Professional Documents
Culture Documents
1) Do NOT use Lookalikes for targeting
People in Facebook learned interesting thing based on a lot of data they have. We should
use lookalike targeting when using conversion objective but definitely not for DABA
campaign. So when you are using product catalogues for acquisition you should go as
broad as possible instead of using Lookalikes. Again it is mostly because of the algorithm
which could have more data and pick the more suitable audience from your target group,
maybe even better than targeting lookalikes.
2) Exclude only purchasers from targeting
Another part of the DABA targeting is to exclude only people who purchased on your
website/in your app in previous X days. That’s all. We should not exclude our visitors from
targeting.
When excluding the purchasers we should think about our business. Usually people just
exclude for example people who purchased in the last 7 days and they don’t think about it
at all. But if for example people start buying on your website again after 3 days, then
exclude only people in the last 3 days and not 7. If the purchase process won’t happen
earlier than before 14 days, exclude people who purchased in the last 14 days. It is
important to know your business and exclude the people based on your real data.
CATALOG SETUP
With the algorithm process being so advanced at this moment for optimisation
Facebookers say that the most important thing about all your campaigns is proper and
100% correct set up of your product catalogues and your Facebook pixels on your
website. The better you have your catalogues and pixels, the better results you could
achieve.
That is why this section will be all about the product catalogues and how to make the audit
of your client’s catalogues. There are some very important things which we need to be
considered.
The more info we can send to Facebook to the catalog about product the better, every
new info will help FB algorithm to find the best customers for us.
Amount of products
It is important to have as many products in the product catalogue as possible because
Facebook is gathering all the signals about your products and is analysing how users of
Facebook interact with those products and then it helps you in prospecting campaigns.
This way it is prefered to have one big product catalogue with all the products from your
website.
If for some reason the product on your website is no longer available it is better to set its
availability as “out of stock” instead of deleting it completely from the source feed. It is
because of the amount of data that FB gathered for this product. Once you delete it and
then for example introduce it in future again then FB will have to get all the signals for it
again from zero.
For example on Allegro we had product catalogue for each sub-category such as fashion
for men, fashion for women, fashion for kids… Once we actually made one big feed and
catalogue for all fashion products together we saw increase in results.
Product categories
It is recommended to have correct product categories for all your products. If you have in
your feed correct Google categories then it is totally fine and enough for Facebook and it
is actually the best practice to use them for your feeds and catalogues.
Catalogue diagnostics
This is very useful and important section for you in the product catalogue section. You
could find here all the problems that your product catalogue has and the reasons why
Facebook is not able to learn the most about it and provide you the best optimisation.
First go to https://www.facebook.com/products/?business_id=<Your business ID> and
select the catalogue which you would like to check. Then you will see the Diagnostics
section. Here you will see the most important errors and mistakes with your product
catalogues.
The RED warning messages are the critical errors and they need to be fixed.
The ORANGE warning messages are just warnings and it is not that important to get them
fixed. But of course for perfect condition of your products you should fix them as well.
The best thing here is that the Diagnostics section will write you also how to fix those
problems. So start working on the red errors you have there and everything will be better.
Another important section which you would need to check is called Event Data Sources.
You will find it also in the product catalogue in the top menu. Here you would see the data
source you are using for this catalogue, so for example all the pixels connected. Once you
click on “See all conversions” at the data source, you will be able to see detailed
information about the events and their match rate:
Once you are at this step, you can finally analyse how “healthy” is your product catalogue.
Facebook is recommending your match rate of your FB pixel with the product IDs in your
catalogue to be at least 90% and more!
Critical events to check are:
Search/ViewCategory -> ViewContent -> AddToCart -> Purchase.
As you can see in this product catalogue we have only 47.6% match rate for ViewContent
event. What does it mean? It means that we are only able to retarget people who viewed
47,6% of all the products. That is super small number!
This way you are basically not reaching the full potential of your ads. You might reach a
point when you can’t spend more or you can’t deliver your ads to more people and you
are not able to scale. This could be the main reason why you are not scaling your
campaigns up to better results.
What you can do is to check the reasons why this is happening. One of the most common
reasons is that the products on the website are not in the source feed or in the product
catalogue. There is a product with correct ID on your website but there is no product with
the same ID in the feed or catalogue.
Luckily for you there is an option here to download the list of unmatched products and
send it to the IT guy responsible for feeds. Before you should check some random
products from this list if really there is correct ViewContent triggering with correct ID and if
the product truly is not in the feed. You can download the list of IDs and also the list of
URLs and FB will also give you the amount of hits of those events for those products.
We worked a lot on this with ZOOT in previous months because they had really low match
rate and we felt like the DPAs are quite limited. Now we are on all events above 90% and
everything is working well.
Product sets
Because Facebook is recommending you to use the biggest catalogues possible and
target broadly for acquisition of new customers, they also recommend to you to build
effective product sets that are broad enough but also logical.
You should always think about your business and create product sets based on what is
important to it. It is recommended to use custom labels for your products to give it extra
information for making those product sets.
Do you want to optimise for the best profit margin? Pass it in the custom label and select
product set only with the best products.
Do you want to promote some special sales? Give your products info in custom label that
they are part of that sales and create product set filtered by this label.
Example of how we mark our products with one client:
Custom_label_0: type of promotion (none, discount…).
Custom_label_1: profitability of products (A,B,C products).
Custom_label_2: amount of discount (1%, 2%, 3%...).
Custom_label_3: reviews of users (This product has 5 reviews).
Think about the logic of what you are selling to whom and then try to be as broad as
possible. For example on Zoot we just diversify products for female and for male and then
run DABA with them.
Image transformations
This is a feature of Business Manager’s catalogue manager which is not so well knows yet it
can be very handy for your products. Here you can change how your products look like in
your ads. For example you have a product feed only for carousel images so you could
easily transform how it should look like when the product is used for single image.
You would find this feature when you enter the product catalogue and then go to Settings.
Here you would go little bit down and you would find “Image transformations” section.
Advanced pixel matching
Advanced matching can increase match rates even for people who are not tracked via
pixel. Send your customer data through the Facebook pixel base code to match more
website actions with FB users. An increased match rate improves your targeting and
optimization of your campaigns.
It is possible to do it automatically now by turning the Advanced pixel matching option
ON and by letting Facebook to gather your data from forms and user behaviour on your
website.
Facebook sees in average +8% increase in results thanks to switching to Advanced
matching!
About GDPR: There is nothing to be worried about because they transfer everything into
hashed data and it is used only for better matching and nothing else. Facebook has
external legal statements which confirm that it is ok with GDPR.
SETTING UP THE CAMPAIGN
When setting up the DPA campaign there are few very important things which we will
cover in this section. When you choose them wrong it could lead to under-delivery, low
delivery or worse results.
Budget
The budget of your advert set is important for the good delivery. When running DPA
campaign you should always think about the budget before you start your campaign.
When you set budget too low it could result in two things:
1) Good delivery in the first day or two and then very limited delivery for the next days.
2) Small delivery or no delivery at all.
Why is that? You should always think about how does the FB algorithm work and what is
your optimisation goal.
When you are optimising for Purchase event 1 days after click and your CPA is 15 EUR.
You should set up budget at least 4x the current CPA. If you would start your advert set
optimisation with budget 10 EUR but your CPA is 15 EUR, the algorithm would never be
able to deliver your desired results.
That is why it is actually better to optimise for events higher in your funnel, for example
AddToCart. If your budget is 10 EUR, your CPA for Purchase is 15 EUR but your CPA for
AddToCart is 0,5 EUR, then you are good to go. Your budget for this advert set is 20x
higher than the CPA and the algorithm would be able to run it.
So when setting the budget always think about the optimisation you chose and what is the
CPA for it. If your budget is smaller or the same as the CPA, you will have big delivery
issues. If the budget is larger, at least 4x, you are good to go.
That is why Facebook also recommends to optimise for events higher in the funnel, with
bigger attribution windows, setting up broader audience (bigger chance to have more
frequent and cheaper CPA) and higher budget.
If you do that you won’t have problems with delivery of your advert sets.
So to sum up:
Budget < CPA for which you optimise -> BAD!
Budget = CPA for which you optimise -> BAD!
Budget > CPA for which you optimise -> GOOD!
When you set up the campaign and you get to the engine - the budget is very very
important. Sometimes Facebook has to prioritize advertisers with bigger budgets also
because the autobid needs to spend the budget.
Optimisation goal
When choosing your optimisation goal, always think about the magical number 50.
Facebook’s algorithm works the best when you have at least 50 conversions per advert set
per last 7 days. It is the minimum for performance campaign. After reaching this number of
conversions weekly the algorithm is able to say where is the limit, how to optimise for the
best cost per conversion etc.
The system needs data points to learn not time.
That is why Purchase is not always the best for optimisation. We should always think about
which goal is achievable for us to have it 50x per a week. So if we can’t deliver so many
Purchases per adset, it is better to optimise for different events in the funnel, for example
AddToCart or ViewContent.
If you have 5 advert sets and each one has around 20 events per a week, it is better to
make only 2 advert sets and each would have 50 events per a week. You can experience
CPA drop even by 50% by changing the structure.
Selecting conversion window
This is another important part for getting the best results. When we select very strict
conversion windows for optimisation such as 1 day after click or 7 day after click we go and
let the algorithm look for the most expensive conversions and most expensive users.
If we also include 1 day after view in our optimisation we go also for people who don't
click much but they can keep the message in their heads and they can convert later.
Different bid types = delivery to different types of people.
Facebook says that “clicky” users are 5,5x more expensive than the 50% of users that are
least clicky. That’s why they recommend to you to include also 1V into your conversion
window for optimisation for better results and higher scale of your delivery.
Bidding
When you start your campaign and you don’t know where are your limits, where does your
optimal CPA lays, automatic bidding is very easy and safe to use here. Usually autobid
helps you to spend your daily budget and get you as low results as possible. But bear in
mind that it needs data to learn and start getting your results down which could take few
days.
But still you should always try to think about what you want to do when setting up your
bidding. That is because the algorithm is optimising based on what you actually choose in
terms of objective, optimisation goal and the conversion window.
If you optimise for video views, don’t expect sales. Autobid will try to get you the lowest
CPV possible but it would not care about conversions.
The same is when you optimise for clicks instead of conversions. The algorithm will try to
find you the cheapest clicks possible across all platforms and placements, but it would not
care if the people generate some sales afterwards or not.
Once you know your optimal CPA it is advised to change from Autobid to some bid cap:
Target cost or Bid.
If you set up the manual bid just remember that Facebook needs to have budget big
enough in order to win auctions and deliver properly.
Example: The client wanted to scale but had KPI and bid for Purchase 6 EUR. They
went back one step in funnel from Purchase to AddToCart optimisation and
actually set up bid for 10 EUR. So much much higher. And what happened? The
Purchase CPA dropped to 3,5 EUR from 6 EUR and CPA for AddToCart was
around 0,8 EUR. So don’t be scared to try and set up higher bids. Try to be more
aggressive.
Placements
It is important to target people across all placements of Facebook, Instagram, Messenger
and Audience Network.
When you are creating new campaigns try to have less advert sets and less segmentation
and leave it more on the algorithm. The best practice is to have only one campaign to be
running in all placements.
Facebook is woking on the algorithm and auto-placement very intensively and they make
sure that the ads look good across all placements.
Don’t forget about Messenger and Marketplace and any other new placements possible.
Think about this: Advertisers are scared to go to new placements without previous results.
This way there is smaller competition and you could have much cheaper CPA there.
Targeting
When setting up a new campaign for prospecting Dynamic Ads (DABA) do these 4 things:
- Exclude all users who have purchased something in the last 10 days. For broad audience
don't exclude or include anything else.
- Use core targeting, be broad.
- Facebook doesn’t recommend you using interest or connections targeting to narrow your
audience and lookalikes. It is much better to work on your events and products in
catalogues.
- Do not exclude website visitors except for purchases.
THE AUCTION
In this section I would like to share with you few charts and explanations about the auction
which we were provided at this Facebook event.
To know which ad will be delivered to the user they always compare the Total Value of
each ad in the auction.
You can see on the image that the Total Value consists of three things: Advertiser’s bid,
Estimated action rates and Relevance and quality of the ad.
This chart explains why it is super important to select the correct optimisation goal and bid
for you. Because you will have advantage when optimising for conversions with people
who are likely to convert but you would have advantage when optimising for clicks with
people who are likely to click but not convert. This can help you win auction for the correct
people.
Factors that have influence on the estimated action rates (pink circle) are:
And then it is very easy to make an example of advertisers with different set up of their
campaigns and we will easily know who will win the auction.
As you can see on our example, Advertiser A has the highest bid while advertiser B has the
lowest bid of all of them. But because Advertiser B chose the best objective for that given
user in auction in the end it was B who won the auction. For advertiser A it was actually the
worst objective possible and A was the last in the auction even behind C.
TROUBLESHOOTING
There are few tools in Business Manager how you can check why your campaigns are not
delivering the way you want them to.
Audience Overlap tool
One of the reasons for lower delivery could be that your audiences are overlapping.
Imagine that you have one campaign targeting 1 lookalike from source A and then second
campaign targeting 1 lookalike from source B. Maybe there are same people in both
audiences. But how to check it?
Go to Audiences in Business Manager, select those two audiences and in Actions select
Audience Overlap. There you will see how many users actually are in both audiences.
If there is any overlapping only one will win the auction and they compete with each other.
If there is overlapping put those advert sets together and create one which will target all of
those users.
If overlapping is over 30% we need to take actions. If it is under 30% then it is still okay
(based on FB data).
Delivery insights
You can learn a lot about delivery of your advert sets also in the Delivery Insights tool. It is
a bit tricky how to access this section, you should go in Ads Manager to your advert set
and there go to the delivery column and you can find the link to this section.
In this section you can check how is your advert set delivering, if there has been any
important change in the delivery and other things such as:
Audience saturation - you will see here how many times the users have seen the ad. If the
users are seeing the ads very frequently and they don't convert, they will simply not
convert.
Check "First Time Impressions" and check if it is not going down. What to do? Broad
Audience - refresh data, make the audience broader. Prospect new users.
Auction competition - this depends on the things discussed previously. Check delivery
insights. Paced Auction bid vs. bid.
Once you will see some problems with your audience insights, it’s time to do changes so
you would maximise your potential.