Return to site

Optimization on Facebook and Instagram: Part 1 of 2. A/B tests

We prepared a client case to describe the Aitarget Audience Splitting Tool

What is Optimization on Facebook and Instagram?

Optimization is the process of assessing campaign progress, collecting and analyzing data, and modifying campaign components based on the findings.

Instagram and Facebook campaign optimization reduce the unit cost per target action (ie. goals or impressions). Campaign optimization has many components including campaign settings editing, audience segmentation and ad creative analysis.

The Facebook Campaign goal optimization is an algorithm that gathers the most relevant and cost effective audiences for your ads.

We’ve compiled a few general principles of ad optimization for Instagram and Facebook:

  • Work with the entire audience, segmenting audiences where possible

  • Measure every campaign element such as comparing the CTR of the same creative on various audiences, and displaying alternative creative to the same audience

  • Test everything: different demographic targeting, buying type, placements, and bids

  • Use multiple versions of ad sets: different sets of creative per mobile and desktop and alternative CTAs

  • Use your CRM data for optimization: retarget users and advertise for lookalike audiences

  • Optimize ads and ad sets: switch on/off ads or adsets, or increase and relocate budget depending on performance

Ad Sets Optimization

Key ad settings for optimization include: budget, buying type, payment type, and bid. Buying type can be auction bidding, fixed price bidding, or reach and frequency buying.

There are two types of payment options: either right after an impression, or after the target user action (calculated from the value of impressions). In the second case, the algorithm manages risk and therefore narrows the audience to exclude irrelevant impressions at auction. Discover more information about buying and payment types here.

The bid type is chosen by the advertiser, and depends on the desired action and payment type. It allows you to inform Facebook of your desired maximum Cost-per-Result for which optimization is going to be performed (ie. Website Conversion, Mobile App Installs, Daily Unique Reach, Page Likes, Page Post Engagement, Video Views, Clicks or Impressions).
When choosing a bid, use the following guidelines:

  • Set real value: define the purpose of each ad set, and determine the maximum amount you are willing to pay to achieve that goal. For example, if the desired outcome of your ad is Page Likes, decide how much you’re willing to pay per Like.
  • You can allow Facebook to automatically select the bid. Facebook optimizes your bids to show your ad to a defined number of people from your selected audience, in order to maximize the results achievable within budget.

Optimization of ad sets within the campaign include strategizing to plan each step of the total budget allocation among Ad Sets, and modifying settings to reduce the overall cost and increase the ideal number of target actions.
Optimizing the ad set level can be in the form of an A/B-test (or splitting), or its various forms and variations with the same audience. For the first few steps, you can measure the effectiveness of a particular ad, comparing results with others in a fixed time by the number of targeted actions, Spent Amount, Cost per Target action, CTR (Clicks / Impressions), Conversion and Relevance Score.

Optimization Options

There are various optimization parameters to pay attention to. When A/B testing, you can create several different ad sets and configure them based on need, and hypothesis to test.

 

Targeting: You can choose any demographic combination, interests, behavioral characteristics, and custom audience if you have the opportunity to test them. Testing the top percentage of lookalike audiences will benefit the campaign.
Buying/Payment Options: Typically, advertisers ask us whether they should select Website Clicks or Website Conversion as the target action. Ideally, the test is implemented with a small portion of the budget, and the data is assessed to determine the most effective and valuable buying/payment options.
Placement: Over time, the placement audience may vary. It’s necessary to assess campaign results in the initial stages for each placement, to estimate the budget required for specified type of placement.
Schedule: Choose how long you want your ad set to run. You can set a start date and time, and then choose to run your ad set continuously or set an end date and time.
Post formats: Advertisers often wonder what ad format is best: picture, video, carousel ad, lead forms, or a post with a link? Here is the simple rule to follow: use video if you care about the broad reach of your campaign, and use carousel ads to increase Online Conversions.
Creative: Of course, creative execution is crucial to campaign success. Depending on the characteristics of the audience, advertisers display one or another creative set.

These parameters complete the checklist of a fully optimized approach.

A/B tests and manual optimization

We reviewed the basic principles of an A/B-test (or splitting). Now let’s look at our tool, which helps to break the audience into non-overlapping segments by gender, age, geography, and devices.
When we take the example of running ads on Facebook: the campaign deploys, but at some point, you find that a given target (CPC or CPA) is quite expensive, and you’re projected to exceed the allowed the budget. What happened?

Often advertisers, especially beginners, face this problem; it’s associated with a lack of A/B-testing. If we take a neutral picture and show it to the entire target audience, then we won’t collect valuable data or achieve effective results. Alternatively, running an A/B test could mean showing advertisements of dresses and skirts to women for example, and snickers, shorts and t-shirts to men. Advertising results improve because the right message reaches the right audience.

The basic principles of A/B-testing

Below we identify the main principles of an A/B-test (which applies to any similar advertising channel).

  • Each audience has its ad set with the same creative. Ideally, the audience is divided by gender, age, geography, and devices for maximum results.
  • Each audience is set a budget and bid, and each audience will have standard measurement parameters
  • Each audience should be shown the same set of alternative creative. Thousands of publicly available case studies show that different audiences absorb and react to information in different ways. This method allows advertisers to assess which set and what combination is most effective.
  • You can add complexity to the hypothesis after each test iteration, for additional testing

The problem to be solved by Facebook Auto Splitting

Despite the fact that the core principles of splitting are relatively straightforward, the action of splitting requires a huge time investment from the ad manager. Here is a checklist:

1. Manually create a vast number of ad sets:

2. Prepare a complete batch of creative combinations (at four pictures and two texts 2 * 4 = 8 combinations) for each ad set:

3. Keep track of the results (for instance, hourly).

You can automate steps 1 and 2 (Part: Ad Sandbox).

Find how it works at our help center.

Client Case: turning off ineffective ad groups on Facebook

One of our clients is a large medical service provider with more than 1,500 clinics in Moscow. The client has a blog and an online consulting option for users. They set a campaign goal of 200 calls a day from Facebook at the cost of up to $8 USD (~500 Russian Roubles) per day. The campaign generated 11 calls per day at a cost of $17 (~1100 Russian Roubles) USD per day. The creative wasn’t reaching the right audience, the client wasn’t optimizing the campaign, and the results were disappointing

We recommended making automatic splitting by gender and age, and by Desktop and Mobile. Each segment had coverage of about 50,000 people. We found ad sets that generated calls within $8 USD (~500 Russian Roubles) per day.

Moving forward, we only ran the ad sets that performed at a cost per call under $8 USD per day, and calls per day increased to 50.

In our next article (part 2), we describe the automation rules tool and prepared a client case. Subscribe to our mailing list to receive the second part of this article and get a weekly newsletter.

All Posts
×

Almost done…

We just sent you an email. Please click the link in the email to confirm your subscription!

OKSubscriptions powered by Strikingly