Google’s campaign forecasting tools can be useful. The keyword tool, for example, remains the best indicator we have of likely volumes, short of putting the money down and seeing for ourselves.
But they’re not perfect (nor should we expect them to be) – and it’s worth getting to know where they tend to over- or under-estimate.
More often, Google tends to be overly bullish. In line with its general push for volume over refinement, and maximal algorithmic freedom, Google tends to overplay the upside.
The framing of its classic recommendation to “Get more conversions at a similar or better ROI by adding broad match versions of your existing keywords” (though not as silly as it used to be) is a good example of that.
But one important area that tends to lean in the opposite direction is in the forecasts around changing your tROAS or tCPA targets.
These are the ‘target simulator’ tables that show up when you click on the chart icon next to your campaign budget, or ad group CPA values, forecasting what you can expect in terms of cost and conversions – or conversion value – over the next week, for a given target input.
To interpret the table in this example, Google’s expectation for the coming week – if we make no changes to the CPA target – is 87 conversions for a spend of $4674. A CPA of $53.72.
If we want to push for more conversions and raise the target CPA by 10%, we’re told to expect an additional $341 cost, in order to achieve an additional one conversion: an ‘incremental CPA’ of $341, and a new total CPA of $56.99.
Pushing harder with a 25% increase will, we should expect, produce a further one conversion for a further $488.
Intuitively enough, each additional conversion comes at a higher cost as we seek higher-hanging fruit.
That pattern makes perfect sense… But I’ve become increasingly convinced that these forecasts consistently underplay the real likely gains from loosening targets… Especially with smaller changes.
How sluggishly conversions or conversion value will be dragged up as you loosen targets depends largely where you currently sit on the ‘incremental ROAS’ curve… i.e. how deep into the zone of diminishing returns we have already gone.
One good indication of that is simply ‘impression share lost to rank’. If there’s a lot more traffic to be had by bidding more, then in all likelihood there’s a lot more converting traffic to be had by bidding more.
But that’s a crude measure… and if it works well, Google’s target simulator should give us a more sophisticated insight into our potential to extract more value from a greater investment.
Why do I say that Google downplays the upside? It’s mainly from experience – which may or may not chime with you… and of course you can only take it so seriously without data to back it up… But we do have something of that data, courtesy of an excellent study by Mike Ryan’s Smarter Ecommerce.
But when we don’t have that ready source of new traffic available, the next move is to seek more raw material, by expanding our targeting in various ways.
The study tracked actual shifts in ROAS based on changes in ROAS target – giving us a much clearer picture of what typically happens when you adjust your targets (on average, and at the upper and lower bounds.
I’ve seen different interpretations of this data, but mine, from the chart alone, is that shifts up to the region of around 15% are pretty reliable when it comes to dragging performance along with it (and although it would be useful for the chart also to show actual effects on conversion or value volume, that can partly be inferred.)
My own experience gives me at least that much cause of optimism in the power to increase conversions/value by loosening targets.
So when I see a likely opportunity to grow by loosening targets, and then see Google’s very negative assessment of the likely impact…
I’ve come to ignore this feature – but to note that anyone taking these forecasts at their word would be missing out on likely opportunities for growth.
It’s important to get to know the likely effects of given changes in Google Ads.
And two great sources of that knowledge are the type of bid data study produced by Optmyzr, Adalysis and – as above SMEC… And of course (best of all) experience.
Sometimes this tallies with Google’s forecasts; sometimes it doesn’t.
And as we triangulate our data points to try and reach sensible decisions, let’s also not become so sceptical that we assume Google is always overplaying the upside. Sometimes the opposite is true.
