It's so easy to see which variable is
actually moving the needle for my clients and you don't end up competing
with yourself in the auction. Hello! I'm back. Made it through. Thanks guys. Really, really appreciate it.
So today I'm gonna be talking about machine learning. Is Facebook smarter than a media buyer?
Now before I get into things, I just want to know from a show of hands, how many of you guys think that hyper-segmentation and manual optimisations is the key to beating the algorithm? Any hands?
I see a couple shy hands okay. That's okay. I just want to get a vibe for how the audience is feeling. This will be a great speech then for all of you guys. My name is Savannah Sanchez. I work at an agency called Common Thread Collective and what I do is I manage the media buying teams. We do solely ecommerce and do Facebook, Instagram, Snapchat, Google Ads. How I got into this world of media buying is I actually have a marketing data science background. I got my MBA in marketing science
and I used to work for other agencies that focus solely on Instagram growth,
organic growth, but I really wanted to get into the paid social side about 2 years ago.
I took a job at this agency called Common Thread Collective where I
worked as a media buyer and I got to work on some really big agency accounts for
ecommerce working on Facebook ads and I absolutely love the data side of it.
I was always optimising data, pulling things through Excel, Supermetrics,
AR Studio trying to find the best way to optimise my campaigns and to such every bit of efficiency out. Just coming from a stats background, Facebook media buying was just a really great fit. After working at the agency as a media
buyer for a year, I ended up getting promoted after a very successful Black
Friday last year. I was working on the agency's largest accounts and now what
my current job is, I'm training media buyers, so the agency is growing over
three times this year, so I had to hire a bunch of media buyers and train them how
to do Facebook ads for ecommerce. In this journey of now
being a manager trying to train people on Facebook ads, I realised it was
actually really complex to try to teach what I was doing in accounts to brand
new buyers. First of all, there's just a very large experience gap in the market.
So trying to find media buyers who understand Facebook buying and can scale
accounts, there's just not that many of those people out there. First, it was
just very difficult to find people, and then once I did find the right talent at
the agency for media buyers, I found that it was very difficult to train them on
my way of optimisation. So what set me apart as a media buyer was that I was
always in accounts, I always was doing optimisations through Excel, tearing
through the data, but to teach that to a new buyer, I found to be extremely
difficult. And then another agency problem we were facing, so as of last
year we started to seeing some plateauing results for our clients. It
was getting harder and harder to scale. It wasn't as profitable as our campaigns
were in 2017. Getting in 2019, things were beginning very difficult for Facebook
advertising. So as is always happening, as I was becoming a new manager training
media buyers, Facebook released this framework called the Power5 and this
was released in February of 2019. And what the Power5 is, is that it's
Facebook's framework for saying that machine learning is going to help
advertisers optimise their campaigns and that there's much less need for very
heavy manual optimisation by a media buyer. And one of the developers that
released the Power5 actually said this quote. He said, the ways of manually hacking your way to add success are no longer, which is a pretty bold statement especially as a media buyer, I was thinking, am I out of a job? Is media buying dead? But really what Facebook is trying to say is they're trying to level
the playing field between small advertisers and large advertisers. And
they want to make it easy for people to enter the market and be able to
efficiently optimise their campaigns. The first two tactics of the Power5 are
auto advanced matching and account simplification. And these serve to help optimise your campaigns
while utilising Facebook's machine learning. The third and fourth and fifth tactic of the Power5 is a CBO, automatic placements, and dynamic ads. And these serve to help automate the process of
media buying and make things a lot easier by relying on the algorithm.
So the robots are coming. I was very excited about testing the Power5 because how
Facebook pitched it to us as an agency was that account management was gonna
get so much easier, so much less time intensive, plus they said that by
following these Power5 tactics, that my clients would actually receive
higher results, higher return on ad spend, and more scale. As I was trying to
train all these media buyers, I was thinking, okay if this actually works,
then this can save a lot of time in accounts, can make it so much easier to
train media buyers, plus if it's true that I can attain more scale and get
better results utilising the Power5, then it seems like a win-win. I was
skeptical but I tested it out on all of the agency's accounts, over 50 ad
accounts, spending anywhere from $0 to $1 million in ad spend, and it
became very clear that machine learning is much more efficient and clearly outbid
a media buyer from what we used to use in our previous tactics. Our
previous tactics of manually hacking our way to ad success, studying everyday
ads manager, breaking out different audience segments, all those tactics were
beaten by just simply utilising the Power5. Signal based machine learning is
simplifying account management and driving better performance.
Now signal based machine learning may seem like a buzzword but it's so
important to understand what a signal is because this is how Facebook is
optimising your campaigns. So a signal is all the information that Facebook is
receiving about their users. When someone adds something to the cart on the
website, when someone's visiting a product page, all of users past history,
past purchases, ads they've interacted with, Facebook has all this information
and they use that information in order to serve relevant ads to them. It's in
Facebook's best interests to provide the lowest cost advertising space for the advertiser, while at the same time making sure that
Facebook and Instagram users are seeing ads that are extremely relevant to them
so that they spend more time on the platform. It's a win-win.
How machine learning works in the Power5 is that it's taking these
signals about customer data, their purchase history, their websites, and
using those signals to inform decisions and optimising ad accounts. And just to
show you, these are a few of the brands that I have worked with that have
successfully used the Power5, and these are accounts that are spending hundreds
of thousands of dollars a month of ad spend, and I was able to achieve better
results and a much more simplified account structure just by utilising
the Power5 methodology. Here's the Power5 again. I know my speech today is specifically about campaign budget optimisation and my testing with that,
but what's so important to understand about the Power5 is that these tactics
don't work best when they're used in isolation. The whole point of the Power5
is that these tactics are used in tandem. So you can't use campaign optimisation
without properly leveraging automatic placement, a simplified account structure.
You could but you're not going to get the same results and fully take
advantage of the power of the platform by using these in isolation. That's
why I think it's so important to fully grasp all the concepts of the Power5
before specifically diving in to CBO. The first Power5 tactic is auto
advanced matching and this is by far the easiest thing you can implement in your
ad account. What auto advanced matching allows you to do is, so when people are
checking out on your website, entering in their phone number, their email, or
filling out their email to subscribe to a newsletter list, Facebook will take
that information and match it with users on Facebook. And what this will do will
help increase your remarketing audiences so when your remarketing website
visitors, you're gonna have more people in that pool because Facebook's able to
take all that information that people are entering on your website and match
it to Facebook users, and also it's an improved attribution, which is of course
as marketers one of the problems we face everyday, so this is improving attribution. And to show you guys, this is how auto
advanced matching is set up. You go into events manager and you just toggle on
this button and then you're set up. It's a set and forget method. It's all you have to
do and it will be able to improve your attribution and help increase your
remarketing audience sizes. The second tactic I want to talk about is dynamic ads.
I'm sure any of you who are running ecom have probably been
leveraging dynamic ads for a while. What this is, is you have a campaign that
you set up a catalogue sale objective and then what Facebook will do is when
someone's viewing content and or adding something to cart on your website,
it'll show people relevant products to stuff they have viewed. What's
really powerful about this is again, it's very much set and forget.
You upload your product catalogue list to Facebook, Facebook's taking the
pixel data of everyone viewing the products on your website, and then it's
showing people relevant products. And then another thing that Facebook has
talked a lot about this year is dynamic ads broad audiences. Not only can
you use these dynamic catalogue sales ads to retarget people up into your website,
but Facebook can take signals of people who have never even heard of your brand
but may be interested in similar products. I've interacted with ads that
are similar to yours and they'll show you the catalogue of your products to a
broad audience. Essentially, it's dynamic ads for prospecting. The
ad on the right, that's one of my clients' where I actually uploaded a custom
catalogue, so that it would show lifestyle images to a cold audience, since I
believe it looks a lot better than product on white. Now to go beyond dynamic ads,
because I'm sure many of you have been utilising dynamic ads
for a while now, but what I'm sure maybe not all of you are utilising is dynamic
creative testing. And dynamic creative testing, I'm convinced, is one of the best
machine learning features that has been released by Facebook today.
What dynamic creative testing allows you to do again, it's simple setup
on the ad set. You toggle on a switch that says dynamic creative on, and what
you can do is on the ad you can upload different videos, different images,
tops, texts, headlines, and Facebook will dynamically test each of
those variations against each other, and tell you which is the best combination
for your target audience. It's very cool and it's completely revolutionised how
we create tests at our agency. To show you the old way we used to media buy, which I'm sure a lot of you can relate to, if we wanted to do creative testing,
testing out different thumbnails, headline, top text videos, we would have
to build out each variation as their own ad. Not only was this extremely time
intensive, but also it was very difficult to get statistically significant
learnings on each variable. Took a lot of budget in order to spend enough on each
variable to know which was actually driving performance for our audience.
The new way of creative testing and what I've been leveraging in all my ad
accounts was dynamic creative. So uploading all the different variations
but it's all within one ad and then Facebook is dynamically testing each
variation against each other to find which combination is going to
perform best for your audience. And to show you a bit about the setup, so when
you toggle on dynamic creative on the ad set level when you build the ad, you'll
be able to upload different photos, different videos, and then if you want to
see the results of your dynamic creative, all you simply have to do is go into the
breakdowns and ads manager, and you could break down by video, headline, top text, so
you can see the best performing variable for all of your creative testing. This
has allowed me to be able to build creative testing much faster and also
get creative results much faster. So I can use less budget and be able to get
statistically significant results to tell my client within even a week.
This has completely revolutionised the way we are creative testing. Plus, the way
the algorithm works is that the more creative variables that you give the
algorithm, the better it can optimise. It optimises on the user level, so if they
know one person is more likely to convert from a short-form video versus
long-form versus image, it's going to serve the variation that that person is most likely to convert on. It's going to be in your best interest from a performance perspective as well
to be utilising a dynamic creative testing. The third tactic of the Power5 is automatic
placements and this is something that's also been available for years now on
Facebook, but it's gotten more and more powerful in recent time. And what
automatic placements allows you to do is you can upload your square video, or
photo, whatever your creative asset is, and Facebook will test it on Facebook
news feeds, Instagram news feeds, stories, audience network, all the available
placements. There's 17 available placements now that Facebook will test
your creative on. Now the old way of media buying would be that maybe you
select feeds only or you select whichever placement you wanted to test,
but where automatic placements is very powerful is, that it knows which users
are most likely to convert on which platform. And also it will go after where
is the lowest cost conversion available. News feeds might be very expensive so
they know that someone may purchase on Instagram stories instead. So they're
gonna be able to drive lower costs over time utilising automatic placements. I
wanted to show you an example of one from my client. This is a client called PupSocks.
And what I did is I uploaded my square image, and these are the results
from the different placements off of $35,000 of ad spend on this one creative.
Now what you guys might see right away is, oh right hand column was at an 8x,
I should scale right hand column and feeds was the lowest ROAS. But the
question you should be asking is how much scale was I able to get out of each
of these placements and how much spend? I will show you guys the breakdown
and as you can see, although right hand column audience network had the highest
ROAS, it also doesn't have that much inventory, didn't have much scale. Where
feeds, even though it had lower ROAS was able to get a lot more scale, while
maintaining ROAS at my goal. I talked to a lot of media buyers and a
lot of media buyers will push back on automatic placements and they'll say, I
only run on feeds, feeds convert the highest, they have the most scale. But
what you're missing out on if you're just running on feeds only is
opportunity to show your ads at a lower cost conversion. I showed the
breakdown on the screen. If I was just to run on feeds only, I would only be able
to achieve a 2.5x ROAS and $31k of spend. Whereas because I opened up to
automatic placements and letting Facebook serve my ads on cheaper
opportunities for inventory like audience network, stories, and I'm able to
achieve more scale - $37k and higher ROAS by taking advantage of these lowest cost placements. And we tested automatic placements across the entire
agency and what we saw is a 23% lift just by toggling on automatic placements. Again,
this is so simple and extremely revolutionising the way that we're testing.
It's eliminating the need to test out every individual placement, and we're
allowing Facebook to use the algorithm to serve our ad where the individual user is most likely
to convert at the lowest cost to the advertiser. The fourth tactic, and I must say is probably my favourite tactic of the Power5 is account simplification. This year we've revolutionised the way
that we buy on our agency accounts, we're minimising the number of campaigns, and
we're consolidating the account structure to follow Facebook's best practices. And to show you, this is how we used to buy and maybe
how some of your ad accounts currently look. With hundreds of different ad sets hyper-segmenting
out different audiences, different placements, even by creative type and you
end up with an ad account that is an explosion of data. And it's so hard to
spend on each individual adset and get statistical significant learnings, as
well as those often so much overlap between audiences, so between all of your
lookalikes, all of your interests, you end up competing with yourself in the auction
and bringing up your costs, while also it's just extremely time-intensive to manage
an ad account that is this built out. Whereas the new way with account simplification is having a minimum amount of ad sets. So consolidating audiences together, similar audiences, broad audiences, putting them in your funnel and then
utilising minimum campaigns. And what this will allow you to do is, because
you're utilising broader audiences, you're gonna be able to get through your
budget more efficiently because the algorithm has the signals in place of
who is your ideal person to purchase, so if you give Facebook a broader audience
of who to serve to, that it's not putting any limitations on the algorithm. And
also because it eliminates a lot of audience overlap by having hundreds of
different ad sets all competing with each other, you're gonna be able to get
better delivery and also get out of the little learning phase faster. The
learning phase is so crucial and important. It's so important that you're
out of learning phase so that you're can be at the full optimisation efforts that
Facebook has available to you. By having hundreds of different ad sets, you're not
going to be able to get statistically significant learnings on each individual
audience, and it just gets very hard to see what variable is actually working. I wanted to show you guys how
I default account structure my campaigns. I have the classic prospecting top of the funnel, where I'm showing a cold audience and I'm utilising broad interests, lookalikes, and then utilising
dynamic creative testing and campaign budget optimisation. Then I have my
middle of the funnel, which is my re-engagement audiences. I'm targeting
people who have interacted with the Facebook page, Instagram, video views.
And then finally bottom of the funnel is remarketing, so people who have
been to the website, past purchasers, all across the board, utilising CBO, account
simplification, automatic placements, all the Power5 tactics. And then lastly the
DPA, dynamic product ads for remarketing. And if any of you guys want a copy of
this, I'm gonna be at the Facebook Ad Mixer later today after the speeches. I'm
happy to email you a copy if you didn't snap a photo. And to show you visually
what this looks like for my campaign. So top of the funnel, I have UGC, people
talking about the product, keeping it very authentic. Middle of the funnel, I
start to get more product-focused, so I move away from the UGC, PR, and now I'm
talking very much about product benefits. And then bottom of the funnel, those
dynamic product ads. I'm showing people the products that they viewed on the
website. And this is what it ends up looking like in my ad account. It's a
very simplified structure. It's so easy to see which variable is actually
moving the needle for my clients and you don't end up competing with yourself in
the auction, so it's truly a win-win having a simplified account structure.
Finally, we are on to campaign budget optimisation, which is my favourite topic
to talk about because it's a bit controversial. For those of you who
don't know, campaign budget optimisation it's actually becoming mandatory in 2020
for all advertisers. I'm sure a lot of you have been testing and trying to see
how to buy on campaign budget optimisation, and if they can achieve the
same results, as manually choosing the budgets for your adsets. I've been
drinking the CBO sauce, also a sauce available at McDonald's, because I am in
love with the abilities that campaign budget optimisation gives you. Because
you're allowing Facebook to decide which audience to put your budget on, it
eliminates the need to be constantly managing budgets and to be doing those
very intensive hourly optimisations, but instead leveraging Facebook's platform and they choose which audience to serve your creatives to. This is how you toggle on
campaign budget optimisation on the campaign level. Toggling it on
allows you to do is you're gonna choose your daily budget for the campaign. If
you want to spend $500 on your campaign per day, you choose three to five
different broad audiences, and then Facebook will decide with that $500
where the best opportunity lies within each of those audiences that you select
for your ad sets, and will dynamically put budget towards the ones where they
think you're gonna get the lowest cost results in real time. The old way in
media buying, which may be how your accounts are currently. Without campaign
budget optimisation on each ad set, you would have to choose your budget
manually, and then by the day or even hour, maybe you have automated rule set
up, you're choosing different audiences to put your budget to. But now with campaign budget optimisation, you are setting your campaign daily spend on the campaign level and then Facebook is choosing in a
real time which audience to put the budget towards. Some benefits of campaign budget optimisation. Most obviously, zero time managing budgets while lowering your CPA. Not only does campaign budget optimisation drive better results but you're gaining that precious time back so that you can focus on the things that actually move the needle like creative, strategy, your offer,
your website. Those are the things that make the biggest difference in your ad
account and so now that you can spend less time managing budgets, you're gonna
have that precious time back to work on the levers that truly make a difference. The second thing is that decisions are made off future outcomes. What this means is that Facebook understands in each ad set where the lowest cost
conversion lies and it will automatically distribute that budget
towards where it thinks the lowest cost conversion is. Whereas a media buyer, so
if I was an ad account, I don't have access to any of that data
that Facebook has about its users and when and where they're most likely to
convert. So all I can do as a media buyer is make decisions off of past outcomes,
which is often very inefficient. I have an extremely limited data set because I
only know past information about my campaign and even if I'm using automated
rules, it's always based off past data, past CPA. Facebook has a data set of how
they think that their users are going to interact in the future, so their data
set is just much more powerful, it could make better decisions than an individual
media buyer has access to in terms of data. The third major benefit is that a
de-duplicates audiences. If you have five different audiences within a CBO,
you don't have to worry about audience overlap. If there's the same person that
lies within multiple audiences, Facebook will just move the budget towards one,
but it won't put the other one in the auction. What this allows you to do is
you won't end up competing yourself and driving up your costs. See, Facebook's
algorithm is smart enough so that it'll just put the budget towards one of the
ad sets while ignoring the other, so that you're not competing with
yourself in the auction and running into delivery issues. And lastly but probably
most importantly, it's that CBO avoids restarting the learning phase, and as I
mentioned, the learning phase is so important to get out of in order to
fully optimise your campaigns, and what Facebook recommends is getting at least
50 conversions events a week in order to fully optimise and to be out of the
learning phase. But if you're not using campaign budget optimisation and you're
changing your budget daily, hourly, using your rules, what it does is that every
time you make a major change to the ad set, it's resetting the learning phase, it has
to relearn who is most likely to purchase on this ad set with your
creative, and you're never going to get out of that learning phase and into the
true mode of optimisation. That's why CBO is so important because you set your
budget on the campaign level, you let it go, and you let Facebook fully optimise
and learn on your audiences and your ads. Now there are some downfalls of CBO and
it requires a volume of pixel data. Like I mentioned before, the signals.
Facebook has to know who your ideal customer is, so if you don't get a lot of
website traffic, not a lot of purchases on your website, it's going to become
very difficult for Facebook to figure out who inside these broad audiences are
most likely to convert. It is important to have a seasoned pixel and
to have enough pixel data coming through on your website so that Facebook can
fully optimise. The second is of course less control with campaign budget
optimisation. And as a media buyer who used to pride myself of manual
optimisations and diving for the data that's just no longer needed and I'm
giving up control to Facebook to best optimise my budget, which is a bit scary.
I completely understand that, especially if you're seeing right away that maybe
results aren't as good with CBO as your prior manually optimised campaigns.
But I think this is probably the biggest concern with CBO and something that I
have to talk to a lot of entrepreneurs and media buyers to get over is almost is
fear that Facebook is not working in your best interests. So I hear all the
time from media buyers they say, if I let Facebook choose what to do with
my budget, they're just gonna blow my money, they just want to take my money.
But that's a very short-sighted way of thinking about Facebook and its true
value. What Facebook wants to do is it wants to provide the lowest cost
advertising for its advertisers while again showing relevant content to its
consumers, people using Facebook. Because Facebook wants advertisers to
put more budget towards Facebook, to scale on Facebook, and not move to other
platforms like Snapchat or TikTok, it's really in Facebook's best
interest to provide the best results to advertisers. Although Facebook is
making CBO mandatory and it seems like it's giving up a lot of control, just
know that Facebook does have the advertisers best interest in mind and
it's going to use these algorithms in order to get you lower cost results. Now
the CBO challenge. I posted a few months ago in an ad buyers group. I said, is
anyone up for the challenge to manually buy on a campaign versus a CBO campaign
worth set and forget? And I'll show you guys the post. I was in
Tim Byrd's Facebook ad buyers group and I said, will the robot take your job? Do
you take on the challenge if you were to buy against a CBO campaign and see
which one would win? I was genuinely curious. Maybe there are media buyers outside
of my circle and my agency that can beat the algorithm. I was completely open
to that. I got 95 comments on here, a lot of people wanted to prove me wrong
that a robot will not take their job and that CBO is not as smart as their media
buying tactics. I was able to enlist a media buyer friend of mine Eric, an
acquaintance. He was up for the challenge of trying to beat the
algorithm. Then I told Facebook about this experiment and they were so excited
to prove the value of CBO, they gave me a $10,000 coupon to test this out.
So how it was set up, is that Eric will get $5,000 from my client and then Facebook
CBO would run another campaign, and we would run it for two weeks, and see which
one would get the better results. And I talked to Eric. I'm like, so how do you
media buy? Tell me about your process. Why do you think you can beat the algorithm? His tactics were very similar to
what I hear other media buyers say. They say to look at ROAS, CPA for a day, what's
the AOV for that ad set, what's the cost per ad cart, unique checkout, looking at
all these metrics, and then making rules about when they would lower budgets and
raise budgets based off of these if/then statements. So if CPA is lower than this,
then I'll put budget here, etc. That was his media buying process. Then I
found a client, one of my agency clients, Road ID, was of course very keen to take
on a $10k free ad credit from Facebook for a test. And so they sell identification wristbands. People who go on runs, live an active
lifestyle, maybe they're out biking, they want to be able to have a form of an
identification in case something bad happens, and emergency services need to
be able to identify them. Average order value is around $30 and they spend
around $80k a month on Facebook ads. This is important because this account
has a lot of pixel data. They were running this account for a couple years
so it's fully ready to take advantage of Facebook's optimisation techniques. So here's the set up, $5,000 to the CBO campaign and then $5,000 to the non-CBO campaign.So that came up to $357 of budget each day per campaign. For the CBO ad set or the CBO campaign, it was completely set and forget. You could not adjust budgets. You could not turn off ad sets. You just had
to let it run for the full two weeks. Whereas for Eric's campaign the non-CBO,
he was able to make daily adjustments on ad sets based off of where which one
was performing best, based off his own decision making. He could also turn off
ad set, so if he found that one wasn't performing as well. But in both campaigns
you could not turn off ads. So we left the creative as a completely constant
variable. I did not want that to affect the campaign results. So creative was
constant. It was always turned on, nobody was adding in new creatives throughout
the two weeks. Again utilising the full spectrum of the Power5, automatic
placements, optimised for purchase, I utilised lowest cost bidding, as well as
doing worldwide targeting. So I'm essentially allowing Facebook to find
the lowest cost conversion within any country. They were able to
ship to any country so it made sense it set it up as worldwide. And then here's
the interest I chose. It's like I said utilising, a consolidated campaign
structure. I'm utilising very broad interests, so I know that this product is
very big with outdoor enthusiasts, so biking interests, running accessories
interests, and the interests are the same for both campaigns. And then something
I've also been utilising is detailed targeting expansion. So when you're
entering your interests on Facebook there's a little check mark under your
interests where you can choose to expand detail targeting if Facebook thinks it
can find a lower cost conversion outside of your details targeting. And what that
essentially allows you to do is give the reigns to Facebook say, hey I think my
people are within this running interest, however if you think someone's gonna
convert based off the signals that you're getting from them and they're not in this interest,
you can go ahead and serve it to them. Here's some of the creatives that I utilised in this campaign. I want to use a wide variety of assets. Again utilising dynamic creative testing, so I uploaded a few different photo variations, different videos, different headlines, different
top text and I'm letting Facebook dynamically test each of these
variations against each other within each ad set. And just anecdotally out of
these four creatives, the unboxing video on the far right was by far the best
performer. I think it's just very authentic, it shows the product, and I
thought it was very interesting that that was the one that Facebook
optimised to very heavily. And so here is the results. This is the CBO test. As you can see in the beginning, it chose to put a lot of budget towards audience 4, but
was still testing each of the ad sets pretty evenly. It gave all the ads
a fair chance as it optimising and then over time you can see that it
really liked audience 1. And so I started putting more budget towards
audience 1, it decided that audience 2, the yellow audience, to decide that one
was complete trash, it just completely stopped putting budget there, and that
way the algorithm is just optimising, and learning, and choosing which audience to
put the most budget to. And so here are the results of CBO versus non-CBO.
Now in the first week, Eric was actually beating CBO.
I was getting a bit nervous. I'm like, maybe this CBO isn't all it's cracked up to be. As you can see the orange lines, non-CBO is actually winning, but this is the thing with campaign budget optimisation, you have to give it time to optimise. You have to get in the learning phase, have to get enough pixel data on each of your ad sets. And once the campaign started optimising, then CBO just shot up and completely was beating Eric's efforts. Like I said, with CBO, it does
take time, so you have to be able to let it set and learn for a bit, and don't get
too jumpy about turning it off if it's not getting results right away. Here's
a question I also get from people who are concerned about CBO. They said,
shouldn't CBO just put the budget towards the one with the highest ROAS?
I mean if that's the whole thing of it's going drive lower CPA then that's how it should work right? But that's not always how it works, because I said that
Facebook optimises campaigns off of future data, out of data they know about the consumers that the ad buyer doesn't. Although one ad set may have a higher
return on ad spend at the time, they know that the audience is exhausted. There's
no more lowest cost conversions within that audience. So they'll move the budget
towards another audience where there's other lower cost conversions. But a media
buyer wouldn't see that. They would just put the budget maybe towards the one
with the highest ROAS and the best leading metrics, when in reality that
audience could be completely exhausted and there's no more scale out of it, so
has to move the budget towards another audience which may have a lower return.
And so here's a breakdown of how the spend was between the CBO versus non- CBO. As you can see for the audience number 4, Facebook put a significant
amount of budget towards that, that would have the highest ROAS. Same for audience 1.
But as you can see, as you keep spending on each audience, like what Eric did,
so he probably saw that audience 2 was doing really well, but he put more budget
towards it, but it ended up diminishing the returns. That's the problem with
having a media buyer optimise your campaigns versus an algorithm is that
they can't see that there's no more lowest cost conversions living inside
that audience, so that it's no longer profitable to keep putting budget there.
And the final results CBO versus non-CBO. CBO had the higher ROAS of 1.69x versus a 1.62x. Now although that's not by a very large
margin, because if it was a magical thing CBO, if I could just switch on CBO and
start printing money, I would be a billionaire, but I'm not. CBO provides
marginally better results but the real savings is the time savings.
Eric logged 10 hours of managing that one campaign over two weeks, where the
CBO there was no budget modifications, it was just set and forget for CBO. So you
get that precious time back and with that time, that way you can put that
towards more higher value activities, like working on your creative strategy,
your offer, your product. Getting that time back out of ad accounts and
utilising that to actually grow the business, that's where the true value of
Facebook's algorithm is and campaign budget optimisation, saving you time so
you can focus on those bigger levers. How to set up your CBO's?
Lookalikes, broad interest audiences, I always turn on interest expansion in case Facebook can find a lower cost conversion. Three to five ad sets per campaign. Dynamic creative if not, I'll allow
three to five ads within an ad set. Utilising automatic placements, so again utilising the full spectrum of the Power5.
Conversion purchase optimised, so if you're running ecommerce, I always
optimise for the event that I ultimately want. So ultimately I want people to
purchase on the website. So I'm going to optimise for purchase conversions. And
then for bidding strategies, if you need scale, if you need to be spending a
consistent amount on each ad set, then or campaign, then I would recommend lowest
cost bidding. Whereas if your goal is to be the most efficient with your ad
dollars and you don't necessarily care about how much you're spending per day, then I would recommend a bid cap or a cost cap strategy, which is gonna give you the most efficient
ROAS possible but just not as much scale. And then lastly my tip with CBO is, give it
time to get out of the learning phase and have enough pixel data in order to fully
optimise. If you're gonna run this test on your own account, which I hope you do, just give it at least two weeks and I
had $5,000 for that campaign, so that I could have statistically significant results. Now what to do with my time? Like I said, I'm not in ad accounts as much anymore, doing the day-to-day optimisations all night, all weekend,
managing budgets. Now that I have this simplified account structure utilising the Power5. I'm spending pretty much all of my time focused on the creative testing aspect.
So utilising dynamic creative testing, testing different angles to unlock new audiences. So messaging angles, long-form, short-form,
UGC. My entire focus now in terms of Facebook is doing creative testing
instead of audience segmentation. The fastest growing advertisers create 11x
more creative, which resulted in 3x more revenue. I know some of you guys might
be thinking, how do I get 11x more creative? What I've been doing for my
clients is these UGC type videos making it very authentic, filming with my iPhone,
boomerang style, unboxing, application videos, testimonials, and these are all
filmed on my iPhone and just to get the volume of creatives up. But what I found
this year is that this style of advertising was actually working better
than the very produced creatives that we would do in the studio, because they're
authentic, and because they're designed to convert on Instagram stories,
they're native to the platform. It looks like your friend or influencer could
be posting it. I would highly recommend if you're trying to get a
volume of creatives, get a bit hacky with it, utilise your iPhone, get your friends,
the products, that's usually where the best creative comes out of in my experience. And so lastly, if you're spending less time in your ad accounts because of utilising CBO and the Power5, this is where media buyers can really
shine in 2020, focusing on the offer. Testing different offers, different
pricing, what deals are on the website that's made a huge different in ad accounts
in my experience. The top advertisers are also focusing a lot on story.
What story your ads are seeing? What story are they seeing at
top of the funnel versus bottom of the funnel? Truly a creative testing focus,
and then experience, so customer experience on the website, post-purchase
experience, these are the biggest levers that are gonna make a very positive
impact on your ad account. It's not as much about manual optimisation anymore
These are levers that'll make the biggest difference. Lastly, I'll leave you with this thought. It's not about effort, it's about outcomes. So what I mean by that is, that it does take a lot of effort, the old way of managing Facebook ads, having hundreds of ad sets, all this
testing, but if you can achieve better results and get your time back to focus
on the more important variables such as creative, then you're gonna end up
getting better results in the long run, while also getting that precious time back. Thank you guys, this was awesome. Thank you for listening and if any of
you would like to debate me on CBO, which I'm sure if you and you in the audience will,
I'll be at the Facebook Ad Buyers Mixer after a couple of these speeches.
I hope to see you there. And also my email is up there, if you want to just
personally reach out to me and ask me any questions about this or I can send
you the deck, that's totally great as well. Have a great day guys.
