I’ve been of the belief that in order for someone to be deemed ‘good’ at PPC, one must have the ability to effectively analyse the data at their disposal. Then, most crucially, act upon what they’ve learnt in order to bring positive change to the account they are working on.
Trust the data, use it smartly and you’re on to a winner. That’s what every piece of content on PPC ever written (including anything you’ll find about the subject on this site) will tell you.
But how much data is considered enough to act upon? How many impressions or clicks or conversions does a keyword, AdGroup or campaign need to accumulate before we can gain any reliable, actionable insights? And in what time period? When does data elevate from the murky doldrums of coincidence to the divine meadows of absolute truth?
What is PPC data’s ‘golden number’? I remember a colleague once telling that in statistical studies, only sample sets upwards of 1000 are used – so, for argument’s sake, we’ll stick with 1000.
Now let’s think of using 1000 as a minimum sample set when analysing PPC data. 1000 impressions aren’t difficult to generate so establishing whether campaign settings, keywords or ads are relevant is doable.
But then we look at other key metrics; an account or campaign, even an AdGroup, generating 1000 clicks – feasible. But one single keyword or advert producing 1000 clicks over 30 days? In most the accounts that I’ve managed, it’s rare for any more than two main keywords to trigger over 1000 clicks. And when talking about conversions, that becomes even more of an anomaly.
Therefore, that means that when working on a conversion-focussed campaign, the insights seen as statistically sound would be very much to do with campaign settings, such as location and time of day targeting, which can be extremely effective in its own right. However, in my own experience, the more you can affect an account’s most granular entities – those being keyword selection, bids and AdCopy – the better chance you have at improving performance account wide.
Some could argue that the likelihood of reaching 1000 whatevers increases the longer the time period that you select to analyse. But PPC is a channel that is built on being responsive and producing results in a short time frame. So one must act quickly on any insights gained.
So does this mean that anyone working on an account should only take actions once 1000 is reached? Considering all the above, that would mean little much could be done to an account other than at a campaign level. If aiming for statistical accuracy, does one run the risk of being crippled by their data (or lack of)?
PPC doesn’t work like that, however. Liberties are taken with data; each manager setting their own thresholds for what they consider enough. And so, the insights gained and the decisions made off the back of them are not truths but nothing more than calculated guesses.
This is the way one must work in order to, well, do some work.
So it bares the question; is PPC and data as closely married together as we are led to believe, or is it just a game of gut-feelings and restrained risks?