" Worried about cannibalization
You've got a sneaking suspicion that one Facebook ad campaign is cannibalizing the other, and need to do some analysis to see if your hunch is right.
GoolyBib Office – Thursday
You're unusually ahead of your work and still have time left in the week. Now could be the perfect time to look into something that has been worrying you: are your ads cannibalizing each other?
With so many targeting options available, it's common that some of your audiences will overlap. Some dog owners also love to travel, so stay at home moms play video games, and some finance professionals play golf (ok, all of them do). If the amount of overlap is small, and your spend is relatively low, it's usually not a big deal: there are plenty of fish in the sea. However higher rates of overlap can cause strange behavior: campaign performance unexpectedly dropping off a cliff, ad fatigue setting in early and inflated costs, as you're effectively bidding against yourself.\nMost modern ad platforms give you the ability to add targeting exclusions as well as inclusions, but first you need to know if you have a problem. The best thing to look out for is a divergent return on ad spend between two campaigns. If performance is changing in one campaign in an unexpected way when you make a change to another, that's a sign that their performance is inversely correlated, because they're overlapping.
Do I contradict myself?\nVery well then I contradict myself,\n(I am large, I contain multitudes.)
Most advertising platforms operate on an 'auction' model, where advertisers set their bids and a winner is determined. Bidding on overlapping audiences can inflate your costs because you could be bidding against yourself in the auction.
Is one campaign hurting the other?
Let's run the numbers: how does one campaign perform when the ROAS of the other increases? Is there a negative correlation?
GoolyBib Office – Later that day
You settle down to do the analysis, but you're unsure of what to look for. What evidence would show you that one campaign is affecting the other? You message a friend.
Hey, yes we had something similar\nThe best thing to do is make a scatterplot of Return on Ad Spend\nIf they have a strong negative correlation it's good evidence that one is affecting the other
ROAS (Return on Ad Spend)
Take the revenue or estimated value attributed by your ad campaign and divide it by your ad spend. This is a useful metric for when you can't easily track profitability and can be a better metric to use than cost per acquisition when conversion value has a wide range of potential values.
If you want to know if one variable affects another, it's useful to start with correlation. This means that one variable moves up or down at the same time as the other variable. A positive correlation means that both variables increase at the same time. A negative correlation means that one variable decreases when the other increases. You can find the correlation between two variables simply by creating a scatterplot in Excel or GSheets. You can also add on the R2 to determine the fit of the data: how much one variable explains the other.
Hey, I'm going to teach you how to make a scatter plot\nI'm just going select this data, this is a pivot table, I'm pulling it from another sheet\nI'm just pulling in the campaign name, and here is the return on adspend value, so this is purchase value divided by Spend\nI'm just going to select this data, and to do that I'm going to select some data and then press control, shift, and then down and then right again\nOn the mac I think it's command, shift and then down\nSo I have all the data selected, I'm going to click insert and then chart\nCool so we have the data in here\nI'm just going to see, over on this side, if we can pull up a scatter plot\nand you can see that it doesn't look right, we don't have everything we need\nso we're just going to get rid of the grand total\nand then with a scatter plot you want to choose one of the variables against the other, and right now we have the date on the x axis, so I'm just going to remove that, and then remove this one here, and then I'm going to add it back in here\nNow we have one campaign on the x axis and another campaign on the y axis\nYou can see straight away that they do look correlated\nThe way you would interpret this, is when the return on adspend for one campaign is ten, its below zero for the other one\nand when the return on ad spend drops to four, it goes higher for this one\nso what this is saying is that these campaigns look linked, in that when the return on adspend is higher for one campaign it tends to be lower for the other one\nthe way we can see how linked they are, is by adding the equation to the chart\nclick in customize, series, and then we're going to click on trendline\nand that just draws a line in between here\nthen we just want to click on here, show R2\nso we can see that 50% of the parents mothers return on adspend can be predicted by the other campaign
Were you able to create a scatter plot as above?
How can the 1% Lookalike (US) campaign have a negative return on ad spend?
Deciding on a campaign hierarchy
Not every campaign can be king: which priority order will you assign when making audience exclusions?
GoolyBib Office – Later that night
Now that you've figured out that your campaigns are overlapping, you have a choice to make. Which campaign should you favour over another?
It's natural that campaigns will overlap each other, and we can avoid any potential cannibalization on most platforms by using exclusions. However how to do we decide a relative priority? The first and most obvious move is to exclude retargeting audiences from prospecting audiences. If someone has been to your website recently they'll behave very differently to a cold prospecting audience, will likely respond to different creative, and will be less incremental (they might buy anyway, but your ads will claim credit).\nThen to decide relative priority between other audiences, ask yourself this question: \"if I only knew one thing about a person, and I had to decide if they were likely to buy, what would I want to know\". You typically want to prioritize the smallest, most niche audiences first, as they're most likely to buy from you, but are small in nature. Exclude them from the broader audiences that you target, which can afford to lose some volume and don't add much in terms of specifying who is likely to buy.
Would you exclude the lookalike audience from the mothers audience, or vice versa? Why would your chosen campaign get priority?
When you have multiple audiences you want to prioritize and it's not clear which to favour, it can help to look at it from a creative perspective. Which of these audiences would respond best to unique creative tailored to them? That usually serves as a good tie breaker and reason to prioritize that audience over the rest.
There can only be one king"