What do you think first when we say liquidity? You will probably think of what everyone else thought – cash flow and finances. But that is not what is about here. In this particular case, the word liquidity has to do with advertising and machine learning systems in advertising.
Such systems are widely employed in different industries and are used for different purposes. And how does it work in the context of digital marketing?
What’s the Deal With Liquidity in Advertising?
In this case, this system is used to help us make important decisions and real-time optimizations related to advertising and bidding advertisements on the advertising platform.
There are many different factors that affect the price of an ad as the auction system we use in advertising is very lively and changes quite quickly. At any given moment, thousands of advertisers are competing in their attempts to reach the desired audience in a unit of time. Clearly, you need a way to scale and systematize the approach to the algorithm. Also, there is a need to choose the right devices, platforms, and placement destinations that give the best effect. This is, in essence, the definition of machine learning on the Facebook advertising platform.
The machine gives us the answer to the question of which ads work best for us and how to most efficiently allocate a budget or creative. With the help of this system, we are given the opportunity to create flexible campaigns without restrictions that can stifle results. It can be said that we set everything in automatic mode, keep everything on, and give the algorithm the freedom to manage all possible options and look for the one that gives the best results. This is where the advertising platform liquidity story begins.
Liquidity in advertising means to allow the machine as many things as possible, i.e. to set as few restrictions as possible when creating a campaign and give a lot of room for maneuver for the algorithm to start learning, measure results, and use all data without limits. Information should flow freely, on all fronts.
What Kind of Campaigns Are Possible?
For example, you have a campaign that targets certain people. What will happen if you strictly limit the target group? You will reach a smaller number of people, for sure. If we expand the target and make it less restrictive, we allow better flow, i.e. liquidity. We give more freedom to the algorithm to reach the best target group in real-time.
For a machine learning system to work properly, you must first know your goals. You need to think strategically, to know what you really want to achieve. Are you working on the brand awareness part or do you want to reach solid sales through advertising? You simply need to be clear about this before setting up the system and filling it with information.
The next thing is what creative solutions we present. Creatives and goals are not something we create mechanically. The system does not control it. This is up to us and our heads. A good example can be this: if you use the GPS to get to a store, the navigation device (algorithm) will take you to it in the shortest and fastest way. But first of all, before any of this happens, you have to ‘tell’ the system which particular store you want to go to.
That is why the following question should be asked first…
Where Am I Now and Where Do I Want to Go?
The machine is quite useless if you do not have a created strategy or the power of strategic thinking. The conclusion is that you feed the algorithm with data, and that data must be of high quality, relevant, and of sufficient quantity for the system to be able to learn based on that. After the learning phase, the system will be ready to make various decisions and optimize your campaigns in real-time.
For example, a lot of the data we feed the machine comes from the site. How do we know which actions are taking place on our online platform that we have built for the needs of the services we offer or the business we have developed, and how do they affect the machine?
We know this if we have a Pixel installed, a piece of code that can initiate various events that measure our conversions. Pixel thus collects information and has the ability to tell the system various data about the behavior of people on the site. It helps you understand what actions happened and when. Also, there is an SDK that is similar to a Pixel and only applies to actions on mobile applications.
Liquidity Tools
There are four tools that extend or restrict the freedom of the algorithm:
Campaign Budget Optimization
Within this optimization, we do not divide the budget by ad sets but at the campaign level. This means that the system will control each ad set. For example, if one set has spent the budget, and the other still has it, and it happens that people from the first one go online, then a new opportunity for a result arises. If we divide the money by sets, we will not reach those people, and if the division is at the campaign level, the system will transfer the money where new potential buyers appear.
In essence, the budget is spent where the goal of the campaign is cheaper. Campaign-level spending is also Facebook’s recommendation for creating a campaign structure and certainly gives the algorithm more freedom to make valid decisions to optimize them.
Creating an Audience
The tool for building your audience is quite effective. You can ‘catch’ people in different ways – for example, those people who put something in the cart on the site and did not complete a purchase. And then you want to sell them exactly that product from the cart. On the other hand, you can expand your target, find people similar to them by demographics or geography. There is a possibility to find people of specific behavior or interest: those who are married, those who are engaged, parents, people who were born in a certain month and their friends…
There are many possibilities but you must not overdo it and do not put too much information and restrictions on the system. You need to leave room for the machine to find new possibilities. If you enter interests, enter more of them. This expands the potential target group. The algorithm sees it as a union of sets, where each set is one interest. If you are adding people similar to the people who performed the action (lookalike audience), do not add additional restrictions by city or interests. There is too much information and the flow is narrowing.
Automatic Ad Placements
There are four major placements for ads: Facebook, Instagram, Audience Network, and Messenger. Each of these main ones also has subgroups. With automatic placement, your ad will appear on all destinations in the Facebook system. The liquidity of machine learning is then very good. The system will push forward destinations with lower ad placement costs.
If we place an ad only on Facebook and turn off Instagram, there is a possibility that we have potentially turned off a cheaper destination that would bring a better result. For more placements we place, there is a greater chance of finding cheaper conversions, whatever they may be. Another very important thing is that you do not have to change the complete campaign if the result is bad or too expensive. This is a simpler process in essence.
Optimization of a Creative
The next thing in the process of automatically placing ads that you have to ask yourselves is how your creative looks on different placements. The dimensions of the visuals for Instagram and Facebook stories are different from the visuals on the news feed. The platform allows you to set up one creative and the system will automatically adjust it for different placements. This option is useful, but it is not always recommended, and it is not a very fortunate solution either. There is a better option: you can directly set different visuals for different placements.
What you also have as an option is dynamic language optimization, but this is by no means recommended. The system translates very clumsily from one language to another. The recommended option is definitely dynamic creative placement. You are able to place several different texts and visuals on one ad, with different titles and calls to action. Based on all this data, the system creates different combinations of ads, and based on the results, it decides which creative to push forward. Again, we come to give the machine freedom even in the creation of the ad.
Conclusion
There are definitely more levels of liquidity. When advertising on the Facebook platform, you see that a lot of things are happening ‘under the hood’. Various tools – from budget optimization, audience, placement, and even the creatives themselves – are very useful and allow more flexibility to the machine learning system which makes it more liquid, has a good flow, so you get better results and lower prices of your campaigns.
In reality, you raise the efficiency of the algorithm over good data. This is the only way to turn the machine to work in your favor. You are the navigators – if you do not guide the machine well, it will not know the way itself.
