For some, forecasting is dull as dishwater, others see it as an opportunity to explore what impact you and the team will have on the metrics that matter - and if you don't have the resource then this is the slide to show the opportunity cost of not delivering on the strategy.
Approach to SEO Forecasting Has Changed Massively Over the Years
Over the last 25 years of working in house, agency side and contractor there have been a huge variety in terms of how businesses view the importance of forecasting and how I was held account for the numbers. In the early days it was exclusively retrospective: you say you're going to do some stuff - you do a monthly report and comment on month-on-month and year-on-year change. That was it.
Latterly in my career much more importance was placed on predicting the outcomes of activities so resource in other departments can be aligned to expectations - think buying, logistics etc.
In larger organisations it comes down to marketing budgets and how bets can be placed where the right balance between risk and reward can be placed - relatively accurate forecasting is central to this need.
It's a Crazy World
I say "relatively" because of the world we live in. Who knows what's going to happen next? Although we live in a predictably unpredictable world where an economic downturn seems to happen every 10-15 years, we haven't had a pandemic for a century and Europe invaded for 75-odd years. I can't even remember when the last time a ship got stuck in the Suez Canal - blocking much of global trade and preventing ecommerce stores getting resupplied as quickly.
Demand shifted quickly during the pandemic. Fashion was down as people weren't going out and people spent more on doing up their homes. There were various stages of ending the lockdown so there wasn't a step change of 0% locked in our homes to 100% freedom. Remembering the nuance of what happened when and how that'd affect the numbers of market demand on daily-on-year impact is almost impossible.
How Should You Approach SEO Forecasting?
There's no right answer to this question as there are many purposes when it comes to forecasting:
Do you need to validate your plans? And is that a strategic or tactical level?
Do you need to isolate different growth rates for parts of your sites?
Do you need to identify risks of certain environmental and technological factors (algo risks, market risks, platform risks)
For all of those objectives its exceptionally important to get a view of historic performance before you even start to build your forecast. Jumping straight into producing a projection when you have no context of past performance will only give you a wildly inaccurate performance.
Download the top level data for at least the last couple of years, chart it up and annotate with significant events relevant to the data set - that could be the various stages of lock-down, replatformings, noteworthy technical events (we noindexed this entire section), noteworthy marketing events (we turned on paid shopping and it took 15% of our attributed traffic), resource events (we onboarded a link building agency and they started getting links about now or our focus was on delivering the new site and we focused less on the old one).
This approach will give you the right grounding in historic performance so when you come to either reviewing a sophisticated forecast or doing a simplistic one of X+8% then you know what's going on and can adjust your numbers.
Finding the Right Way to Skin the Cat - For You
SEO forecasting can be contentious as different people will see it in the sense of they'll see it as a nail as all they have is hammers. If you're a person that loves data and wants to spend months pulling data together and using sophisticated forecasting methodologies then you'll have some great spreadsheets and code to show people - it may be more or less accurate than replying on a good, logical sense or feeling for what's going on. Remember if you're spending months on this sophisticated model then you're not driving change.
And this point is exceptionally valid. If you read the book Range by David Epstein then you'll understand at how poor forecasting can be when done by experts. Generalists - those without a deep ability to create "sophisticated" models often miss what is blindingly obvious to people that have a wider approach to a subject area as they tend to apply the same approaches to problems, while generalists are more willing to try different approaches and have a much wider optic - they're less blinded by how they've historically solved problems.
When it comes to forecasting for a department level strategy when senior leadership is often well-aware of how things rarely don't turn out as planned over a year or longer. If that's the case why spend months delivering a forecast with utter confidence on the technical approach when the world is changing so fast? Think: progress over perfection - spend most of that time actually identifying specific problems to be solved and bring that into the definition or iteration of the plan.
What Are The Different Approaches to SEO Forecasting?
Past +%
You could view it on a past+% approach where you break down all the different sections of your site - whether that's homepage, product listing pages, product details pages, content pages, support pages, functional pages for ecommerce sites (perhaps by territory) or a different approach for non-ecommerce. Then view your plans for each page type - i.e. your technical delivery for PLPs may be extensive but your backlog for PDPs is much slighter. Or your plans for a content section is extensive so you may say we'll phase in a net 100% increase.
Get that views of expected deliverable and implementation dates (delivered isn't the same as implementation - one is you've got the data or created the capability, the other is getting it in front of search engines and customers)
Also understand the wider marketing landscape and what other plans are due to be delivered by other teams and how that affects your assumptions.
Keyword + CTR Level
Another approach is keyword level. You may say that accross the board you think you should be able to move all rankings up by an average of five places due to technical or content enhancements, but you may say for these keyword groups, say around key categories may improve by an additional three places because they're central to a set of link-building initiatives.
Using keyword + ranking data with CTRs etc and the curve of position and those CTRs specific to your keyword set and industry you could forecast what would happen to the CTR and clicks if you moved up by those five (+three) places.
It won't show you the potential impact of adding largescale, new content sections to show the potential gains because you're not showing the impact of the potential traffic on keywords you don't currently rank for - or aren't in your dataset at least.
That's more sophisticated than the Past+% approach but there are a great deal of assumptions being made here about the transition of rankings between page types, intent etc.
Conversely, it does help you play around with scenarios for different initiatives and the relative importance of different scenarios of allocating resource.
Detailed + Sophisticated
Whereas the Past+% approach and simply applying a phased growth (or otherwise) to historic data, overlaying your insight around the environmental and internal factors that could have impacted that previous performance has the benefits of simplicity. It would smooth over more granular insights that would be useful. The Past+% approach is best suited for monthly forecasts - but you can bury many dead bodies in a monthly forecasts. A more sophisticated approach would be to get daily or weekly data and then forecast using a robust forecasting procedure such as Facebook Prophet which utilises R and Python to give potentially more accurate views of how the metrics could move over these more granular data points.
This approach has the same caveats as the others, that you've got to understand what drove past performance as you'll be forecasting based on past performance - it doesn't just pluck numbers out of the air and be a rock solid view of the future.
You may have to build in a dampening or fortifying factor based on these wild events - but it should give you the ability to much more adeptly identify the causes of these events.
This 3rd approach is excellent at individual keywords (with sufficient data) or groups of keywords which the first just isn't capable of - its too reliant on subjectivity and not enough of the objectivity that reliable data provides.
Remember the Why
It all comes down to the 'why' you need to start there. If you need a broad, directional view then the first approach seems more sensible. The second would be reasonable when you're covering the specific pillars (e.g. content, links and technical) and the third more appropriate when you're in the delivery phase and you're in that start of the quarter and you want to build your weekly schedules and want to know where to place your bets. But can also provide valuable directional insight for those pillar plans that you can tentatively roll-up or quality the overall view of the impact of the combined strategy you see it as at the time of delivery to your audience.
I'll cover how to present the data, insights and priorities in another post, but I wanted to get across the message of making sure that the level of insight, granularity and sophistication should be greatly aligned to the purpose of the forecasts and that there are opportunity costs of progress when it comes to sophistication.