I have been forecasting *a lot* recently. It’s a part of my day to day as an SEO that’s become more and more commonplace over the past few years. I think, perhaps, as SEO continues to shred it’s reputation as a dark art and starts being taken seriously as a very worthy, grown-up marketing channel, where an SEO is give responsibility to make real business decisions, we’re being KPI’d more than ever.

Which, I think, is ultimately a good thing.

OK so *everybody knows* that SEO forecasts are pretty much just educated guesses. Everybody knows that we can’t foresee competitor activity or Google changes, that CTR rates are generally inaccurate and that we really do struggle to say if a piece of content is going to really drive value and we all know that attributing growth to specific SEO activity when there might be several other things at play is a little bit like claiming it’s only the bricks that build a house and forgetting the like…erm mortar and frames (?) and stuff. *(I just made that up and that is the most laughably bad metaphor ever and I clearly need to learn the basics of construction).*

However! I still feel that as a channel forecasting gives us credibility, makes decisions to invest in SEO easier and means that we can actually define what success looks like rather than just blindly saying

Hey! Client! You’re number 1 now for your hero term and your traffic is up! Pay me more! (

Kirsty Hulse 2011)

Just because something is difficult, does not mean it is not worthwhile.

Firstly I’d like to call out something which I think needs to be defined. For me, there is a difference between an ‘opportunity analysis’ and a ‘forecast’. As far as I see it an opportunity analysis is when we take a KW set, look at search volume, take average CTRs and say “if we get to 5th position, then we might drive X traffic”. This is looking at the opportunity within the market, it’s excellent for winning pitches and getting an upsell because this can often be very compelling data. However a forecast (IMO) is something a little different; a forecast is looking at historical data, other channels and on this making a *mathematical assumption* of what can be achieved. The opportunity analysis data can be used to feed in to the forecast, but alone, I do not feel as though this is enough.

Secondly in the examples below I am considering only organic traffic. The same, in theory, works for revenue, though swap your CTRs for CRs.

So based on that, I thought I’d share with you how I’ve been doing it. This is not the only or right way, just a way that’s been working for me.

**Getting your Baseline**

So first, we need a baseline forecast which we can then tweak to account for trends, seasonality and growth.

Take your historical data (in this instance looking at organic traffic) and look for any outliers that may skew the accuracy of the forecast.

For the example above, although there’s typically always a peak in January, this was unusually high due to massive investments in TV and display campaigns that year, so, unless those same investments are being made next year, we need to adjust this to the traffic spike which is more typically seen in January *without* investment.

(There is a clever mathematical way to identify statistical outliers, beyond my unempircal method of just “looking at the graph”, which is beautifully explained here).

So we have our data, now we need to start making some projections.

Excel is, as ever, is our best friend. There’s two ways in which you can super quickly create a forecast based on mathematical assumptions.

1. Using the liner trend data

When you add a linear trends line to a graph in excel, excel creates this based on an equation, we can then take that equation and use it to map out what that trend will continue to look like in future, **all things being equal**.

Then pop that equation in against your actual data like so and you are left with the continue progression of that data sets trend. Cool.

Option 2 is to use the forecast function within excel, which is:

The FORECAST(x, known_y’s,known_x’s) function returns the predicted value of the dependent variable (represented in the data by known_y’s) for the specific value, x, of the independent variable (represented in the data by known_x’s) by using a best fit (least squares) linear regression to predict y values from x values.

If you assume that data pairs are plotted in a scatter plot with x values that are measured on the horizontal axis and with y values that are measured on the vertical axis, FORECAST returns the height of the best fit regression line at the specific value x on the horizontal axis. FORECAST is the value of y that would be predicted based on both the value of x and the regression line (characterized by its slope and intercept that can be found by using Excel’s SLOPE and INTERCEPT functions).

I have no shame in saying that I don’t really understand what any of that means. I equally have no shame in saying that I trust Excel to know what’s it’s doing and am more than happy to blindly follow it. Like so:

So now we have 2 baselines which have taken mere moments to create. In some cases, this might actually be enough. However what this is missing is connecting this to SEO specific activity, which is one of the most crucial pieces to keep clients happy, so naturally is the hardest part.

**Adding what you know to the numbers**

OK so this is where the obstacle course of variables come in, so there’s no set way of doing this. Though a few things to consider:

Do you have an increased budget?

If so, this is where an opportunity analysis can come in handy. Look at where you think you are going to get your wins, take the traffic projects based on search volume and CTRs and add these traffic projects to the baseline.

Any penalities?

If so, this is a fairly easy one to consider, where was it before? When will you hope to be fixed? Map on to the baseline over a time period you’re comfortable with.

Creating new pages?

In this instance let’s hope history repeats itself. At some point in the websites history, new pages will have been added. What happened then? If not, take a look at your opportunity analysis for the KWs relating to those pages.

Running activity on other channels?

This is crucial. Make sure you understand the interplay between SEO and other channels. I personally have seen big impacts on organic when display or TV is running, be sure to get historical and future media or marketing plans.

Seasonality?

Are there big Christmas spikes? Or any other seasonal peaks and troughs? These will likely remain relatively consistent year on year, so you can take these from your analytics package or gain an idea from Trends. Also consider industry news and/or events that may have an impact on this.

And lastly…

**But is it accurate?**

Well, as accurate as it can be. I did exactly the above 6 months ago and this is where we’re currently tracking on forecast vs target

At the moment we’re 99%. So, not too bad at all.

Again we all know this is really hard and not an exact science, but wanted to share where I’ve got to after slogging at it for months. My main bit of advice, try and remain as true to the baselines as possible and only add in additional things you are confidant (as you can be) on.

Oh, and of course, **never forget your caveats**. 😉