This interview is part of Kameleoon's Expert FAQs series, where we interview leaders in data-driven CX optimization and experimentation. Alex Birkett is a co-founder at Omniscient Digital, a content marketing & SEO agency that works with B2B SaaS companies. He lives in Austin, Texas, and writes at alexbirkett.com.
I have a massive amount of content live on my site but optimizing it all one by one is very time-consuming. Is there a way to do it at scale?
Yes and no.
You can audit at scale, and you can sometimes group pages together if there are similar issues causing performance issues. This is common with technical issues.
But it’s similar to any sort of conversion rate optimization process: you start with research and diagnosis, you move towards prioritization, and then figure out your action items (experiment, fix it, more analysis, etc.).
When it comes to content optimization, you’re typically trying to optimize for one of two outcomes:
- Drive more traffic
- Drive more conversions
There are other secondary goals, like improving readability or user experience, but these goals tend to ladder up towards increasing conversions or traffic at the end of the day.
So I take each of these goals as a separate process. To drive more traffic, you’ve got to increase search rankings. So the most valuable thing you can do during your research phase is to find content that is either:
- *Almost* ranking for key terms
- Is falling in rankings and traffic
For content that is *almost* ranking, you just have to plug your blog or domain into Ahrefs and filter for pages that rank in positions 5-30 (your range could be different, but I like to start here), and then prioritize by the highest traffic or impact:
If you have a lot of content, it helps to pull this over to a spreadsheet, but you’ll probably already see some opportunities to update content. Like, here, I can see that updating my articles on “email blast examples,” “best newsletter software,” and “best keyword tracking” would be advantageous.
For falling traffic, the easiest way to identify this early is to use a rank tracking tool like Ahrefs or Semrush. Set alerts and proactively update pages that seem to be falling (avoid whiplash by looking only weekly or monthly, as there are short-term fluctuations).
You can also use your own Google Analytics data to compare time periods. Filter for organic traffic, so you don’t include social or direct spikes, and look at the current time period (say the past six months) and compare it with the previous period. Doing this, you have to correct for seasonality, but sharp drops can still be observed.
As Matt Gershoff explained, with every personalization rule used, you have a new experience to manage and maintain. There may be marginal benefits to unique offers, but at scale, it becomes more expensive to manage.
Plus, to run experiments on CTAs, you often need to group together tons of pages to get an adequate sample size. For example, on the popups on my personal website, I run the same one across all my blog posts and use that to experiment with new offers. If I have a specific high-traffic page that differs in intent from the rest, I’ll create a unique offer for that page. But typically, I rope them together.
How can I use the “What If” analysis to identify content optimization opportunities within my old content?
By “what if” analysis, I mean if you build a growth model, you can do some scenario modeling to determine where your time is best spent.
You’ll need access to some historical data as well as some assumptions. For a content growth model, these could be:
- Keyword search volume
- Average SERP position
- Assumptions on SERP position CTRs (averages available here)
- Historical average conversion rates
- Lead to customer conversion rates
- Average lifetime value
Obviously, these could be different depending on the type of business you work on and your go-to-market strategy (for example, ecommerce would operate under totally different assumptions, as would, say, a freemium SaaS product or an enterprise sales-led product).
But once you have this data and assumptions, you can estimate the relative impact on your revenue if you were to focus more attention on creating new content (raising overall traffic) versus working on increasing your lead-to-customer conversion rates.
At a certain point, you should be working on all of these nodes in the system, but you’ll still have to determine how many resources you allocate to each bit of the funnel.
It also helps you determine, based on the costs of content production, if and when you’ll be able to break even, given the assumptions in your model (i.e., average SERP position, search volume, etc.).
I’ve modeled traffic assumptions for prospects and learned that it would be an incredibly long time before we hit breakeven based on the costs of hiring our agency. In these cases, I refer them to more affordable freelance writers.
All you’re doing here is building a model, one that is close to representing the reality at hand, in order to determine where to best allocate your time and resources. Even if we pretend otherwise, we all have to make tradeoffs and incur opportunity costs.
What does optimizing even mean in the context of readable content—have readers' preferences changed in recent years?
Anecdotally, much attention has shifted towards short-form content and video. But that story is too simplistic.
My personal website is filled with 5,000-word articles that get read and bookmarked, and our podcast is long-form and is picking up steam. But I deliberately created content like this to attract advanced readers and listeners.
I will say at our agency, we’ve shifted a lot of our focus to capturing attention, at least initially, with short-form social media content - namely, LinkedIn posts and Twitter threads.
I don’t know if it’s necessarily a shift in how readers consume content. Rather, it’s where they’re spending their time, and you have to mold your message to the medium.
When writing blog posts, the same standards apply, whether you’re writing short-form or long-form content: make it scannable, no walls of text, an image is worth a thousand words, and the copy has to be substantial and engaging.
You talk about diversifying and rebalancing a content portfolio using a variation of the Barbell investing strategy. Can you talk me through how to apply this strategy?
The main idea of the Barbell Strategy is that you should bifurcate your approach to include,
a) stable and predictable bets and
b) volatile, “high risk, high reward” bets.
Each piece of content you publish should fall clearly into one of these buckets, even if it has other benefits.
Your volatile bets will typically be aimed at garnering backlinks, social shares and generating conversation. These are less trackable, but they feed into the flywheel to make it easier for your stable bets to rank.
For example, if I were running content marketing for a company that makes AI copywriting software, my stable bets could be listicles like:
- “Best AI copywriting tools”
- “How to write landing page copy”
- “[competitor] alternatives”
Depending on your stage and site maturity, you may be able to rank for these right off the bat, but you probably need to accumulate links and build up your own site’s authority in relation to competitors.
So your volatile bets may be contrarian takes like, “Why Every Marketer Will Use AI Tools by 2025.” Or you could compile research to create an industry report, like “The State of Copywriting 2022.”
These pieces may also drive conversions, but their primary purpose is to drive attention (and links), thus making it easier for your stable bets to drive ROI over the long run.
Earlier stage companies, especially in competitive industries, usually need to lean more into the volatile and buzz-worthy content. Companies with a stronghold can generally build most of their portfolio on stable SEO bets.
Should I run A/B tests on blog posts that have ranked and receive a lot of traffic? How can I do so without messing up SEO?
Most marketers shouldn’t be running A/B tests on blog post content.
Most blogs don’t have enough traffic to warrant A/B tests. If they do, your experimentation will likely fall into one of two buckets:
- CTA experiments to drive more conversions
- SEO experiments to drive more traffic
In the second bucket, your goal is to mess with SEO. But you need a lot of traffic and a lot of similarly templated pages in order to run controlled SEO experiments (think: Pinterest, Thumbtack, etc.). You can still run quasi-experiments using something like Causal Impact, but the marginal benefits are probably not large enough to warrant the time spent versus simply using your best judgment or tools like Clearscope.
For the first bucket, you normally don’t need to change any copy on the page itself. Rather, you experiment on a CTA button, image, or popup and test different messaging, offers, or placements. These types of experiments, by and large, shouldn’t impact SEO.
I noticed you have learned to code. Do you think it’s a worthwhile skill for people in “non-technical” roles? And how should a newbie start down the coding path?
100%. One of my favorite essays is Simo Ahava’s “The Myth of the Non-Technical Marketer”. An excerpt:
“The whole polarization of non-technical vs. technical is silly and artificial, and nothing irks me as much as this constant undervaluing of the human capacity to learn new things. Code allergy should be a thing of the past by now.
Why not instead embrace the fact that our industry is rife with opportunities to not only understand more about the technology stack we work with, but also to combine this technical know-how with our marketing skills for some true hybrid insight?”
I’ll add a few points:
First, learning to code is fun and teaches you how to think and build stuff. It opens up new potential solutions to problems you may never have thought of before. It’s a new tool in your toolkit.
Second, depending on your role, you’re almost certainly going to prove yourself more valuable if you’re technical. If you work in CRO / experimentation, this is undoubtedly the case - particularly with analytics and front-end development.
Even for content marketers, the ones that are the most valuable are the ones that can whip up a landing page, edit HTML, crawl a website, and write SQL queries.
As for advice, start with a project. I tried and failed many times before I finally had an actual project I was working on. Luckily, Peep Laja gave me some leeway to learn R to build data-driven user personas (though not too much leeway – I had about a month to put it all together). Having a tangible project to work towards gave me guardrails instead of trying to learn and conquer every aspect of the language.
Second piece of advice: CodeMentor.io. You can hire code mentors to review your code and help you get better. This was super helpful for me when building the SERP Tracker at HubSpot.
Final piece of advice: learn SQL. It’s useful in almost every role, and it’s surprisingly common that PMs and marketers don’t know how to write basic SQL queries.