Stop forecasting, start optimising

Posted on 13. Jun, 2012 by in Digital Marketing



Have you ever gone back to a forecast produced before launching a search campaign and been surprised at how different the data looked from the actual results of the campaign? At how your “killer” keywords ended up being the ones that disappointed in terms of performance, while others outperformed any expectation and prediction? Agency incompetence or carelessness? Poor forecasting tools?

Maybe, but it is ultimately questionable that the time invested in producing a credible model, will pay-off with with useful and usable data.

First of all there’s the question about realistically forecasting search traffic volumes. Of course there are obvious seasonalities that can be considered and some of the search engine data has accumulated for a significant amount of time so that pre-empting some of them becomes feasible.

However, how interesting is this data really? In January I will sell more flights in the UK, just as certain as there will be a surge in retail sales before Christmas. B2B business will be slower in the summer months, etc., but these trends won’t generally look different in the search space compared to any other advertising space.

But let’s say I am looking to verify trends experienced in other spaces, within the search space. Is the best approach really to try and predict how many people will be searching for keywords I deem important for my business? Once I have some numbers for possible impressions, what will I do with that data?

If the search volume is there, I will still have no idea about the cost effectiveness of that specific keyword for me. What if I’m going after a set of keywords that will ultimately not convert users into buyers for my business (because the price is too high, because availability is low, because the path to conversion is more complicated than the competitions)?

Predicting impression and click volumes could be almost worthless information if not linked to actual business objectives and goals.

But let’s say my product offering is relatively stable, prices competitive and seasonalities consistent year on year. I can “trust” search volume predictions and rely on the fact that there won’t be any significant fluctuations in conversion rates.

How would I go about predicting the impact of external factors that are out of my control, but can have a massive impact on the performance of my search campaign?

What if one of the products I sell suddenly becomes “news-worthy” for bad reasons. What if as a consequence users search for it far more, but to find out more information rather than buying?

What if a competitor launches a massive TV campaign, that boosts search volumes and that I can unexpectedly piggy-back off to boost sales?

What if through word-of-mouth products I am selling become known by terms different than their actual definition?

None of the above is particularly unlikely within a search campaign and can have a massive impact (either way) on the campaign. Rather than spending significant amounts of time trying to “guess” what will happen in the search space, it is advisable to build flexible campaigns and to spend time optimising them for best results, regardless of what initial predictions may have suggested.

Tags: , , ,