Dawn Anderson owns Bertey, a boutique digital agency.
Dawn Anderson delves into the detail of natural language processing and information retrieval, IR and believes information retrieval, a branch of computer science, is the backbone of SEO, that the two are inseparably intertwined.
When Dawn is not attending an IR conference, presenting research herself, writing for a plethora of SEO news sites or doing client work you can find her exploring the countryside with her Pomeranian dogs, learning to code, or baking.
This post is part of a series called ‘Conversations in Search’. I discuss the current state of SEO practice with other SEO experts and discover their views on the future of SEO.
Damien: How are you, your friends and family at the moment, given all that’s going on?
Dawn: I’ve not seen my mum for weeks now. They’re all fine we’re keeping in touch like everybody else is. I know a lot of people kind of rushed you back out there and the news reported there were three-hour queues outside Primark the other day, but we’ve literally just all stayed in and just want to be on the safe side and be careful.
Thankfully, we’ve got a beautiful garden, and we’re in touch with everybody, and I’ve worked at home with my husband John for a few years now anyway, so it’s kind of business as usual for us, but it’s extended on the weekends. Ah, my cat has decided to join me.
Damien: What’s your cat named?
Dawn: I have two cats. This one is Tia, and I have a very evil one called Evie. Everyone is scared of Evie, even me.
Damien: Tell us about you before SEO, Dawn. Where did you grow up, how did you get started in your career?
Dawn: I am from Rochdale. I’m not sure you can tell by my accent, you probably can, it’s not quite as broad as it used to be. At one point wherever I used to go in the world, on holiday, somebody would always say, ‘Are you from Rochdale?’.
A few people will be surprised to know that I didn’t do a Bachelor of Science. I never went to University in my early days. I got my first jobs when I was 14-years-old, doing paper rounds and working in a local bakery doing the washing up, an after school job I’d inherited from my older sister.
Then when I left school, I got a summer job, ironing skirts in a factory, which was my first ‘full-time’ job. I was due to attend college after the summer holidays to become a legal secretary, but I suddenly realised I had some money from the summer holidays job and started going out, so my parents were furious.
One thing led to another, and they decided they weren’t going to keep me through college or university, so I had to carry on working. Thankfully I moved on from the Summer job to working in an insurance office as an Office Junior, and spent many years in those office / secretarial roles (despite missing out on the college course).
At 23 I started running my own businesses, and continued to do so for the foreseeable, eventually finding the need to ‘optimise’ my own business websites in 2006/7 and accidentally falling into SEO.
I never got to University until I was in my 40s, where I studied part-time with the IDM in Digital Marketing, and then I did a Master’s in Digital Marketing back to back at the Manchester Metropolitan University. It took me five years; in total, part-time. I now guest lecture at the Manchester Metropolitan University on search engine marketing and optimisation.
I started my businesses from my early 20s, and I discovered the internet. I started building websites, mostly because I couldn’t seem to find anybody to do it, so thought to myself I’m going to do this myself.
I wanted to be a developer before I wanted to be an SEO. I was in a funny position because I was older, and I was effectively still junior. Nobody wanted to employ an older junior in the world of development, and I wasn’t concerned about earning a fortune, I just wanted to get my foot in the door.
I took a year out to learn PHP, Drupal, Joomla all these different things for my projects. I also I took a year out to learn PHP, Drupal, Joomla all these different things for my projects in 2006/2007. I also needed to SEO the websites that I’d built, so I did loads of courses and read loads of stuff on SEO. Then I went to Latitude for an interview.
A recruiter said to me, ‘Hey you can’t do SEO can you, do you know much about SEO?’ and I said ‘Well, yeah, because I do SEO on websites I work on’. I met Phil McKechnie, and he was the Head of SEO at the time at Latitude. He interviewed me, and I got a call on the way home from the interview, saying they’d like me to start.
Damien: Then you started your own consultancy, focused on growing your business and eventually you began to guest lecture at Manchester Metropolitan University. How did that come about?
Dawn: I started working with clients, and I had my affiliate projects as well. I had the affiliate projects first and found more people asking me if I’d do SEO for them. It was a natural progression to create a brand and agency, more of a consultancy. You know, we all make choices, and my preference is not to have a big agency with lots of people in a room. I like having something that has more flexibility.
A person who’s been really good to me is David Edmundson-Bird. He was leading the digital marketing course that I was a student on at Manchester Metropolitan University. I met a lot of people through my studies who I’m still in contact with to this day, and who I value.
David Edmundson-Bird is one of those people. David asked me whether I would be interested in doing any associate lecturing because I enjoy sharing knowledge. They were looking for an associate lecturer in the Fashion School within the University to teach marketing analytics, and thought it would be an excellent addition to what I’m doing with clients and working on my projects.
I try to teach them [students] the core, the ‘red thread’ or fundamentals. I tell them to be careful because there is a vast ‘link economy’ online. There’s the whole influencer space that is driven by the link economy, where a lot of money changes hands. I make them aware that these things are against Google guidelines.
I think forewarned is forearmed. I tell students about the things that they can do that are in keeping with search engine guidelines, creative ideas that are about building those advocacies and building brand communities.
For me, SEO is about helping a search engine to understand fully the value that you provide, rather than artificially inflating popularity.
Damien: Who would you say has had the most profound influence on you, which you see reflected in your daily life, Dawn?
Dawn: A mentor I met back in 2008, Nisal Kooragoda. I was working on building my affiliate project at the time. Nisal is an enterprise architect who I got chatting to, and he gave me a considerable amount of advice about information architecture.
He taught me about the core principles behind organising information, structuring knowledge within an ontology and so on. And one of the first things he said to me when a lot of the significant changes came in SEO. When the Penguin algorithm update hit, many people rushed towards creating lots of content that was just pointless.
It was like a knee jerk reaction, to make more content given the low-value links were under attack. Initially, it was getting more links, and you ranked higher. It didn’t matter about quality. The reaction to Penguin was that nearly everybody swung in another direction. Get more content, and it didn’t matter what it was.
I was talking to Nisal at this time, and he said to me, ‘Dawn, where is the ontology?’. Always think to yourself, where is the ontology when it comes to creating anything that you put within a site. Having an ontology front of mind means you’re building from a body of knowledge about a specific topic.
Then the Panda algorithm arrived, and those same people began to prune their otherwise pointless pages, because of the impact, or fear of this new algorithm.
There are so many examples of people who saw their websites just tanking when they pruned away without consideration for the concepts of ontology and co-occurrence. They took away the relatedness and co-occurrence of adjoining pages that were about a specific topic and contribute to the domain topic.
Damien: On their own, those pages appeared thin, but they were lending weight to those topics and providing signals beyond their content value?
Dawn: Yeah, you know they were providing semantic signals. If you think about pages in clusters, then if you suddenly chop away half of a bunch of grapes, you’ve not got a whole bunch of grapes any more. I see these reactions from people, and they’re often dangerous.
For me, there is a difference between pruning, taking a machete to something, and trimming. Trimming is where you’re just slicing off the useless edges (ages which are just not important when you consider their value for users or overall importance to search demand).
It’s sensible to trim the pages that are least important or the lowest quality. Google is selectively indexing because not everything is important. It follows the Pareto 80/20 rule, and it’s what the Zipfian distribution is based on.
That is, 80% of the pages satisfy 20% of the need. The largest websites of the world probably meet a few percentage points of need to be fair. That means 80% of the content on the site will be less useful, and probably in a Zipfian curve with decreasing levels of importance.
Damien: The emphasis on trimming and not taking a machete to how you clean up your pages is a top suggestion. What else do you suggest people focus their attention on to optimise their website?
Dawn: It’s crucial to structure your website with importance as a primary consideration, where the important parts of a site are found very easily. Thinking about the value that those very weak, long-tail pages without search demand have or don’t have, and trimming those pages.
The concept of Zipfian distribution is everywhere in the world. It’s a Power Law, and it applies to everything. It’s the reason why page rank doubles from between 1 and 10. It’s the reason why the first position in search is twice as relevant as the second.
Can you help users with their information needs? If you’re targeting products, and you never had any of those products in stock, or you know you’re not likely to get them, trim. If you don’t meet with an informational need or you don’t stock something that you’re trying to target, trim it. You can consolidate it and create a page which is of greater depth and before you know it, you’ll start to move the Zipfian distribution of relevance.
People often say ‘I improved from position 50 to position 30, aren’t I doing well?’ No, you’re not. Why? Because in search, only results in the top 10 or perhaps 20 matter, probably because of ‘Top-K re-ranking‘ for efficiency and precision.
Know the difference between what search engineers call precision and recall. Precision is ‘these are exactly right results’, and recall is ‘there’s mention of this’. When you search, it says 55 million results, that’s the recall. Precision is the top 10. Just because you’re in a recall where you get an impression, that impression could be positioned 100 or 300. That doesn’t matter because search engines are not concerned with anything beyond the top 10 or 20 results.
Search engines have the concept of ‘Top K Re-ranking’ because of efficiency, they rank everything, they take the top 10, or perhaps 20, and they refine that because that’s mostly what matters.
Damien: I’ve seen that mentioned in one of your articles. Speaking of your writing, you wrote about BERT recently, can you tell us more about that and why it is a critical topic for SEO?
Dawn: BERT goes well beyond the field of information retrieval and more into the field of natural language processing. The two worlds are merging, three worlds really because of machine learning as well which has created a sea-change in the way that search engines can understand the language and the intent and the context behind what a user is after.
For instance, BERT was created by Google. It’s many things, it’s an algorithm name, but it was initially a paper in 2018, a research paper that they then gifted as well after presenting it academia as a model to the machine learning community. Many big tech companies have run with it, including big players such as Facebook, Baidu, even Toyota. Massive brands are using it to build their bespoke models.
It was costly, and it used a lot of computing memory etc. Now it’s becoming more efficient with communities like ‘Hugging Face’ for instance that are working very much on creating more efficient and lite versions of BERT where you can train models to understand language. Even Google has worked on producing a more efficient and computationally inexpensive version of BERT.
Search engines are doing this at scale, and then eventually the worlds of research and practice collided with BERT.
Last October 2019, Google announced that they were implementing, in part, into their search machinery, BERT (or a version of BERT, since BERT from the original was greatly improved by the research community and presumably tech companies in that year between first academic paper and Google announcing its implementation in part in search in the wild).
BERT is needed where there is ambiguity and particularly, where there are ambiguous words that may have many meanings, either in text or speech. I always use the example of ‘four candles’.
A lot of the humour of the past was very much about the misinterpretation of words. That was between humans, so how much more difficult has it been for search engines and just dumb machines to understand when they don’t have the context of a scenario?
BERT tries to understand the context of a word, particularly in the body of a text. BERT was trained on the English version of Wikipedia, and it was also trained on the ‘Brown Corpus’, a collection of books.
Another leap forward this past six months has been something called Reformer, which is an extension upon some technology that partly powered BERT called Transformer. A Transformer focuses attention on particular words, in the context of all the other words in a text.
Reformer has extended upon that so that models like BERT when used with Reformer can understand the context of a word in a much larger body of text than previously. Before, BERT was able to understand what a word meant in the context of a paragraph and then in an article. Now BERT and these models using this thing called Reformer can understand the context of a word within the body of a book.
It’s so important to have informational content in blogs, articles and guides. A luxury fashion website will have lots of images and is never going to have a million words on a page. And, it shouldn’t because the fashion world is very much an image-led field, where users want to see pictures rather than read about what a dress looks like. If I am shopping for a dress I want to see pictures and run-way videos. I don’t want to read ‘War and Peace’ about a dress.
They might have just a paragraph of description. Those word-rich pages that are informational from blogs, interviews, fashion show write-ups and industry news, linking across and being part of the site itself will help to add context to the things that are its neighbours, like those image-rich transactional pages and galleries. The search engine realises pages with few words will be related to the images shown within the website.
Damien: It’s not about manipulating an algorithm, it’s about thinking holistically, how your site meets the needs of users?
Dawn: Yes, exactly. Features will become increasingly important — features such as video, imagery, 360-degree runway videos or even user-generated content, which is natural language ‘naturally’, and illustrates the fostering of a community (and naturally attracts links without manipulation since advocates generally want to share what they love and respect).
Damien: When is it right to make use of user-generated content?
Dawn: User-generated content is vital for the future because you want your audience to be part of your community. You want them to become advocates, you want them to join, taking them up the loyalty ladder, so they’re spreading your message. All of that creates buzz around your site and helps your SEO.
Those are the kind of things to think about for now and the future. Building audiences, creating armies of advocates and providing features that bring people flooding back to you.
It’s imperative at the same time, to not think that these things will happen without SEO. It’s easy to start building a juggernaut without direction. Juggernauts, once they begin to go in the wrong direction, can quickly go off the cliff.
The site grows bigger and bigger, it doesn’t take a lot for it to become a bit unwieldy mess and suddenly the ontology is all over the place, particularly when there are the likes of faceted navigation which is all accessible to search engines.
It starts off well, but eventually you end up with many combinations which either make no sense for indexing and / or compete against many URLs which output the same result in terms of value-add for searchers, the value of which search engines divide mathematically between these pages, meaning none of them can rank highly, because of the shared value.
I have this other notion as well, something call ‘overflow SEO’. That is where a site grows and you end up with an inventory of content to manage the visibility and importance signals on intelligently the content that was on, you know the first page of a category suddenly gets pushed out of the first page but it was important, and that’s where logical sub-categorisation and strong internal linking, scaled well ideally, becomes very important, to maintain visibility for important content.
Damien: The devil is in the detail. Do you want to know all of the details?
Dawn: Well, I do like to know the details; every detail matters. Because on a website, everything matters. Every single word that gets added has an impact in some way. It’s like the game of Jenga, where every block you remove has a potential impact, so it’s a case of being intelligent in the game of Jenga you play on a website.
Damien: The fight for equality in lots of ways is ongoing. Thinking specifically about gender equality, as a woman in the SEO industry, what’s it like?
Dawn: It is still an industry very heavily skewed towards males, as is the tech industry as a whole. I can’t deny that it takes a while for people to take you as seriously as they would do if you were a male, I think.
Even to this day, I’ve been overlooked or dismissed in some ideas I’ve put forward and I think that in part is because I am female. Even with clients sometimes, when you’re in a very male room, there is that bias still in place in some of the traditional industries.
You only have to look at the business development teams of the vast majority of SEO agencies and tool providers out there, which are heavily skewed toward males in the senior roles. Except for a few. For example,here’s Ali Parcell at ‘Builtvisible’ and Jen Hoffman formerly at ‘DeepCrawl’ who do great jobs of building their marketing teams. And of course, there’s the team at SEMRush who have lots of female senior players, such as Olga Andrienko.
The vast majority of the traditional business owners would still be more convinced about SEO if males walk into a room and convince them. At some point, I hope it will disappear. I’ve earned my stripes regardless of being male or female. Being disregarded means I’ve been able to sidle past a load of people, and they didn’t even notice because I was just busy getting on with it and moving forward.
It does mean in a way that you have an advantage. It’s never a good thing to be disregarded for being female. If it has happened to women out there, they should take heart and turn it on its head, use it as a competitive advantage. Because when a competitor doesn’t realise you’re passing along on the inside lane, you have power.
Damien: Luvvie Ajayi gave a talk at TED Women in 2017 focused on as she puts it ‘being the first domino’, and getting comfortable with being uncomfortable. Have there been times where you felt like you were the first domino?
Dawn: If I am honest, it’s no, I’m a lot more introverted than people might think. I tend just to want a quiet life where I can just stick my head into server log files or look at code. At the same time, I respect and admire people that are pushing forward on important topics and empowerment; I respect and support them, albeit from the sidelines.
Damien: What questions do you think SEO should be asking in their businesses or of their clients that they aren’t?
Dawn: One significant thing is to be selective when taking on clients. It is very easy just to take on any client because they’ll pay the bills. The client relationship should be great because you have the potential to make or break their business. With search, people are actively seeking something. Clients need to invest in SEO for the long term and realise that even when things look like they’re going well, SEO investment needs to continue.
SEO is forever. SEO is not a three-month project, is not a two-year project; it is forever. It’s a continual part of the product. I believe SEO and marketing work together; SEO is marketing with technology. You have to be continually thinking about how can we emphasise that this meets the users’ informational needs. Your product may be pretty, but how does it help the user? Is it understandable by a search engine?
It’s very easy to think ‘Well we’re doing well now, we don’t need the SEOs’ and that is where it all just becomes chaos. Because you know you have a content inventory which needs to be maintained, updated and kept accurate. If the content is not valid, you will lose all kinds of nice things. For example, the featured snippets will start to fall away if their content is not maintained.
Damien: Dawn, what’s your view on the future of SEO?
Dawn: Search will become more assistive everywhere. It will be an ‘all-around us’ experience. I know it’s a way-off, and things don’t move as quickly they think they do.
There’s a massive amount of research going into Voice Search. Voice is now going into things like our TVs, gaming etc. One day we’ll turn around and see voice search is ubiquitous. Remember how mobile-first took a long time, but then it felt all of a sudden? We need to be spending a lot more time, as SEOs, on these topics, and not be on the back end of the information.
I read the research and most things I’ve learned over the past 20 odd years have been self-taught by scrambling around, looking at videos and reading research papers. Because I am keeping myself updated, digging into the research, it’s been years since a search change has surprised me.
I’m impressed by the new head of search, Prabhakar Raghavan, has a strong IR background. He co-authored the Introduction to Information Retrieval with Christopher Manning and Hinrich Schütze, and I encourage people to read it.
I hope that they’re going down the path they are increasingly talking about search in the context of information retrieval.
Damien: What do you think is most misunderstood about SEO by business people?
Dawn: I think people underestimate it. Undervalue it. Think it’s a bolt-on. That it’s a nice to have. They think sometimes they are doing well despite SEO, and not because of SEO. They don’t understand in most cases.
The conversations we have with the layers of bureaucracy between senior management and SEO agencies or consultants, can break down, and the concepts don’t get through to those people that are making the decisions that matter.
Some of the best organisations exceeding online have flat structures, and are much more agile. Big traditional brands and companies generally would do well to de-layer and lift SEO to the critical decision-makers.
Damien: Dawn, there’s a lot of insight and more to read now, thank you. I am looking forward to speaking again soon.