Maximize Facebook performance by leveraging the algorithm
A successful Facebook campaign structure is based on objective, budget and target audience, which was explained in detail by Michelle Morgan during her SMX West session.
In this SMX session recap, Michelle Morgan of Clix Marketing covers Facebook ad tactics, along with tips for driving results with other Facebook ad formats.
This session definitely packed a punch.
Finding the right amount of volumes in the Facebook algorithm is similar to a space shuttle landing. You don’t want to overshoot it, but you don’t want to crash and burn by undershooting it either. You want to enter just right for landing success.
Before getting started you should note that traditional search structures do not work inside of Facebook. They are either too broad or too restricted. For success, you should follow structure based on these three components.
Three main components to Facebook campaign structure:
- Campaign Objective
- Target Audience
First and foremost, you’ll want to determine the campaign objective to optimize performance.
The Algorithm’s focus: The conversion action
The algorithm will optimize for the conversion action chosen during the campaign. You will then be able to leverage any goal you create, including standard and custom.
Choosing the conversion action that best fits your needs
Determine what is important for your account and campaigns so Facebook can help you optimize for it. You can choose from custom actions such as:
- Subscription Levels
- Order Value
- Number of Items Purchased
- Employee Accounts
- Annual Revenue
Thinking outside of the capsule (because in Michelle’s world it’s not a box it’s a space capsule): You can track and optimize for anything that Google Tag Manager can fire!
B2B company example: Driving in larger accounts
The company’s website had a sign-up form that provided a drop-down for users to input the number of employees they expected to manage with the tool (i.e. the more employees, the more money the company makes). This was used to leverage custom event audiences detailed below.
Use Google Tag Manager to get specific
Customize the Facebook script
Creating custom conversions
In order to create a custom conversion, the example changed the Website Event to the conversion action and then changed the Rule to Event Parameters > value. The rules were then setup for different employee counts.
The conversions were ultimately setup by employee counts.
Balance volume and specificity: Ideal conversion volume
To optimize for conversions you will need volume. One conversion a week will not cut it. Facebook recommends about 50 conversions. This may seem like a lot, but choose the action with the most volume and as specific as you can get.
Budgets: Find the happy medium
When it comes to budgets you want to consider not segmenting your budget too much. Here’s an example of an audience-segmented campaign that, from a high level looks great with a daily budget of $82.25 per day, unfortunately, it was segmented so much that each ad set only had $6.25 attributed per day.
Which ultimately lead to a terrible CPA… $527 when the goal was $150.
- Condensed audience groups
- Kept the same audience conversion action
- Total daily budget actually decreased from $82.25 per day to $50 per day
- CPA dropped from $527.64 to $114.75
Lifetime budgets vs. daily budgets
If you are using a lifetime budget, the budget will be balanced out for the duration of the campaign. Facebook will optimize spend (increase/decrease) based on conversion action performance. If you are using daily budgets, the spend will be consistent each day. However, Facebook will work to spend the entire budget regardless of the conversion performance.
Lifetime budget: This can be perfect if you know your total campaign budget. It also is best for fluctuating conversion performance. You will want to ensure you are using ad scheduling and your conversion tracking is in place and functioning. *Not ideal for campaigns with changing date ranges or budget fluctuations. Every time you make a change, the algorithm has to re-evaluate.
Daily budget: Responds better to change than lifetime budget and there’s no need to use ad scheduling. While these will be the most consistent in regards to ad spend, budget may fluctuate during the run. This also can be used to share spend across sets without using the lifetime budget. *Great for evergreen campaigns or undetermined campaign end dates.
Budget optimization for daily budget
Take budget from the ad set to the campaign level. Facebook will optimize for the ad set that is performing best.
Budget optimization: Change in budget can mess with algorithm
Budget trend line: The top line is conversions, the bottom line is daily spend. As the budget scaled up, Facebook was also scaling up with conversions. Campaigns were then pulled back based on budget restrictions which resulted in a small dip. The campaign was left alone after the decrease and it still ultimately gained increased conversions. Then the campaigns were pulled back drastically due to budget restrictions and the results tanked.
Facebook budget best practices
- Set your ad set daily budget at 5-10x target CPA
- Ideally, you should shoot for 50 conversions per ad set per week
- Do not alter daily budget by more than 20 percent unless absolutely necessary
- Pausing campaigns or ad sets can hurt performance progress
Target audience: Start with low hanging fruit
Think of the Audience as Mission Supplies: The Longer Your Mission, the More You’ll Need
For longer run campaigns, you’ll want to test to extend reach without oversaturating your market.
Understanding scale in audience building
When we narrow our audience we make it tougher for Facebook. You can keep track using the sidebar “mission control” as you build out your audiences.
What if reach becomes unavailable?
It may be hidden for privacy reasons. You’ll want to test removing certain portions of an audience targeting, likely customer lists until it comes back into scale. When you’re ready for launch you can add other targeting back in.
Long term success in audience targeting
Start with the lowest hanging fruit and then expand out to find additional reach. If the audience is limited, those users will see the same ads again. To help guard against this, enable Interest Expansion.
Lookalike audiences: Lean in based on your root list
- Engagement Audiences
- Website Audiences
- Page Audiences
- Custom Audience Uploads
When using lookalike audiences it is important that you are starting with good root list data.
If you put good data in, essentially you’ll get good data out
If you put in a good patterned list where people behave similarly, you’ll get a good lookalike audience back. If you put a list of just anyone who shows up, you have no idea what’s going to come out the other side.
Finding meaningful segmentation
Don’t just target anyone who comes to your website, unless you have relatively low traffic. Use some of these root examples:
- Thank You Page Visits
- Pricing Page Visits
- Video Viewers
- Company Size
- Returning vsbefore One Time
- Service Used
- Product Category
- Job Title
- Sale vs Regular Price
These type of roots will provide a good audience model back. However, you’ll want to find the happy medium when building. You’ll want to ensure it’s not too narrow or broad when determining what these should be.
So how many users in a root list? Facebook’s official answer is 10,000 to 50,000 users in a root audience list. However, 500 should be at the low end. If you are under that number you can still upload the list to see what happens. However, if you are in the high end of 50,000, you should see how you can segment the list better.
Houston we have a problem: 3 most common issues with campaigns
Low conversions: If your campaign experiences low conversions, you should revisit the conversion action you’ve chosen to optimize for. You may need to determine if there is something with higher volumes you can use.
Small budgets: Look to condense small budget ad sets where possible. You also should revisit the budget modeling you are using. Does another model make more sense to use?
Small target audience: Lower budget for this group so you don’t oversaturate. You can create lookalike audiences from converters to have a larger, constantly refreshed audience to target.
When things are going well: Ad testing
There are some challenges when testing ads on Facebook. You always end up with one ad with a ton of impressions and the rest with not nearly as many, and then it doesn’t really correlate with the cost per result. This is due to Facebook only caring about relevancy score.
Challenge: Cost per result
Actual focus: Relevance score
To impact your ad testing you have to understand how Facebook rates the relevance score.
Relevance scores are based on how well your ad is performing as well as its positive feedback and negative feedback. However, it’s completely estimated based on only 500 impressions. Here’s an example of what it might look like (this is only to provide an idea: NOT SCIENTIFIC):
How to avoid negative feedback: Don’t oversaturate people so they hide your ad.
How to leverage positive feedback: You can use top-performing, existing posts in your ads by using the post ID or you can set up an engagement campaign, let it run for a few days and use those top performing post IDs.
Getting your top post IDs:
‘They’re not just going to find a bunch of losers on Instagram’ – On placements
The audience and quality set is the same across placements. Choosing placement prior to understanding performance can limit results.
Before deciding, make sure you understand the placement options and that they differ by device. The best test for knowing if you should run the placements is asking yourself “would my ads look good here?” If the answer is no (looking at the preview), remove the placement from the list, or you also may decide to create new media for those placements.
Concluding SMX insights:
- Leverage the algorithm in every way
- Focus on the goals you really want
- Keep budgets and target audiences to scale within duration of the campaign
- Ad testing takes a lot of work
- Change your perspective on placements
Opinions expressed in this article are those of the guest author and not necessarily Search Engine Land. Staff authors are listed here.