Getting SEO done can be challenging, especially if you need to convince someone else that it is worth spending their money on it.*
The vast majority of people want to know what to expect before investing in a new SEO initiative.
It is hard to know what to expect from some SEO initiatives until you test them.
Yes, we can forecast and approximate based on previous work, but relying on that too heavily can over-expose the campaign or cause significant forecasting issues.
Using Proof of Concepts, Pilots, or Tests for SEO
Call it a PoC (proof of concept), a pilot, or a test; the end goal is the same, we create a small version of the main initiative to get sign-off to test it and prove it can work.
A “test and learn” approach like this is highly effective at getting things moving - something I’m a huge advocate of - but it isn’t just a case of scaling something down from 100% to 10% and then testing; it needs to be more structured than that.
The other essential step in the process - before you start planning how to implement the new test - is validating the approach to ensure you are setting yourself up for success.
A Mental Model For New SEO Initiatives
Running faster “test and learn” style approaches with SEO makes sense, but these will still take time, and you will only be able to do a finite number of them. Selecting correctly is key.
Each section of the mental model will help fine-tune or remove potential tests to give a robust list to work with and help ensure a higher degree of success. A spreadsheet you can copy and use for yourself can be found here.
Form your Hypothesis
What is it you are looking to change, and what do you think it will do? Expressing this in a hypothesis is usually the best way to start. You will need to confidently communicate this hypothesis to those who don’t understand SEO, so it needs to clearly show the value.
To borrow from Will Critchlow, there are only three SEO hypotheses:
Are new keywords being targeted?
Can the change improve existing rankings?
Will the appearance in the SERP change?
The details about the expected result—improvement or decline—and the how/why it might happen can be nuanced and different, but the essence is you want to test one of these three things.
If your hypothesis does not align with one of these points, it is likely not an SEO hypothesis OR it is not well-formed.
Detail - What is needed to run this test?
Can you describe what you are proposing, to a non-SEO? A developer or a senior stakeholder, for example. What are you planning to do and why? Where on the website will this change will be known and can you draw the paramaters clearly?
Creating a complete version of this detail description will take some honing and collaboration, but if you can be conscise and specific here it will be much easier to get others on board with what you need.
Do you have a Clearly Defined Idea?
This is the first “gut check” of the process. When you read your hypothesis and short description of what you are planning to change and what the results are, is it clearly defined?
Sometimes it is tricky to do this with your own work, so a peer review can help. At a minimum, you should get into the mindset of the person who will have to sign off on this effort and consider their expectations and objections. Is your hypothesis strong enough to withstand that?
Estimated Opportunity
What do you expect to gain if you successfully run this test? Or, if you are mitigating risk, what could be lost if you don’t? We won’t get into exactly how you calculate this here, but having a robust, well-researched, and believable appraisal of the opportunity increases your chances of getting anything put into action. Opportunity in monetary gain is the best, but click gain should be the minimum for what you are looking for.
Some SEO work is hard to approximate in click/revenue uplift. This work is most often best practices/housekeeping and may not have an easily reportable improvement attached. It is best to do this work as part of a portfolio of SEO best practices changes rather than a PoC or testing approach.
Metrics/Indicators for Success
Really simply, how will you tell if it has worked or not? This should be pretty straightforward for most SEOs, but my Metric Chain helps to frame this better if you are less sure.
Is the Change Reportable?
This leads directly from the previous step and establishes whether you believe you can report on your outcomes. Even if you know the metrics you are using, do you have access to them right now? Or do you think the change might be too small or hard to report on at all?
Required Team(s) to Fix/Implement
If you don’t have buy-in or resources from the other teams required to make the test a reality, then it won’t work. Changes like this are only real when realized, so speak with the teams you believe will need to help, and understand their ability and level of resources.
At this stage, you get bonus points for seeing how you can help their team’s KPIs/targets by assisting with this test. The best pilot projects make all the teams involved look good - investing time upfront is worthwhile.
Confidence Level in a Successful Outcome
This is the second “gut check” in the process. Based on each previous step, what is your confidence level that you will be able to positively demonstrate your hypothesis and get sign-off for a larger rollout?
If this feels like being asked to predict the future, you’re not; it enables you to establish your own level of confidence and helps you rank each hypothesis against each other.
Your confidence can and will change over time, so it is important you check back in on this if the facts change.
Finally, PIE Score
I’ve written about the PIE score before. It can be a powerful way to help compare recommendations numerically based on Priority, Impact, and Effort. This is a more three-dimensional way to review rather than just based on the potential gain, which isn’t the only factor.
How Many Tests Do I Need to Run?
The number of tests you’ll need to run will be linked to your goals, level of resources, and the opportunity you have identified—a kind of “it depends…” there.
But if you need to roll out 4 successful tests:
You’ll likely need at least 6 live tests.
You will have pitched nearer to 8-10.
Maybe even planned or short-listed as many as 16.
These numbers aren’t set in stone, but you should use this to set your expectations of the work you need to put in up-front.
Each of these tests is a small gamble, hedging your bets across multiple potential ideas, and validating each one will only increase your chances of getting them launched.
Long-list many ideas, refine and shortlist the ones you are most confident of, and only pitch the ones that add the most value and will not overwhelm people or resources.
Example - Internal Linking Improvements
Hypothesis - Adding new internal linking modules on L3 and L4 pages within site structure will increase relevant horizontal linking between pages, improving rankings of those target pages.
Detail - The site structure is highly siloed and pages from level 3 and onwards receive a very low amount of internal linking. These pages also receive a low amount of clicks from search, but they do present viable keyword targeting opportunities. Relevant internal links from other pages within the structure will increase the exposure to the pages and help Google see the relationship and increase ability to rank. We would introduce this to 10 sections of the site (rouhgly 10% of the total available) to prove whether this will drive a benefit when rolled out across the site whole.
Clearly Defined Idea - Clear (enough).
Note, the theory here is that internal linking is one of the issues impacting the ranking of these pages. There would need to be a portfolio of changes to help reduce crawl waste and index bloat to help better focus search engines.
Estimated Opportunity - Based on other incremental technical changes, this presents a 1-1.5% growth annually in organic clicks.
This sounds like a low number but this kind of growth from a template-level change could present a great ROI.
Metrics/Indicators for Success - Clicks & impressions to target pages and visibility of the keyword ranking group (core keywords tracked for those target pages) would help illustrate the improvements the changes could drive.
We can also track other indicators from a technical standpoint - Google bot hits to the target pages, number of internal links as measured by Screaming Frog, Site Bulb Etc. These can show technical progress, but they wouldn't cut it for senior buy-in of the overall test though.
Is the Change Reportable? - Yes - if successful. There's a risk that internal linking changes - if over too few pages - can be hard to measure against the usual background noise of search activity. We may need to run Causal Impact and Time-Based Testing against this beyond simple uplift of clicks + ranking.
Required Teams - Development teams would be needed to make a variation of key page-templates to introduce the new internal linking module.
For this test, we can manually dictate the internal linking, but for the full roll-out a programatic solution would be needed.
Confidence in Successful PoC - Medium/high
There is a great sense that internal linking is clearly lacking - and not as strong as other competitors - but whether this is enough to move the needed does need a test & learn approach.
PIE Score - 6
Priority - 7
Impact - 6
Effort - 5
The score here isn't as meaningful without seeing the scores of the other potential tests, but a 6 means it'd be considered medium/high in priority. More confidence of the impact (i.e. more research) or a reduction of effort in implementation would make it more appealing.
Example - New Content Page Creation
Hypothesis - Introducing expert-lead, inspirational content help gain non-brand exposure whilst customers are in research/consideration phases. This would drive additional mid-upper funnel traffic where currently there is none.
Detail - We would work with the SME (subject matter experts) within the brand to make five, brand-new content pieces to live within the pre-existing content area. We will use product reviews, live chat questions, search trends and conventional keyword research to identify questions which we can answer especially well. The emphasis here is on content that no one else can write about or answer as compellingly.
Clearly Defined Idea - Clear
Estimated Opportunity - Key competitor pages are driving 18,000+ visits p/month with equivalent content.
Whilst we're not going to push to position 1 over night, there is a clear feeling that we can and should be able to outrank the alternate pages eventually. A great aspiration and most will agree we should be able to achieve it.
Metrics/Indicators for Success - Traffic driven to new pages, incremental revenue generated (via assisted conversions).
Most reports report on last click revenue, but as these are top/mid funnel pages we will struggle to show this as part of the PoC. Using assisted conversions will help tell the story of the success.
Is the Change Reportable? - Yes
Required Teams - Content team, brand sign off, legal sign off & CMS teams to add the copy.
The requirement for brand and legal sign-off makes what is - on paper - a really great opportunity much more likely to get delayed or stoped altogether.
Confidence in Successful PoC - High
Keeping the number of pages small and getting strong buy-in from the start should be enough to overcome a lot of the resistance
PIE Score - 7
Priority - 8
Impact - 7
Effort - 5
There's some real potential to drive incremental traffic gains to top/middle funnel. Just to note, we need to strengthen the attribution/reporting to help tell the story of the importance of this traffic.
To Conclude
Successful implementation of SEO tests requires careful consideration of goals, resources, and identified opportunities. Utilizing metrics like the PIE score and conducting multiple tests can enhance the chances of launching impactful initiatives.
By prioritising ideas, refining them, and selecting those that add the most value without overwhelming resources, teams can maximize chances of selection and implementation. How have you gone about getting buy-in recently? Have you tried someone else altogether? Is this just not an issue and are you in the fortunate position of having less friction in your backlog? Please reach out in the comments below.
Great idea.