top of page

4 Dimensions of Liquidity

Machine learning in advertising

Machine learning is now a fundamental part of digital marketing and helps show the right ads to the audience you want to reach. As you learn more about how we use machine learning, you can better understand how to set up your ad campaigns to achieve the optimal return on investment.

Questions machine learning can help answer

Machine learning uses serial algorithms and predictive analytics to find the best audience at the right time at the best price. It also processes countless data points in real time to help you answer questions such as:

  • Which devices work best for your message?

  • Which placements are more effective?

  • How should you allocate your budget based on the audience?

  • What audiences are most likely to take your desired action?

To get the most out of machine learning, you must first understand liquidity. Liquidity is the condition that gives the delivery system the flexibility to move ad spend where it’s likely to achieve the best results.

You achieve liquidity when you increase campaign flexibility across four areas: budget, placement, audience and creative.

When you embrace machine learning and use flexible campaigns, you can access new efficiencies and insights

Before machine learning can search for opportunities, you must first select the best ad objectives and optimization events for your business goals. Let’s consider an example.

Case Scenario

The marketing department at the big grocery chain X Market wants to increase its online presence. They just released an app that lets users order organic vegetables and have it delivered right to their door. Their objective is to have more people download their app.

Installs of apps are clearly the optimal goal in this situation. However, choosing the ideal objective by itself wouldn't produce the desired business outcomes in the most effective manner.

The team should use their data to filter out users who already have the app installed in order to direct the algorithm toward better results. They may prevent users who already have the software by using such optimization, which would help their campaign be more successful.

Therefore, the system need input to direct it toward attaining your business objective.

Translate business goals to ad objectives

To achieve the desired results, first inform the system of your business objectives. This is accomplished by selecting the appropriate ad objective for each campaign. Examine the examples of business goals provided below to learn which objectives make sense and why.

  • Brand awareness: Our delivery system looks to reach the people most likely to remember your ad.

  • Reach: Our delivery system shows your ad to the maximum number of people in your target audience.

  • Video views: Video is one of the best ways to tell your brand story, because it builds deep connections with people who see it. With video views, our delivery system targets people who are likely to watch.

  • App install: Our delivery system looks to reach people most likely to install your mobile app

  • Traffic or conversions: With both objectives, you can optimize your campaign for specific actions within the app.

  • Lead generation: Our delivery system looks to reach potential customers interested in a business. This may be in the form of subscribing to newsletters, scheduling appointments or filling out “contact me” forms.

  • Messages: Our delivery system looks to reach people most likely to start a conversation and move from consideration to action


What is Advantage campaign budget?

Budget liquidity refers to the degree of flexibility with which the system can allocate an ad campaign budget.

Advantage campaign budget is one tool that allows advertisers to increase liquidity and improve the effectiveness of their advertising. You can enable Advantage campaign budget at the campaign level in Meta Ads Manager. When you enable Advantage campaign budget, you define a single budget for your campaign, allowing the system to dynamically distribute funds to the best-performing ad sets.

Let’s compare using ad set budgets to using Advantage campaign budget. In both examples, we're using lowest cost bid strategy.

Ad set budgets versus Advantage campaign budget

Ad set budgets

There are two ad sets, each with their own budget. Halfway through the campaign, you realize that ad set 1 has spent all of its budget, but there are still abundant opportunities in that ad set. Meanwhile, in ad set 2, there are fewer opportunities and plenty of budget remaining.

Shifting budgets manually from one ad set to another would result in resetting the learning phase, and there's no guarantee that the opportunities would still be there.

Advantage campaign budget

The identical situation would occur if you used Advantage campaign budget, which would automatically allocate your campaign budget to the ad sets with the best chances for cost-savings. By doing this, it makes campaign administration simpler and removes the requirement for reentering the learning phase.



Advantage campaign budget works continuously and in real-time to capture the most results for your budget, effectively lowering your total cost per result.


Using Advantage campaign budget simplifies campaign management as there are fewer budgets to track. It also eliminates the need to recalculate budgets mid-flight.

The "breakdown effect

When using Advantage campaign budget and the lowest cost bid strategy, the system finds the lowest cost per optimization event for your overall campaign, not a specific ad set. As a result, you may notice that the ad set with the highest budget allocation doesn’t have the lowest cost per action. This is due to the “breakdown effect.”

Ad Set A

CPA: $1.46

Budget: $450

Ad Set B

CPA: $1.10

Budget: $50

Let’s say ad set A receives 90% of budget spend and has a higher CPA. Meanwhile, ad set B receives 10% of budget spend and has a lower CPA. If ad set B has exhausted all of its lowest cost outcomes at $1.10, then the system moves to ad set A to find the lowest cost outcomes there. While the CPA in ad set A may be higher, that doesn’t mean the system is spending inefficiently. It simply means the opportunities in ad set A were more expensive than the lowest cost opportunities that were already taken in ad set B.

Remember this when you analyze performance. Success with Advantage campaign budget occurs at the campaign level, not within a specific ad set.


  • Use campaign level budgets to increase budget liquidity. This gives the system more opportunities to allocate spend where it yields the best results.

  • Remember that Advantage campaign budget optimizes distribution across the overall campaign, so analyzing its results at the ad set level can be misleading.

  • Avoid using spend limits as it reduces the overall efficiency of your campaign budget.

Advantage+ placements

Placement liquidity is achieved by allowing the system to determine which placement and impression combination delivers the best results.

Advantage+ placements are the simplest way to boost placement liquidity. Advantage+ placements enable the system to take into account all possibilities and place ads where they have the best chance of achieving your campaign's goals. The technology can determine where ads produce the best outcomes at the lowest cost in real time thanks to more possibilities. This makes managing campaigns simpler and generally results in a reduced cost per outcome overall.

Best practices for brand safety:

Check each off as you review it.

  • Determine whether there are websites where you don't want your ads to appear.

  • Consider whether ads within publisher content are an issue.

  • Determine whether you want to use the Audience Network placement.

  • Evaluate who your customers are and what they consider appropriate.


The Law Office of Jeremy P. Harville

Jeremy P. Harville joined the US Army in 2007 and served for over 10 years. He transitioned from active duty to the civilian world in 2017, attended law school at Atlanta’s John Marshall Law School and graduated with a Juris Doctorate in 2020. Today his law practice encompasses numerous areas of law, including personal injury, general civil litigation, DUI defense, criminal defense, family law, veterans disability, social security disability, estate planning and probate, and criminal law.

Jeremy P. Harville was looking to attract potential clients and increase consultations.

The Harville team wanted to find the optimal way to reach more potential clients and strike up an introductory conversation, known as a consultation. So it partnered with the team at Meta to run an A/B test that compared two different ad targeting strategies:

broad targeting and detailed targeting.

In cell A of this campaign, the law office used broad targeting to show ads to US adults in Alabama aged 18 and over who lived in or near the counties of Mobile, Baldwin and Escambia. In cell B, it used detailed targeting to show ads to that same base audience, refined to only people who were also military veterans or had an interest in politics.

Both test cells used the same photo ad, which showed the balanced scales of justice and the messaging: “The Law Office of Jeremy P. Harville, LLC aggressively represents individuals and families in all of their legal matters.” The ad included a “Send message” button that, when clicked, opened a chat in Messenger where people could connect with the law practice for a consultation.

The law practice used the Meta Advantage+ placements feature to automatically deliver the ads across Facebook, Messenger, Instagram and Audience Network, based on which were most likely to drive campaign results at the lowest cost at any given time. It also used campaign budget optimization, now called Advantage campaign budget, to automatically distribute the budget across the best-performing ad sets in real time.

The Law Office of Jeremy P. Harville determined the results of this January 25‒February 15, 2022 campaign using an A/B test in Ads Manager, which revealed that broad targeting drove stronger results compared to detailed targeting:

  • 88% more new messages from broad targeting compared to detailed targeting

  • 15 new message consultations

  • 46% lower cost per result using broad targeting compared to detailed targeting

3 Audience liquidity

The technology finds persons more likely to do your desired action when audience liquidity is increased. This may produce better outcomes and assist in identifying unforeseen audience segments for upcoming marketing.

To increase audience liquidity, set broad parameters in your audience targeting. This increases the size of your audience and gives the system more opportunities to reach the right people for your objective. It also allows the system to better predict who's likely to take the action you are looking for. To achieve the best results for your campaign, pair larger audience sizes with customer interactions from your data sources. This guides the system towards optimizing delivery and also helps achieve the best results. Below are some specific tools and strategies you can use to increase audience liquidity.

Broad targeting

Broad targeting occurs when an advertiser specifies location, age or gender, but doesn’t provide any other audience parameters, such as interests. While narrow, detailed targeting may result in lower delivery, broad targeting tends to deliver more cost-efficient outcomes and higher reach.

Let’s see a real world example.

CASE STUDY: Halls Thailand success story

Introduced in Thailand over 56 years ago, Halls is the number one candy brand in the country. It is well known for its cooling “mentholyptus” flavor that combines menthol and eucalyptus to help refresh breath.

When government restrictions related to coronavirus (COVID-19) began to ease in Thailand, Halls wanted to find a way to raise brand awareness, positive message association and sales, while also encouraging people to reignite the Thai travel industry by visiting domestic destinations.

Halls partnered with the Tourism Authority of Thailand on a campaign titled “Breath of Thailand.” For a limited time, its product packaging showed images of locales around Thailand and each online purchase automatically entered customers for a chance to win a trip to a local destination. Halls also created a microsite that featured little-known vacation spots to spur interest in traveling there.

Halls Thailand worked with its partners Spirable, a Meta Business Partner, and Spark Foundry, a media agency, to launch the campaign. After receiving insight from Facebook that people viewed travel more positively on weekends than on weekdays, Halls wanted to incorporate that into the campaign structure. So the teams crafted a narrative that included the word “escape” on weekdays and “explore” on weekends, and Spirable helped Halls create more than 90 mobile-optimized photo and video assets based on five scenic locations across Thailand.

Spark Foundry then assisted Halls with setting up the campaign. There were two versions: one targeted to a group of people in Thailand between the ages of 20–44, with interests in travel, outdoor activities, culture and food, and another one that used dynamic ads for broad audiences. The broadly targeted dynamic ads campaign used Advantage+ placements, formerly known as automatic placements, to make the most of the campaign budget by automatically choosing the best mix of creative assets, ad copy and ad placements. It also included a rule to use specific creative and copy on weekdays and different creative and copy on weekends.

To measure the results of the six-week campaign, Halls Thailand conducted a brand lift study and an A/B test. The A/B test compared the campaign targeted to demographic and interest groups with the campaign that used dynamic ads for broad audiences.

The results:

The “Breath of Thailand” campaign received the highest number of campaign entries in Mondelez Thailand’s history, with 1.7 million entries. After measuring the performance of the “Breath of Thailand” campaign, Halls also discovered that dynamic ads for broad audiences were more effective, resulting in a:

  • 3.6-point lift in purchase intent.

  • 1.1-point lift in message association

Custom Audiences

When advertisers use a Custom Audience, they can use their own data and connect with people across Meta technologies in a secure and safe way. To create a Custom Audience, import a list of contacts (email, phone or user ID) or use the actions people take on your website or mobile app.

As a standalone product, Custom Audiences are a great way to identify where current and potential customers are in their purchase journey and move them down the funnel. For example, if you'd like to increase product sales for a specific item, you can create a Custom Audience based on the “add to cart” Pixel event and exclude people who've already purchased that item. From there, you can tailor your ad to that audience and encourage them to complete their purchase.

Advertisers can use Custom Audiences to drive effective remarketing, upselling and cross-selling campaigns:

Remarketing allows advertisers to reinforce marketing messages that people have already encountered. You can create a Custom Audience to reach people who researched a product on your website but didn't make a purchase.

Upselling encourages customers to move up from one product to one that's more premium and retails at a higher price. If you have different tiers of products or services, you can create a Custom Audience of customers in each tier and encourage them to move to the next level.

Cross-selling is the act of recommending products which are complementary to those a customer already has or is planning to purchase. For example, the owner of an online clothing shop could create a Custom Audience of customers who purchased swimwear from their website and send them a promotion for hats and sunglasses.

Custom and lookalike audiences

You can add a lookalike audience if you want to reach more people than are in your Custom Audience. Lookalike audiences enable the system to show ads to users who share traits with your Custom Audience, expanding your audience as a whole. The system has additional options to find the optimal outcomes for your target because to this rise in audience liquidity.

You can choose the size of a lookalike audience during the creation process. This size is expressed on a scale of 1–10, which is the percentage of the population of a particular country that you're asking the system to match with your source audience. Smaller audiences (1–5) match your Custom Audience much more closely than a larger audience (6–10).

Remember to consider the nature and specificity of the source audience. For example, a Custom Audience based on existing customers may lead to better results than a Custom Audience based on everyone in your CRM database. When you create your source audience, quality is often better than quantity.

This audience strategy is useful both domestically and internationally. Let’s review an example of how Custom Audiences can work with lookalike audiences to achieve results on a global scale.

CASE STUDY: DMCI Homes Success Story

DMCI Homes Success Story

Established over 20 years ago, DMCI Homes is a Filipino real-estate developer specializing in communities and condominiums that feel like resorts. The company employs engineering experts and builders with the aim of developing modern-day residential solutions for urban families.

DMCI Homes wanted to earn high-quality leads and boost sales and reservations for its newly launched property, the Oriana, while making the most of its budget.

Because Filipinos who reside locally and abroad often use Messenger, DMCI Homes wanted to test whether the platform could help qualify leads for its new condominium complex.

To gain insights, the real-estate developer set up an A/B test comparing two lead-generation campaigns using the same video creative: the business-as-usual campaign used lead ads and the test campaign used video ads with the Messages ad objective. When people clicked the lead ads, they arrived at a pre-filled form where they could confirm or modify their details and submit the form. When people clicked the video ads with the Messages objective, they arrived in a Messenger window where they could learn more about reserving or buying a unit at the Oriana via an automated question-and-answer experience.

To reach its target audience, DMCI Homes created a Custom Audience and lookalike audience of current customers, people who had engaged with its Facebook Page and people who had previously indicated an interest in buying a condominium. In addition, the real-estate developer targeted ads to Filipinos living overseas who were keen to invest in properties back home, and measured sales conversions through its own customer relationship management software.

After its monthlong campaign and A/B test in December 2021, DMCI Homes learned that using lead generation in Messenger successfully engaged the target audience and delivered high-quality leads more efficiently than the business-as-usual campaign. Results included:

  • 91% lower cost per click using lead generation in Messenger

  • 25% more qualified leads using lead generation in Messenger

  • 20% lower cost per qualified lead using lead generation in Messenger

Lookalike audiences can both expand your reach and increase audience liquidity. When using lookalike audiences consider the following:

REACH: Use source lists that are specific to the objective you want to achieve, such as people who tend to convert online versus offline. Then find people who look like them.

AUDIENCE CUNTRY SIZE: If your source audience is large relative to the size of a country, ads may perform poorly, because your source audience is too diverse. In this case, segment your audience into a smaller target group and create lookalikes of that audience.

AUDIENCE SOURCE SIZE: The size of the source audience you use for your lookalike audience should be from 1,000–50,000. Remember to use a high-quality source audience of high-intent customer actions from your data sources.

Advantage detailed targeting

Advantage detailed targeting is another tool that can broaden audience size and increase audience liquidity. When you create an ad campaign, you may be accustomed to using more detailed targeting selections for the people you want to reach.

In such situations, targeting expansion can show your ads to people who the system determines might get you more or cheaper results. The expansion is only implemented if it's likely to achieve those results.

Targeting expansion is specific to detailed targeting, and it doesn't apply to location, age or gender targeting options.

4 Creative liquidity - Personalizing ads at scale

Creative liquidity helps advertisers deliver more personalized ads at scale. This lesson discusses tools to increase creative liquidity and how machine learning can help you identify what creative combinations work best for different audiences.

Placement asset customization

Advantage+ placements or multiple placements generate better results. Without asset customization though, your ads all look the same everywhere they appear. With placement asset customization, you can easily and efficiently optimize the creative format of each of your ads for each placement. This is a powerful way to take advantage of multiple placements and also match the viewing behaviors of the audience on each of those placements.

Let's review how it works.

Once you decide which visuals to use in your ad, go to select a placement to customize.

From here, you can crop or edit per placement and placement group.

Or you can change the creative entirely and replace it with content that better suits the viewing behaviors of that placement. Asset customization provides a variety of ways to personalize the ad experience and increase creative liquidity.

Dynamic creative

Dynamic creative is a tool advertisers can use to increase creative liquidity. At the ad set level, enabling dynamic creative allows the system to take the basic components of your ad, such as visual elements or text, and automatically generate optimized ad combinations for your audiences.

Dynamic creative

Dynamic creative is a tool advertisers can use to increase creative liquidity. At the ad set level, enabling dynamic creative allows the system to take the basic components of your ad, such as visual elements or text, and automatically generate optimized ad combinations for your audiences.

The system then serves these ads across placements to compare the performance of each creative element with a given audience. Dynamic creative is an effective optimization tool when you’re unsure what resonates with different audiences.

Dynamic language optimization

Dynamic language optimization automatically translates the copy of your ads into multiple languages. This gives your campaign the flexibility to reach the right people in the right language at the right time.

Dynamic language optimization works well domestically and internationally to deliver ads to people in their preferred language.

To use dynamic language optimization, first define your objective, audience, placement, budget and creative elements. At the ad set level, select add languages.

Select your default language.

Then select the additional languages.

Note: Some placements don't support automatic translations. In such instances, people who speak the added language don't see your ad on these placements.

CASE STUDY: Bluelambda & EMUAID success story

Based in Singapore, Bluelambda is a full-stack ecommerce growth agency that helps online businesses scale beyond their home country by growing their current traffic and sales channels. Bluelambda wanted to help EMUAID test and improve the click-through rate and conversion rate of their campaigns in Canada by delivering ads in multilingual audiences’ preferred languages.

Bluelambda set up an A/B test that compared the results of a campaign using dynamic language optimization to a campaign using English as a single language. Both campaigns reached a lookalike audience based on a Custom Audience of EMUAID customers and used Advantage+ placements, formerly known as automatic placements, to deliver ads across Meta technologies efficiently.

Using an A/B test, EMUAID gained insights into how dynamic language optimization helps reach audiences effectively and efficiently. Bluelambda applied the results and learnings to build a cross-border business infrastructure to help EMUAID scale into APAC and Europe, as well as other global markets.

As a result of the campaign, which ran from August 4 to 8, 2021, EMUAID achieved the following results based on reporting data from Ads Manage

  • a 2.9X higher click-through rate

  • a 45% decrease in cost per action

  • an 80% increase in sales

1 view0 comments


bottom of page