Monday, 6 March 2017

Breaking down the Facebook auction: How to manage rising CPMs and deliver sales

Facebook advertising has come a long way in the past few years, and it provides a highly profitable way for brands either to engage an existing audience or grow new ones.
Paid social media shares some common foundations with paid search, but there are some significant differences both in how its auction-based model works and in the interaction users have with its advertisements.
As such, this maturing-but-still-nascent form of advertising provides a huge amount of room for testing and innovation. It comes with its pitfalls too, as we have seen with the measurement scandals.
A graphic depicting video thumbnails flying into the Facebook logo from the left, and money pouring out of it to the right.
However, there are other challenges that affect paid social practitioners daily and they require novel solutions. Below, we have looked at what makes a successful campaign – and what to do when everything best practice tells us just doesn’t work.

The ABC of Facebook Advertising

A successful campaign can be broken down into 3 areas:
Three graphics depicting the ABC of Facebook advertising: Audiences, Bidding and Creative.
  • Audiences: Deciding who you want to communicate with and how you want to utilize your audience lists and website visitor information.
  • Bidding: Selecting to use manual or automated bids and choosing the metrics for your campaign, which will affect your campaign cost.
  • Creative: Testing the right format for your audience since it will directly impact the effectiveness of the two categories above.
Let’s begin with the theory behind a best practice Facebook campaign across these three areas, with reference primarily to the auction that decides so much of how well our ads perform and how much they cost.
By way of contrast with AdWords, Facebook is inherently driven by images and their power to create an aspirational projection in the mind of the consumer.
AdWords works on a very pure, direct response model that is often based on text-based communication. A user types in a keyword and is met with a text-heavy response, although this stance is softening over time as results pages become more visually arresting.
Facebook allows a different approach to targeting, based on specific consumers rather than keywords. However, both are underpinned by an auction-based bidding system.
Facebook uses an auction for two main reasons: to create maximum value for advertisers and to improve the user’s experience. In an ideal world, this would see advertisers attract sales cost-effectively by providing a timely, relevant and enticing ad to the right consumers.
It is in understanding which aspects of this auction we can directly impact that we can start to affect our campaign costs.
Bids are defined by the bid value an advertiser sets (more on this below), multiplied by the percentage chance of their defined action being taken.
So if we want to bid on clicks as a metric, that value will be multiplied by Facebook’s estimated probability that that click will occur. Furthermore, bids in the same auction can target different outcomes: clicks vs. conversions, for example.
This is then combined with relevance and quality factors to come up with the final bid price and the auction winner. These last factors will be affected by things like image quality, negative comments on posts, and click-through rate. Note that this can change over time, affecting your costs on an ongoing basis as Facebook hoovers up more data on your performance.
An easier way to summarize and memorize this is B.E.A.R.:
Vector graphic of a stuffed bear above the words Bid = Bid value for the desired outcome x Estimated Action rate + Relevance & quality.
These bids can be set as either manual or automatic. You are told to use automatic if you don’t know what you are willing to pay for that action. If you do have a price in mind, you are told to bid your “true value” of what your action is worth to you. For instance, if you can only afford to pay $30 to acquire a new user, Facebook suggests you set your manual bid at $30.
If setting it on manual, you can choose a maximum or an average bid. For maximum, the algorithm is stricter on the threshold and is supposed to only find conversions below your bid, but unfortunately this is not always the case.
For average bidding, it will find conversions above and below your bid to find one that will eventually even it out. As seen below, if your bid is $10, Facebook will find bids anywhere between $2 to $12 as long as it averages out to $10.
Regardless if bidding on maximum or on average, you should ensure to always have a budget that is five times your bid to give Facebook enough breathing room to learn.
But this theory doesn’t always hold true and best practice sometimes lets us down.
So, what are some of the most commonly-faced challenges, and how can we deal with them most effectively?

Audiences

Hyper-targeting hampers performance

It seems counter-intuitive at first, but getting a bit too carried away with Facebook’s impressive targeting options can actually slow your progress.
It is best to provide Facebook with as much data as possible for each ad set so that its algorithms can find the optimal match between creative and audience. As such, it is advisable to avoid audience segmentation at this stage unless it has a clear and defined benefit to your campaign goals.

Ads stop showing

This issue can be caused by many different factors. The first port of call (after checking your bids and budget, of course), should be your website. Check the implementation of the pixel (the Chrome extension is very useful here) and look into Analytics to check conversion rates on key landing pages.
If all seems fine on the site, check the interactions your audience has been having with your ad sets in the past. If these are overwhelmingly negative, this could be enough to convince Facebook to stop showing the ads altogether.

Audiences stop performing 

If there is excessive overlap between the audiences you are trying to target across different ad sets, Facebook will prevent the list with the lowest performance history from entering the auction. Excessive overlap can cause significant issues, but it can be avoided by using Facebook’s audience overlap tool.
The audience overlap tool will show the percentage overlap between lists (we recommend keeping it below 30%) and provide insight into where you should consider consolidating your lists.

Bidding

Rising CPMs

The auction is increasingly competitive and CPMs are rising as a result. Therefore, sometimes that “true value” that Facebook advises us to bid just isn’t high enough to compete.
However, there are many ways to arrive at the same CPM, so it is worth trialling new campaign objectives such as link clicks. This can be risky, but if you are confident that your landing page will convert well, it can allow you to hit the same revenue targets (or better) for a lower campaign cost.

Pixel problems

Another challenging area we encounter is with Facebook’s pixel. This needs to be implemented correctly at each stage of the conversion journey to give you the accurate data you need. It is also crucial to ensure that each of these events is tagged accurately.
This is a fundamental foundation if you plan to use sequential messaging to target your audience at different stages of their purchase journey.

Analyzing GA data

A challenging question – and one with no obvious, catch-all solution – is the attribution of Facebook conversions within a wider marketing strategy, particularly at impression level. The main symptom of this issue is conversion data on two everyday platforms (Facebook and Google Analytics) that simply don’t match up.
Vector graphic featuring a rocket with two crossed flags behind it, one featuring the Google G and the other featuring the Facebook F.
We could say that Facebook is rather generous in how it weights its own importance, with a 1-day view through and 28-day click through window as standard. Multiple clicks from the same user within a short timeframe will also be classified as separate sessions, unlike in Google Analytics.
Analytics and DoubleClick do not have access to Facebook impression data, meaning that it is difficult to square this circle comprehensively. We also need to accept that we simply aren’t comparing apples with apples; these two platforms facilitate a different form of advertising and, as such, their metrics will inevitably differ.
That said, a great way to mitigate this issue is to trial conversion lift testing. This works best retrospectively and requires some initial investment, but it does provide insight into the ‘true’ impact of a Facebook campaign.
This same logic applies to Facebook ad sets, if you want to understand the impact each is having on your overall performance.
There is some optimism to be found in Google’s data-driven attribution too which, although missing impression-level Facebook data, will still provide a clear view on clicks. Here, we would typically expect Facebook to appear higher in the funnel, demonstrating its importance as an assist channel.

Creative

Ads are not relevant enough

If we return to our friendly B.E.A.R acronym, it is clear that even the best bidding strategy will flounder if the ads are irrelevant to the selected audience.
This falls down further if the ads contain too much text or if they have received negative feedback and a low CTR in the past. It is therefore worth investing time into the research stage of your strategy, to ensure adequate planning of creative assets and target audiences.

Creative fatigue

Creative fatigue is a widespread challenge and a natural by-product of the experience of using social media platforms. Users are bombarded by images, videos and sounds, and they have attention spans reflective of this. Even the best-performing creative has a shelf-life and can end up having a negative impact if left active for too long.
They key here is to read the signs carefully in your performance data and know when to switch up either the creative or the audience. As a rule of thumb, we would recommend updating creative every 2-4 weeks.

Lack of creative resources

The important thing to remember is that you don’t need an extensive creative team to produce great creative assets. There are some fantastic, free apps available to use that can help you turn simple images into the thumb-stopping media Facebook craves.
Below we have listed our three everyday favorites:
Boomerang: This is a free app that will allow you to loop actions back in forth within an image. It’s great if you want to emphasize a specific element like a product, for example.
Cinemagraph Pro: In a cinemagraph, one part of a photo moves while the rest remains still. This app is free to use and will allow you to create cinemagraphs from any image.
Layout: A simple one, but an effective app nonetheless. Layout lets you create collages and show multiple products within one image space.
There is still a learning curve when it comes to Facebook advertising for the majority of marketers, which makes it an exciting but challenging place to be.
The landscape is constantly shifting, but the above will hopefully provide some guidance on how to navigate the main obstacles we have encountered thus far in 2017.
‘Breaking down the Facebook auction: How to combat inflated CPMs and bigger budgets to deliver more sales in 2017’ was originally presented by Clark Boyd and Ximena Sanchez of Croud at Social Media Week NYC 2017. You can view the full presentation here.
This article was co-authored by Ximena Sanchez and Clark Boyd of Croud.