The need to develop an SEO hypothesis
This week’s newsletter is sponsored by the Digital PR agency Search Intelligence. See their case study linked at the end of the newsletter.
Are you interested in taking a cohort-based course for product managers to learn about SEO that I am building with Kevin Indig? Please share your insights in this survey to help inform what we should teach and join our waitlist.
Everyone likes to quickly assume any traffic decline was due to a Google penalty, while on the flip side, they assume a traffic increase is due to the great work they have done.
This is the wrong approach, and there’s a better one that I like to use.
Declare a hypothesis
Before conducting any traffic decrease or increase analysis, I have the collective group (or myself if it’s a solo analysis) declare a hypothesis. I then work to support that hypothesis with data. If the analysis supports the hypothesis, only then would I determine that hypothesis to be true.
To be completely thorough, I might develop a second hypothesis and go through this supporting effort again to ensure no other cause was missed.
Don’t be lazy
When the data does not support the hypothesis, the hypothesis is discarded, a new hypothesis is generated, and the analysis process begins again. It is supremely lazy and highly inefficient to assume an SEO cause and then run without confirmation. Making incorrect assumptions will also inevitably lead to further SEO pain if, in fact, there was something else broken that needed to be fixed, and no one even looked.
Here’s my process for setting up this hypothesis analysis process for SEO.
Define the hypothesis: Make a defined hypothesis that uses detail and is timebound. A poor example would be “Google doesn’t like our site and penalizes it.” A much better hypothesis would be that on 11/1/23, we experienced a severe decline in Google traffic due to a Google manual penalty.”
Time as a boundary Since the hypothesis is specific and timebound, only evidence that fits in that hypothesis window would be relevant. Traffic declines before the stated date should be ignored, and the specificity would allow the hypothesis to be supported if there is no conflicting evidence.
Create a view: Based on the specificity and times in the hypothesis, develop a view of what the data should look like if the hypothesis is correct.
In the example of a manual Google penalty on 11/1/23, traffic should be decreased to the penalized URLs only on that particular date. Manual penalties are not rolling like algorithm updates, so the time window for this event would be no wider than 72 hours to account for visibility in different data centers.
To restate again, the traffic declines should be dramatic, and the URLs should either not be indexed or not visible anywhere near the top for previously ranking queries.Discard the hypothesis: In analyzing the data for the chosen hypothesis, findings that run counter to what is expected of this assumption should cause us to question whether the hypothesis is accurate.
This is the moment where I usually find that an assumption about a Google penalty usually falls apart. While traffic might have declined, the URL is still visible for the exact keywords before the traffic drops. Rather than a penalty, this would be more symptomatic of seasonality or any significant change in demand. An actual penalty would make it very hard to find the URL, or they might even be deindexed.Create an alternative: As this supporting data is analyzed, this is the right time to develop a new hypothesis if the evidence does not support the original one. In our penalty example, seeing that the URLs are still indexed and maybe even visible at the previous positions might suggest that the new hypothesis is that there was a seasonal traffic decline.
Rinse and repeat
This process will repeat every developed hypothesis until the collective group is convinced that the data supports the hypothesis. For completeness, here are the supporting data points for a seasonal hypothesis.
The URL’s need to be still indexed at about the same rate
The URL’s need to be visible in a similar position to where they were before any traffic changes
Ideally, an external event should be identified that would explain the seasonality (a holiday, occasion, trend, or news story as examples)
The seasonal changes should be visible on similar sites when researched using SEO tools. The hypothesis might be invalid if the other sites don’t have the same traffic changes.
There are many advantages to using this process versus just making assumptions; however, it is not just about being more thorough. Having a solid, proven hypothesis makes it a lot easier to communicate to executives what is genuinely happening. Furthermore, when the data supports the hypothesis, the executives will be comfortable making these same assertions to whomever they communicate to, like board members or even Wall Street earnings calls.
Budget requests
The hypothesis approach will also make it a lot easier to make successful requests for resources, even in a hostile environment. CFOs might not be willing to open up the purse strings if they think SEO will keep declining, but if the declines can be explained away in a logic-supported story, the CFO will feel far more comfortable making investments.
Most importantly, using a hypothesis-driven approach will allow any SEO team to learn about how various actions cause reactions, and then those learnings can be deployed for future growth. If an SEO team throws up their hands and declares that they have been penalized or that it’s just seasonality without looking into the data, they have not learned anything of value that will help reinvigorate growth.
Of course, this works on the inverse, too. Every traffic jump should be just as rigorously investigated so these jumps can be sustainable, repeatable, and serve as lessons.
[Sponsored]
We started working with this e-commerce health client over a year ago.
The first months were slow, as it is with many PR campaigns.
But then the ball started rolling.
Throughout the year, we’ve secured over 230 links and placements to their website, in top-tier national and regional news outlets.
These are the campaign types we leveraged, in order to get the links flowing:
Data-driven PR
• We analysed official data-sources related to health
• We normalised and sometimes combined the data with other sources
• We created unique angles that we knew the news outlets would love
• We pushed the stories out to hundreds of health journalists
Expert commentary PR
• We wrote up expert tips from our client related to health
• We pitched it to journalists who write about health
Read the rest of the case study
