It’s been quite some time since I’ve seen a conference that got so many well-respected SEO’s and digital marketers genuinely excited.
On November 30th, 2017 – Paul Shapiro and Catalyst Digital hosted TechSEO Boost, the industry’s first conference dedicated solely to technical SEO in Boston, MA. Below you’ll find all 11 videos, 10 slideshare presentations, notes and every referenced link that I could track down.
And a HUGE shout out to Ruth Reedy Burr (site / Twitter) of Upbuild, who also presented at this event. She was a live tweeting machine that day, posting copious notes and references.
Wanted to genuinely thank @fighto for putting together what was easily the best SEO Conference I’ve ever been to. To all the brilliant folks I had the opportunity to meet at #TechSEOBoost thank you. Can’t wait for next year.
— nick Ξubanks (@nick_eubanks) December 1, 2017
Caught a bit of #TechSEOBoost yesterday and it's exactly what a conference should be. Focused, insightful and actionable. Definitely the way forward for marketing conferences.
— James Whitelock (@jwhitelock90) December 1, 2017
Hopefully marketing event organizers are watching #TechSEOBoost. This is what nailing the programming looks like. (cc: @fighto)
— Ronell Smith (@RonellSmith) November 30, 2017
Table of Contents
-
- The State of Technical SEO
- Working Smarter: SEO Automation to Increase Efficiency & Effectiveness
- SEO Best Practices for Javascript-Based Websites
- How Machine Learning is Shaping Google and Technical SEO
- Everything That Can Go Wrong Will Go Wrong
- Enhancing Your Local Presence With and Without Physical Locations
- Amping Airbnb.com
- Implementing and Tracking Semantic Markup with Google Tag Manager
- Technical Tips & Success Strategies from Engineers
- Your Search Universe: Links, Log Files, GSC, and Everything in Between
- Making the Web Fast
- Further Reading
The State of Technical SEO | Russ Jones, Moz
“Leading vs. Lagging: filling out a technical checklist is not a career!”
What is Technical SEO?: Any sufficiently technical actions undertaken with the intent to improve search results.
▶️ Most SEO’s are working somewhere in the technical stack, whether in the server, the code, the DNS, etc.
▶️ Technical SEO is wholly focused on results, not interested in one-upmanship – it’s about innovation.
▶️ Adoption rates: The top 1000 sites are using AMP heavily, while rates decrease as you go further down the list.
▶️ Technical SEOs are the “R&D wing of our industry.”
▶️ Highly performing sites are early adopters of new technical SEO.
▶️ The industry follows the technology. Interest in JSON-LD drives interest in technical SEO.
▶️ “The difference between technical SEO being a job and being an industry is when people have a career doing it”.
▶️ DistilledODN vs Cloudflare for tech SEO benefits. The WP plugin allows a low barrier to entry. Industry theme: Specialty vs. Scale, e.g. DistilledODN (specialty, $) vs. Cloudflare (scale, free).
▶️ The bad news: If your knowledge scales, it will be automated. To be successful, you have to stay on top of the latest and greatest – or create it yourself. Have a specialized clientele – people who can afford & adopt new tech.
▶️ Developers: convert specialty to scale. It’s your job to put consultants out of their jobs with scaling & automation.
▶️ Researchers will find opportunities and new techniques even when Google is silent. Invest in research; it is the future of technical SEO…and SEO in general.
▶️ “If I had one word of advice it’s that consultants, and developers should be looking to invest in research. Your researchers are the future quite literally of your industry. They will produce for you what you can sell, and if you aren’t, then you’re only reselling what Google has told everybody else to put on their shelves, and that’s only going to last a year or two. We know it well because we’ve seen the graphs that prove it.” – Russ Jones
Working Smarter: SEO Automation to Increase Efficiency & Effectiveness | Paul Shapiro, Catalyst
“We love this tool called KNIME, that you can get that through KNIME.com. It’s free and open source and what I really like about it, is it’s literally this drag-and-drop interface, so the barrier to entry is very very low.
Even some of our non-technical staff can get their feet wet and start playing around with it. It’s completely free, and there’s this huge library of things already built into it.”
▶️ Nodes are pre-built modules designed to complete a task. String them together to automate more complex tasks.
▶️ To automate you’ll also need a scheduling system like Cron, to schedule recurring tasks.
▶️ You can automate keyword research: add your seed keywords and automate pulls from keyword tool APIs.
▶️ Instead of just spitting out your research into a spreadsheet, use simple visualizations to make dynamic decisions. Use interactive data visualization like @tableau or Google Data Studio to make decisions on the fly.
▶️ Automate reporting. “The most obvious thing you should be thinking about automating is reporting. I remember my first SEO job, and we spent hours upon hours making reports in Excel. That’s really dumb because this is literally something that should take you five minutes.”
▶️ Any procedure that can be broken down into smaller tasks that a computer can do can be automated.
What can we automate for SEO?
- Reporting
- Data collection
- 301 Mapping – Automate 301 redirect mapping from old URLs. [see graphic]
- Test javascript rendering
- Performance testing with lighthouse
- Anomaly collection
- Content ideas – Reddit data mining
- AMP validation
- Link building
- Tech audit
[source]
Links
How to Automate Your Keyword Research | Paul Shapiro
Script to extract performance metrics
Crawler with node.js to test rendering
Fixing historical redirects using Wayback Machine APIs | Search Engine Land
SEO Best Practices for Javascript-Based Websites | Max Prin, Merkle
▶️ Search engines are not just crawling anymore – they’re rendering. Google is a headless browser now.
▶️ Crawl > Index > Rank has become Crawl > Render > Index > Rank.
▶️ “Google is working on updating their Web Rendering Service (WRS) that they use when they crawl your site based on the latest version of Chrome (currently Chrome 41)”.
▶️ We don’t build websites for search engines; we build websites for users and make sure search engines can understand them.
▶️ APIs won’t be activated/used by a web rendering service. Crawlers and platforms that don’t render the page won’t see things that are in the DOM but not the code. Bing, Facebook, LinkedIn, etc. aren’t rendering as well.
▶️ Twitter cards/OG tags that are dynamically rendered via JS may not be visible to Twitter/Facebook.
Bing is yet not rendering JS at scale, though there are signs that they can do simple things like follow JS redirects.
▶️ A single-page application should not be a single-URL application. Content has to live somewhere.
▶️ If the content changes but the title/stays the same, that’s a crawler issue. A lot of websites change the main content that’s rendering, but don’t change the. Lots of important SEO signals get lost that way.
▶️ URL structure: escaped fragments don’t always load for search engines. Google is deprecating the hashbang – they’ll stop requesting the escaped fragment altogether. Leverage pushState to serve a clean URL instead of using a #.
▶️ Internal links: onclick + window location doesn’t mimic a classic HTML link as far as a search engine’s concerned.
▶️ Internal linking is a big deal. It can overwrite canonical tags. Just make sure it’s done via traditional The 5-second rule for loading content. The 5-second rule: Google might not see things that take 5+ seconds to load on your page.
▶️ Chrome Dev Tools: Console – Lighthouse check for JS rendering errors.
Links:
Nozzle
TechnicalSEO.com
Fun with Machines. How Machine Learning is Shaping Google and Technical SEO | JR Oakes, Adapt Partners
▶️ TF-IDF sounds very fancy, but it’s not the magic elixir to dominating on Google.
▶️ Wikimedia released their MjoLniR model on GitHub last month. Look at information retrieval mechanisms to get insight into how Google might be handling it.
▶️ Machine learning is significantly more effective as an information retrieval mechanism than BM25.
▶️ BM25 beats TF-IDF as an information retrieval algorithm because it considers more factors like page length. Wikipedia transitioned to BM25 in 2016 – now they’re using machine learning models.
▶️ TF-IDF won’t work because it looks for related phrases, but is a frequency, not a relevance metric.
▶️ Content loading on the page after 5 seconds? Probably not going to be indexed.
▶️ Instead of just clicks, Google wants to use (e.g., mouse) movements to gauge user satisfaction. Google is interested in click data and is thinking a lot about how to extrapolate value from user activity.
Google Search Quality Guidelines is the ground truth for Google’s Search machine learning problem.
▶️ You may not need to “optimize” your meta descriptions anymore because Google will figure out the “best” story for you. Google could take all of our content and make a better page out of it.
▶️ LSTMs pick up referential information in documents and tie that information together. Focus on: query disambiguation. When the query is vague, Google gives a nudge via suggestions.
▶️ Use Google’s query disambiguation (“don’t hit enter”) to disambiguate your own site, e.g., your navigation. Understand how Google uses content/semantic relationships to improve your content’s internal relevance.
▶️ Google knows the frequency with which words co-occur on the internet. They can use that knowledge to determine whether or not content is expert-level.
Links
Duplicate detection with shingling notebook / Semantic Relevance & word vector notebook
Learning to Rank with Selection Bias in Personal Search | Research Paper
Everything That Can Go Wrong Will Go Wrong | Patrick Stox, IBM
▶️ Noindexing pages via meta robots won’t do anything to remove pages from the index if you’re blocking them with robots.txt. Disable caching of .txt files unless you want to flirt with accidentally blocking everything from being crawled.
▶️ Google is processing up to 5-6 hops on redirect chains. Redirecting based on browser language can result in Google indexing English pages instead of other languages.
▶️ Bing does process JavaScript when crawling pages.
▶️ React JS can result in unfriendly URLs / uncrawlable links – but it’s very customizable so just educate the dev team on what needs to be done.
▶️ “You’ll never get things exactly as you want in enterprise technical SEO. Pick your battles.”
▶️ Thinking of changing all the links in your CMS (including canonicals) to relative? Don’t do that. Create an exception for canonicals and use their absolute paths.
▶️ Custom 404 page returns a 200 – redirecting all 404s to that 200 page. DON’T DO THAT.
▶️ Using UTF-8 characters in your URLs? Check your whole stack. Just because the CMS supports it doesn’t mean everything does.
▶️ AMP is really hard to implement on an enterprise site. (see Airbnb presentation for more on this). Companies’ legal requirements for hosting content is one of the big barriers to AMP adoption.
▶️ Look at any new system that’s on the way in – try to get ahead of any problems.
▶️ You will never fix everything as a technical SEO. Sometimes it’s actually cheaper to not fix stuff.
Links:
You’re Going To Screw Up International SEO | Patrick Stox
Host A Website Inside Robots.txt | ohgm
HTTP to HTTPS: An SEO’s guide to securing a website | Search Engine Land
What you need to know about Referrer Policy | Search Engine Land
The AMP is a Lie | Search Engine Land
Enhancing Your Local Presence With and Without Physical Locations | Joe Esposito, Rio SEO
▶️ “You don’t necessarily need physical locations to rank well at the local level. You don’t necessarily need physical locations to have any kind of impact effect on local, conquering the local space, and ultimately ranking at the local level.
SERP’s are all local.All the ads that are coming through from AdWords, all the organic listings, everything is tied back to local. There’s no modifier, there’s nothing geo about the actual search, but the results all end up being tied back to local” – Joe Esposito
▶️ “When you lack local content, it takes the user too much time to find what they are looking for… make sure you have content across the hierarchy.”
▶️ Use breadcrumbs with a site with a store locator to send stronger location signals.
Amping Airbnb.com | Gil Birman, Brian Ta, Airbnb @gilboxme @fanfavorite_bta
▶️ 30 different teams work on Airbnb’s core search landing pages.
▶️ AMP pages were timing out about 16% of the time. They implemented CDN caching to get around it.
▶️ New features like horizontally-scrolling image containers broke AMP implementation.
▶️ AMP is displayed in an iframe – no custom JavaScript, no custom styles.
▶️ AMP doesn’t support inline styles – only style tags which appear in the.
▶️ Use an ERB template that calls render_react_component & send to Hypernova AMP to render there.
▶️ User flows for AMP work well for 1st-time/logged-out users, but it’s a confusing experience for logged-in users.
▶️ Use logic and reason to reduce CSS that mucks up AMP.
▶️ Deeplinking – Google does NOT open a deep link when clicking on an AMP result. Users have to click again to get to the app.
▶️ Implementing AMP targeting a limited audience increased bookings from SEO by 6%.
Links
AMP For Retailers: Is It Worth It? | Smashing Magazine
Implementing and Tracking Semantic Markup with Google Tag Manager | Mike Arnesen, Ruth Burr Reedy
“Semantic markup should be SEO101 these days. As basic as title tags.”
▶️ Rich snippets are evolving into rich cards on mobile. Search engines are doing more and more and using markup to extract that data.
▶️ Schema.org ‘s latest version (3.3) includes “how-to” markup. This means no more using the recipe markup for your non-food how-to content.
▶️ Structured data in RSS 2.0 helps assistants like Google Home play and understand information about your podcasts.
▶️ “There are early signs that Bing will support JSON-LD markup too.”
▶️ “Structured data is driving voice search.”
▶️ JSON-LD is faster, easier, cleaner and better in every way (vs. inline)
▶️ Upside of dynamic JSON-LD: unlimited power and flexibility. Downside: it takes some persistence.
I absolutely love how @Mike_Arnesen and @ruthburr are breaking down this code example. That’s not an easy thing to explain in a presentation. #TechSEOBoost
— Paul Shapiro (@fighto) November 30, 2017
Links
Implementing Semantic Markup with Google Tag Manager | Upbuild
The Search Marketer’s Guide to ItemRef & ItemID | Moz
Using Google Tag Manager to Dynamically Generate Schema/JSON-LD Tags | Moz
Technical Tips & Success Strategies from Engineers | Moderator: Andrew Ruegger, Catalyst
Moderator: Andrew Ruegger, Head of Data Science, Catalyst/GroupM
Panelists: Gil Birman, Software Engineer, Airbnb; Derek Perkins, CEO & Co-Founder, Nozzle; Tanner Linsley, Co-Founder & VP of UX, Nozzle; Mike Atwood, Lead SEO Engineer, Wayfair
We’re all working towards the same goal, but may not be speaking the same language. Here’s a panel of engineers sharing insights and ideas that will advance your technical knowledge and
facilitate better collaboration with your engineering counterparts. This panel will provide you with strategies and tips for how SEOs and engineers can work more effectively together to accomplish business goals.
Your Search Universe: Links, Log Files, GSC, and Everything in Between | Sam Marsden, DeepCrawl
▶️ Log files will show you where search engines go on your site, and how frequently.
Look for:
– non-indexable pages with Googlebot requests
– indexable pages w/o Googlebot requests
▶️ Look at backlinks to nofollow pages – where is link equity evaporating?
▶️ Duplicate pages can split your backlinks. Canonical/redirect to consolidate link equity.
Making the Web Fast | Patrick Meenan, Google
“sprinkle machine learning on everything and magic comes out” @patmeenan
▶️ Performance metrics are the biggest contributors to bounce rate – DOM Content Loaded was the biggest predictor.
▶️ Over the years Google has been evolving more toward user-facing metrics like First Paint. Try to get real performance data in w/your biz metrics, including Time to First Paint
▶️ If your website loads too fast, your conversion rate will start to drop.
▶️ Don’t optimize towards a full page load. Instead, give them the most important stuff immediately and worry about the rest later.
▶️ By focusing on visual metrics, chrome drove a faster visual experience but regressed on interaction. Get more content in front of the user as soon as possible – Speed Index explained
▶️ Time To Interactive: The time it takes before a user can actually do something.
▶️ First interactive looks toward the beginning of the waterfall, TTI looks to the end. If you see a lot of red at the bottom of your waterfall, look at how much code you’re running: do you need it all?
▶️ Your experience with a high-end phone won’t match that of users with slower phones. Get a couple of Moto G4s or other $100-ish phones and use them to test your mobile site to compare.
▶️ Averages suck – look at the histograms. What is the distribution of your performance?
▶️ Weekend/weekday variations – people are on fast computers at work, slower ones at home. When measuring UX, look at real user data.
▶️ When trending performance, lab data works better because of fewer variables.
▶️ Every web developer should frequently be testing their site with the “Slow 3G” network setting.
Links
High-Performance Web Sites: Essential Knowledge for Front-End Engineers | Steve Souders
Easiest place to get started w/recommended parameters
Page Speed Case Studies – Case studies and experiments demonstrating the impact of web performance optimization (WPO) on user experience and business metrics.
Reducing Domain Sharding | Code as Craft
Chrome User Experience Report | Google
Leveraging the Performance Metrics that Most Affect User Experience
Why Marketers Should Care About Mobile Page Speed | Think With Google
“Don’t A/B test with JavaScript. Do it server side unless you hate yourself and your site” paraphrased.
Further Reading
All of the #TechSEOboost tweets