes I know ‘reavow’ isn’t really the right word
  152. 152. Factors affecting the ability of links to pass
  153. 153. Given that links do still influence visibility what might prevent them doing so ?
  154. 154. Disavow weight Reason 1
  155. 155. Disavow activity could be used to qualify a link for further action or attention
  156. 156. But only crap links get disavowed ?
  157. 157. All disavowed in Kerboo’s data more than 100x each
  158. 158. At Kerboo we use PropensityToDisavow
  159. 159. Its logical that the same data could provide Google with signals of trust at domain level
  160. 160. Negation through the action of Penguin* *no penguin picture slide as we have all seen enough of those damn birds Reason 2
  161. 161. For the record… I STRONGLY believe we have underestimated the impact of Penguin real time so far
  162. 162. Penguin is a complex beast now
  163. 163. Previously….
  164. 164. Penguin was a flag applied to the domain
  165. 165. Since Penguin 4 …
  166. 166. Penguin 4 is real-time • Penguin 4 is now part of the core ranking algorithm • That means it runs in ‘real-time’
  167. 167. Penguin 4 is more granular • Rather than site level • Penguin 4 operates at page level
  168. 168. Penguin 4 removes the ability of problematic links to pass equity
  169. 169. By removing the equity from those problematic links the ranking ability of the page is reduced
  170. 170. That’s GOOD right ?
  171. 171. It means things are harder to understand
  172. 172. So why haven’t we seen a major impact?
  173. 173. To explain that we have to work through the mechanics of how Google apply this
  174. 174. Real-time is defined in an interesting way
  175. 175. Real-time for you and I means: -
  176. 176. As soon as something changes its implemented there and then
  177. 177. That’s NOT how Google uses the term real-time
  178. 178. Real-time for Google means: - “Once we crawl the page again”
  179. 179. How Penguin is applied
  180. 180. Each time “Page A” is crawled
  181. 181. Its Penguin status is apparently recalculated
  182. 182. BUT real-time Penguin operates by removing the ability of spammy links to pass equity
  183. 183. So in reality…
  184. 184. Real-time Penguin relies on the signals from the linking sites on whether they are spammy or not
  185. 185. And that status is only updated for THEM when they are crawled
  186. 186. WHAT ????
  187. 187. Lets look at a pretend example
  188. 188. • The linking domains for Page A each have a status on whether they pass equity or not. • Determined by Penguin or another algorithm. • But only updated when they are crawled (like all Google signals are)
  189. 189. • Over time this status is updated as the linking pages are recrawled • This can take some time
  190. 190. • Over time it may well be the case that more and more of a link profile gets negated by Penguin
  191. 191. Page A will only feel Penguin as the links pointing to it are updated
  192. 192. Those links are assessed and updated as they are recrawled by Google
  193. 193. That process can take MONTHS to fully apply
  194. 194. Which means …
  195. 195. Changes to the Penguin algorithm can take a while to implement across the link graph
  196. 196. Google Penguin changes will only be felt slowly
  197. 197. As that link graph is slowly updated
  198. 198. “But surely the link graph is always in a state of update so it shouldn’t matter?”
  199. 199. True but we have been made to expect the changes will apply in real- time
  200. 200. Google real-time and our real-time aren’t the same thing…
  201. 201. And its likely that links that lose their ability to pass value will never regain that ability
  202. 202. So it becomes a slow eroding of the total link equity pointed at Page A
  203. 203. As that eroding of the link equity happens the rankings for that page will slowly slip
  204. 204. And to make matters worse…
  205. 205. The fact that it applies once the links are recrawled means any changes we make are hard to attribute to Penguin
  206. 206. A change we make today that worsens a page’s links might not be fully felt for weeks or months
  207. 207. Link profile, crawl dependent issues are felt slowly and are therefore harder to diagnose
  208. 208. Or people assume because it didn’t happen in “real-time” it isn’t really caused by Penguin
  209. 209. Meanwhile Google is slowly eating away at the link profile
  210. 210. “IF Google are just removing the ability of some of my bad links to pass value – surely I can just throw stuff in and hope it sticks for a while?”
  211. 211. No, No you can’t ….
  212. 212. You’d be ruining your trust and opening up a major link debt issue for the future
  213. 213. TLDR: Penguin might be eating your link equity slowly, all you’d see is a slow decline in ranking
  214. 214. Lack of relevancy to the user intent sought by the machines Reason 3
  215. 215. Link relevancy is more important now (topical and co-citation) than ever before
  216. 216. Crazy SEO experiment warning
  217. 217. • I took a number of sites I control • I went through the link profile for relevance NOT intent or quality • I disavowed those links that didn’t contribute to relevancy
  218. 218. Guess what happened?
  219. 219. The target pages ranking popped
  220. 220. A summary of the link picture in 2017
  221. 221. It’s clear that links haven’t gone away but…
  222. 222. Their role has changed somewhat
  223. 223. Links qualify your trust
  224. 224. And in many cases they still exert their influence from equity alone
  225. 225. We never see the true link picture
  226. 226. You can no longer view raw link data and assume you can understand or compare
  227. 227. • We never get a complete view • So much of the profile is subject to negation through external factors
  228. 228. • Never have all the links • Can’t see disavow activity on other sites • Can’t see negation through Penguin • Can’t see lack of trust through disavow weight • Plus other signals now compete for the status of primary signal
  229. 229. Links used to take the place of the user.
  230. 230. Links now qualify the user data
  231. 231. • Its hard to diagnose changes in the link graph • Its hard to attribute risk and value • So much is hidden to us
  232. 232. It is clear that they still play a role
  233. 233. Trying to define their value
  234. 234. Since mid 2016 I have been trying to build a new metric for Kerboo
  235. 235. Caveat: This is purely about link value itself and takes no notice of all the other reasons why a site might perform or fail
  236. 236. Traditional link value • Link equity via CF / PageRank style • Modified link equity measure like TF • An aim to show true authority like DA / PA
  237. 237. Building LER We have a rather unique mix of modifiers to play with
  238. 238. Letting the idiot play with numbers… • Take the basic mechanics of the link equity for the link • Dampen or boost that based on known or assumed trust signals • Negate portions of its value via the potential negation factors we discussed earlier • Qualify that opinion with humans
  239. 239. Running the LER across profiles changed how I look at the whole link profile
  240. 240. • Take a link profile • Ask people to vote on what the best links are (definition: most important for SEO only) • Run the LER calculation across them • Scratch head…
  241. 241. Running the LER calculation across the profile highlighted links that weren’t as obvious before
  242. 242. Sites like this rise to the top
  243. 243. LER helped me reaffirm that I need to look again at links that can pass equity and be trusted
  244. 244. Think about the passage of equity
  245. 245. Feels like an old fashioned way to view the world But I’m finding it hard not to at least think about it a little more again
  246. 246. “But I want links that send traffic”
  247. 247. So do I but I also want links that…
  248. 248. • Send traffic • Send equity • Show no intent
  249. 249. So…
  250. 250. • Think about Intent • Think about how likely Google are to trust the link • Think about traditional link mechanics
  251. 251. Don’t fall into the cycle of intent again…
  252. 252. If we all chase the same types of links their intent increases and their trust decreases
  253. 253. What is ideal today is LinkDebt tomorrow
  254. 254. Intent is your risk
  255. 255. Links are important
  256. 256. But they are one of many signals
  257. 257. Their role is to qualify
  258. 258. They qualify a page is worthy of consideration
  259. 259. They qualify trust and intent
  260. 260. They qualify the user data
  261. 261. If you pass those tests they help boost your visibility
  262. 262. But ONLY as part of a more complex ranking journey
  263. 263. Links aren’t Google’s primary ranking factor
  264. 264. Huh?
  265. 265. Being the right page to match the true intent of the user is the primary ranking factor
  266. 266. Links just help your content become visible to the machines for consideration
  267. 267. And in some cases links are the loudest signal
  268. 268. And in some cases other signals are as loud
  269. 269. Links shouldn’t be forgotten
  270. 270. But they’re a function of other marketing activity
  271. 271. Without links you are still going to find it hard to compete
  272. 272. Even if you’re relevant
  273. 273. But if you’re relevant then links are part of the mix of trust and confidence to rank
  274. 274. How much they influence depends on many many factors
  275. 275. Niche Competition Brand influence Trust Negation from Google
  276. 276. It’s complicated people
  277. 277. It’s not going to get any less complicated
  278. 278. But we will adapt and succeed
  279. 279. Its what we do….
  280. 280. Because WE ARE SEO’s
  281. 281. Thanks for listening @pauldavidmadden Kerboo.com

×