SEO can be a hard craft. It can feel a slippery, fickle topic; the downside to the industry’s fantastic agility, adaptive streak and drive for constant improvement is a need to continually adopt new techniques, knowledge and tools.
That does however give us the perfect excuse to always be trying out new software and tools, and who doesn’t like to kick the tyres on some new toys?
It was while checking out one of the new keyword tools on the block, CanIRank, that I got chatting to Matt Bentley, the tool’s founder. Matt kindly agreed to turn our banter into a full interview. We discussed about how the tool fits into an increasingly crowded market, why he wanted to build it, and how he’d like to see the industry develop.
During our conversation Matt showed he’s not only very passionate about SEO and building a tool that helps marketers, but is also one of the industry’s most articulate and forward-looking thinkers.
CW: Hi Matt, thanks for taking time to speak with us; can you tell us a little about CanIRank – who is it for, and what can users achieve with it?
MB: We call CanIRank an SEO Intelligence tool: it’s part keyword research, part competitive analysis, and part artificial intelligence. The purpose of SEO intelligence tools is to improve your productivity by giving you a better understanding of why each site ranks where it does, which factors are holding you back, and what your biggest opportunities are.
In short, we let computers do what they do best — compile and analyze data — to free up more time for marketers to focus on what they do best — coming up with creative ways to attract and engage potential customers.
In the past 5 years SEO has become almost exponentially more complex. To really understand why a site ranks where it does you need to consider a huge list of factors (keyword usage, anchor text, topical relevancy, website authority, trust, speed, mobile friendliness, https, user behavior, etc…). It’s no longer possible for humans to keep all of that in their head and really understand how sites compare in these hundreds of factors, and which ones might be making the difference.
In the past 5 years SEO has become almost exponentially more complex
But computers can make sense of it all. CanIRank collects 4,000 data points from 11 different sources for every single keyword analyzed. Our software can take 40 randomized URLs, collect data on them, and re-rank them in the same order as Google to a 60% correlation. I don’t think the best human SEOs in the world could do that, even if you gave them a week.
And honestly, SEO is just a whole lot more fun when you can focus on the creative aspects, the problem solving and relationship building, rather than spending all day poring over data in Excel.
CW: I think for many folks it really is (though some of us still like geeking out in Excel!) For you, what’s the most exciting thing about CanIRank?
MB: The most exciting part for me is seeing the results we’ve been able to help clients achieve. First is the massive time savings. So far we’ve analyzed over 200,000 websites, which means collecting 40,000,000 raw data points and distilling that down into 23,000 Action Recommendations. That’s a 99.9425% reduction in data you have to pay attention to!
For small teams or those trying to get results quickly, it can be pretty overwhelming trying to figure out where to start, so it’s a big deal to know specifically which issues you should address to make the biggest positive impact on your ranking. SEO is just too big these days to do everything.
Of course, all of that wouldn’t be worth anything if our AI models didn’t make accurate recommendations, but we’ve tested that and clients who don’t follow CanIRank’s Action Recommendations actually see a slight decrease in ranking on average, whilst those who follow at least some Action Recommendations for just 1 week see an average ranking increase of 12%, and those who spend 1 month completing Action Recommendations see an average 25% improvement.
Obviously, that’s just the average. We had 1 client who managed to increase search traffic 533% in just one month — and he was brand new to SEO!
CW: Is that what you built CanIRank for, to help those with less experience or technical nous or is it for everyone, including agencies? Who is the tool aimed at?
MB: Our initial goal was to improve the transparency of SEO — to take all these ranking factors and distill it to something where anyone can just look at a CanIRank report and say “aha! so that’s what’s going on.”
So that’s certainly helpful for someone who’s a bit newer to SEO and maybe doesn’t have the years of experience that really badass SEOs do where they can just look at all these numbers flying by and instantly see the big picture, kind of like operators in the Matrix.
But as we’ve grown we’ve discovered that agencies and professional SEOs also appreciate being able to quickly diagnose issues, automate their keyword research, and find keywords they can rank for quickly.
Basically, everyone these days is under pressure to deliver more results in less time, so even the most expert SEOs benefit from the productivity and efficiency increases of automating the most time-consuming analysis.
You could give me detailed instructions on how to build a house, and I’d probably get through it with considerable time and effort, but it’s not going to be the same result as you’d get from a master carpenter
CW: You’ve built CanIRank to let users really develop their SEO understanding and ability – do you think it allows people to do SEO for themselves without outside training/expertise? Was this a goal?
MB: The combination of increased transparency into search ranking factors along with detailed instructions means that CanIRank is a great way to learn SEO. It’s “learning by doing” where each step is tailored to addressing the needs of your site, like a book that was written just for you.
That said, there’s always going to be a role for SEO experts. You could give me detailed instructions on how to build a house, and I’d probably get through it with considerable time and effort, but it’s not going to be the same result as you’d get from a master carpenter following the same instructions.
Experts can look at an issue that CanIRank uncovers and immediately think of a dozen different approaches for addressing that issue, some of which they have absolutely dialed. So for most businesses it’s going to be worth it to let the experts do what they do best.
But even for businesses who aren’t doing the work themselves, CanIRank can serve as a kind of bridge between the business and their SEO agency. Put it this way: hiring an SEO shouldn’t feel like a trip to the auto mechanic. I take my car in for an oil change, and the mechanic tells me there’s a problem with my rear flux capacitor turnbuckle coupler, and I should pay £3,000 to get that fixed unless I want to put my entire family at risk.
I have no idea if he’s just making that up; the car seemed fine to me! Looking at a CanIRank report, even a novice can see where their website falls short of the top rankers and have a real understanding for the challenges that lie ahead and the value they’re going to be getting from the work done by their SEO agency.
CW: I think that’s a fantastic analogy – bridging the gap between what we do and what our clients understand is a big part of the job, especially when it can sound like gobbledygook…
We’ve seen a fair number of new SEO tools enter the market as it starts to reach a new level of maturity; how do you see CanIRank fitting into an SEO’s toolbox?
MB: It is a very crowded market! Only a few years back the problem was we couldn’t get enough data. Now, the problem is there’s way too much! It’s almost a full time job keeping track of all the great data coming in from all these different marketing tools, there’s no time to act on what you’re learning.
And I guess that pretty much captures where CanIRank fits in: we’re a layer of machine learning intelligence on top of the tools that you’re already using (Moz, SEMRush, Ahrefs, social media, crawlers, keyword tools, etc.) that automates the data collection, makes it understandable, and helps you figure out which things you need to pay attention to.
We have a sister company (http://www.lightship.me/) in stealth mode that kind of hints at where we’re headed: automation and intelligence that turns all your raw data into directly actionable opportunities.
CW: What other online marketing tools have you used that you really liked? Have any of them inspired CanIRank at all?
MB: Well of course we’re long time users and big fans of the companies that provide data to CanIRank — especially Moz, SEMRush, and Ahrefs — all do a fantastic job and are really the backbone of SEO analysis.
More generally, I’m super excited about the potential of predictive analytics technology to transform online marketing. I think within a relatively short timeframe we’ll all be doing our jobs in collaboration with an AI of some sort — companies like Adometry, Pretarget, TellApart, Preact, and 6Sense are some other leaders in this space.
I owe much of the inspiration behind CanIRank’s technology (not to mention the direction of my career in the past 7 years!) to coaching and advising from Dean Abbott of Abbott Analytics who helped me get started as a data scientist and lay the foundation for CanIRank’s predictive analytics technology.
I think within a relatively short timeframe we’ll all be doing our jobs in collaboration with an AI of some sort
CW: Now that is interesting. We’re big fans of SEMrush, Ahrefs and Moz here as well. I think the possibilities for AI predictive input into what we do is fascinating, probably deserves its own dedicated conversation!
As you’ve brought up your background, I wanted to ask; you’ve been an online entrepreneur for many years, as well as working in data science, what’s your experience with online marketing? Is it something you’ve always been keen on?
MB: Online marketing, and particularly growth marketing, has always seemed like the most interesting challenge in startup companies, and I’ve been lucky enough to have spent time in a variety of online marketing roles at startups, some successful, some not so much. SEO has always been a part of that, but I’m definitely more of a generalist than some of the other folks who have spent 15 years doing nothing but SEO.
In a way, I think that’s been beneficial as the tool we built was what I always wanted as a startup growth marketer trying to grow traffic on a limited budget: it’s much more accessible than other tools that might only make sense to professional SEOs.
Online marketing, and particularly growth marketing, has always seemed like the most interesting challenge in startup companies
CW: What made you want to take on building an SEO tool? Was it simply about a gap in the keyword analysis market, or was it more about filling a need in the increasingly competitive industry overall?
MB: I talked with a lot of startups and small businesses about the challenges they faced with their online marketing, and one of the most common scenarios went something like this:
- Pick a target keyword that’s way too challenging
- Learn just enough SEO to be dangerous, start building links in all sorts of unsavory ways
- Get penalized by Googleand end up worse off than when you started
So we thought: what if we could build an app that could tell people whether or not a keyword was reasonable for them to target? They could avoid wasting time chasing impossible dreams, stay out of trouble with Google, and see results much more quickly.
Since then of course we’ve added many more features and it’s as much how can I rank? and what are my biggest opportunities as it is can I rank, but still the original name stuck!
CW: It still works well as a different way to present the information – feeding in the terms you feel are important, and then getting a checklist of ideas.
CanIRank gives out detail on next steps to take for on-page, off-page & technical SEO aspects – with so much debate on the best way to do things, how did you select the right advice to give? Is it all tied directly to the app’s machine learning?
MB: Yes, the algorithm learns which factors are most influential based upon what’s working well for each keyword. This is an important differentiator between us and other software that follows a standard “checklist” approach for every website: if we’ve learned one thing over the past few years, it’s that SEO changes extremely quickly. By the time something becomes generally accepted as a “best practice” and starts getting pushed heavily by all the gurus, it’s already on its way to being a penalty trigger.
From directories to reciprocal links to guest blogging, many SEO tactics have tended to follow a similar evolution:
Discovery -> Promotion -> Broader awareness -> Automation and abuse -> Penalty trigger
One or two experts stumble upon something that works, they promote it to help build their brands and expertise, it becomes more widely known including amongst spammers, someone figures out how to automate it or otherwise do it at massive scale, it’s now so widely (ab)used by low quality websites that it becomes a penalty trigger.
So we rely on the results to teach us what’s working, rather than expert opinion.
CW: What’s next in the pipeline for CanIRank? What do you hope the tool will eventually grow into?
MB: Our primary focus right now is on making our Action Recommendations more intelligent and personalized.
Whereas now we can identify a certain issue and recommend some specific actions that will address it, the next step iteration won’t just describe the action, but will show you specifically where you can go for your website. For example, if you need to build more Website Strength, we’ll be analyzing your competitors’ links to determine which ones might represent an opportunity for you.
Lots of actions involve something like looking for industry resource pages, or influential bloggers, or niche communities to participate in. Rather than just describe how to do that we’ll go out and find resource pages/ bloggers/ communities/ etc. and analyze them to determine which ones represent the best opportunity for you.
So in this case the AI will be used to answer questions like:
- Can I get a link on this page?
- Is it worth the effort for me to get a link on this page?
- How trusted/ relevant/ authoritative is this page?
And then presenting you with the best opportunities based upon that analysis.
So basically, new applications of our core theme: collect a ton of data, analyze it to determine what’s most relevant and actionable for your website, then present just the best stuff to you so you can focus on getting things done!
CW: Wow, that sounds like a hugely impressive ideal to reach for. I’m curious, how much does CanIRank’s ability to correlate with Google’s order depend on stability?
For example, has the recent Phantom update meant you’d had to reassess any of your learning? And what would happen if another Panda (the original affected 12% of English queries) was launched?
MB: If you think about it, most search engine updates don’t involve new signals, but rather new ways of processing and prioritizing those signals. So some of those shifts we’ll pick up automatically, others we’ll pick up second-order effects that still let us reflect the change. Of course, as search engines start to prioritize new signals, like https or mobile friendliness, we’ll have to add as many of those things as is feasible into our models.
But as hard we try, we’re not Google, and never could be. No one has access to the kind of data they have. We’re just trying to make the data that is out there and available as actionable as possible.
The cool thing about SEO is that it’s a series of little competitions. In competitions, you don’t have to perfect to win, you just have to be better than the other guy. So if (other things being equal, which they’re not!) a tool gives you even a slight edge, say increasing your productivity by 5%, that can mean the difference between page 1 and page 2, translating into a 1,000% increase in ROI.
On a side note, that’s why it’s usually worth it to hire the best SEO you can afford, since the difference between high rankings and low rankings is a lot bigger than the difference in their rates!
We’re just trying to make the data that is out there and available as actionable as possible
CW: I love diving deep into keyword topic models and user-led content development, which is one of the reasons I was excited to try CanIRank. The way we analyse keywords with lots of data, their intent as well as relevancy, is becoming hugely important. CanIRank also values understanding what works for a keyword in detail – do you think this is the future of keyword research?
MB: Yes, in any case it’s certainly been the biggest change in the past couple years, as the actual keyword string has greatly declined in importance, and the amorphous “topic model” and even more amorphous “intention” have risen. In some ways this has made keyword research a lot easier: you no longer have to analyze every keyword micro-variation (singulars, plurals, sizes, etc.) or try to find a less difficult variant by adding superfluous modifiers or awkward plurals like very cheap iPhones 6+ case…
Now the search engines just ignore your silly obscure keywords anyway, so you might as well focus on the real meat keywords, and instead use the extra time to dig deeper into truly understanding what it will take to rank.
We still have some customers who come in with the mindset of wanting to analyze thousands of micro-variations in order to uncover some hidden gem that no one else has thought of. And unfortunately a lot of 1st generation keyword research tools that only look at keyword string occurrence will tell them “yes! very cheap iPhones 6 case is way easier because nobody’s targeting that right now”. So they waste time adding silly pages to their website, and still don’t end up ranking.
CW: For me the rise of a topic model makes things easier, as you can concentrate on creating several excellent pieces of content, rather than being asked to develop tens of near-identical Panda fodder. It not only works for SEO, it just build better websites.
Finally, as someone who has been a keen observer of the industry, and has now jumped right into it, what do you think the future holds for SEO and online marketing? Is it data-driven tools working alongside creatively-minded brand building? And what do we need to be doing better?
MB: Well, it’s definitely going to be an interesting time! Marketers have experienced this unprecedented explosion of data. Where once we were swimming blind, now we’re all ADHD. When it’s literally possible to know more about your customers’ interests and behavior than previous generations knew about their own spouse, how do you even process that?
I think online marketing right now is still going through the awkward teenager phase, struggling to incorporate all this knowledge, or worse, using it in creepy ways. As it grows up, we have an opportunity to reach what I consider the ultimate pinnacle of marketing: Helpful Marketing. Marketing that informs rather than deceives, and helps you make the best purchasing decision.
Helpful marketing is some future version of FitBit or Apple Watch that can monitor my health metrics and say “Matt, you’re starting to get some inflammation in your rotator cuff that could lead to impingement, here are the top rated shoulder specialists in your area. Would you like an appointment?” Or my Tesla monitoring its own performance, automatically identifying any issues and suggesting nearby mechanics or appropriate aftermarket parts.
And perhaps some not-so-distant future version of CanIRank that monitors all of your online marketing metrics and can connect you to the appropriate experts for whatever challenges your site is facing.
So when we’re all constantly getting advice from Future Siri/ Google Now, our Tesla, even our refrigerator telling us when we’re out of milk, will SEO evolve into helping companies better meet the demands of these billions of new “implicit search queries”? I hope so!
In any case, for marketers it’s going to be a great time. Our data processing tools will help connect us to people who have the exact problems our clients are able to solve.
Only helpful and relevant messages will make it through the blockers and filters anyway (that’s a hole we dug for ourselves!) Nothing feels better than being able to authentically connect with someone and help them out!