Facebook Doesn’t Work – Understanding the metrics

One of the biggest misunderstandings on Facebook is around what the analytics data mean. All too often users respond to the wrong metrics but then this is universal rather just a Facebook problem. Even for those who have been fortunate enough to be an Ad Campaign Manager, Account Manager or Data Analyst can be guilty of incorrectly reading the data. It’s a little like Comprehension in English, there is an obvious and direct meaning then there is the underlying meaning, what is inferred.

The optimisation of a campaign simply refers to the action you take based on the insight understood from the campaign data. Your actions will be based on the campaign goal and while this sounds like an obvious point, the wrong optimisation decisions are regularly made because of looking at the wrong metric in relation to the campaign goal. A typical scenario would be when driving low CTR but high conversions or generating a high CTR but low conversions. CTR is often used as the indicator of the health of campaigns. However, in the first scenario above where the campaign might be seen as failing, what should be looked at closely is what is driving the conversions. Interests keyword analysis is good example of how to identify this or looking at elements in the different ad creatives. Once you’ve eliminated all the things that aren’t working, build on what is. In the second scenario, there is an indication of right audience and message but not what they are expecting – the messaging needs to change to relate to the product. Or if it’s a product to be downloaded, is there a problem when downloading – a UX issue.

If you are a Fashion eCommerce business, there will be multiple factors that you will be observing. If, say, the objective is to generate sales some of those factors will include the customer acquisition cost, customer retention, customer reactivation, average basket value, lifetime value and all these things differ by product category, discount and full price sales etc. For example, you may be achieving a very low CPA during August but the average basket value is low in comparison to existing customers and 6 months later those customer have lapsed only buying once or twice. As this is Sales season, whilst you’ll acquire lot of new customers and a higher volume of sales, the average basket value will always be lower and the likeliness of retaining a significant percentage as loyal customers is very small. These are discount driven customers shopping not out of necessity or loyalty but price atractiveness. As mentioned in a previous post, it is important to understand your customers (behaviours and drivers). Create a relevant Custom Audience to cluster these customers together plus create Lookalikes and target them during your Sales period only.

Or certain product ranges may have a lower average basket value but the volume of sales are high more frequent hence a lower cost per sale. This is true of basic items like T-Shirts so the focus will be on driving volume of sales using Lookalike Audiences to find similar profile customers and making sure you understand how frequently these product ranges are bought to make sure these customers are coming back to you when they need to replace or add more of the same.

Again, both of the above may sound very obvious but these facts are frequently overlooked.

One of the other most misunderstood factors is around attribution. Many are unaware that Facebook has a 28 day attribution window which is why it is so important to hold your nerve with Sales, Downloads and Lead Generation campaigns. Plus, continue to observe the analytics information for another 28 days after the campaign has officially ended (7 – 10 days if is it’s gaming, if a prospective gamer hasn’t downloaded and taken whatever action you want of them within this time frame, you’ve lost your opportunity. Some gaming companies will swear to an even shorter 3 day conversion window). For those using a third party analytics tool, this inevitably results in contradictory data due to differences in definition of an action, the attribution model etc. For instance, when the ability to track downloads was first introduced on Power Editor, the download numbers never matched up with the third party mobile analytics providers that Facebook partnered with or with internal downloads numbers. Why? Facebook defined a download as when a user clicks on the download button whilst the accepted universal definition is when a user completes the download process. The importance of understanding the difference in definitions of actions, processes and behaviours by different tools cannot be overstated. It prevents unnecessary grey hair.

Whilst on the subject of attribution, I’ll make a small point about attribution modelling. Again, this easily creates false positives. For instance, in considering the Cost per sale, depending on the different campaigns that you ran to acquire that sale, you may have to also include CPA + frequency + retargeting. If you are running a campaign across multiple channels and the user came through multiple touch points are you attributing the sale or customer on a last click wins basis or are you apportioning the different channels a percentage. If the latter, how will these portions be weighted?

There are multiple factors that affect your Facebook advertising campaign but understanding what the analytics data is telling you and how to act on in it is key to success.