Optimization
Once targeting and amplification efforts are in place, ongoing optimization is necessary to
maximize efficiency, reduce wasted ad spend, and ensure that campaigns continue driving leads
and conversions. For local service businesses—such as plumbers, roofers, pest control and
pressure washers — this means refining targeting, adjusting budgets, and improving content
performance to achieve the best return on investment (ROI).
For Johns Hopkins University students, this phase focuses on analyzing real-world campaign
data, identifying areas for improvement, and applying systematic optimization techniques.
Dennis Yu | Optimization
❏ Apply Metrics Decomposition.
❏ Compare the current period against last period.
❏ Using Audience Insights, create new saved audiences.
❏ Review budget allocation by channel and ad set based on performance (watch for
statistical noise).
❏ Increase relevance positive and negative audiences.
❏ List 3 to 5 top recommendations to execute in the next 7 days.
❏ Apply Top N to the data set and explain the results in terms of Goals, Content, and
Targeting for each level.
❏ Refine lookalike audiences.
❏ Update Success Tracker.
Optimization Checklist
Apply Metrics Decomposition
● Break down campaign results into key performance indicators (KPIs) such as Cost per
Lead (CPL), Click-Through Rate (CTR), and Return on Ad Spend (ROAS).
● Identify which metrics are underperforming and need adjustments.
Compare Performance Periods
● Analyze the current campaign period vs. previous period to track trends.
● Look for seasonal variations (e.g., increased demand for HVAC services in summer, roof
repairs after storms).
Use Audience Insights to Optimize Targeting
● Evaluate which audience segments are engaging and converting.
● Create new saved audiences based on high-performing segments; Exclude
low-engagement audiences to focus on high-quality leads.
Review Budget Allocation by Channel & Ad Set
● Adjust ad spend to favor the best-performing platforms (e.g., if Facebook Ads are
outperforming Google Ads, reallocate budget accordingly).
● Avoid statistical noise by ensuring enough impressions and conversions before making
changes.
Increase Relevance by Refining Audiences
● Expand or narrow audience segments based on CTR, conversion rates, and
engagement metrics.
● Improve positive targeting (focus on the best-converting groups); Adjust negative
audiences (exclude irrelevant or low-performing segments).
List 3 to 5 Actionable Recommendations for the Next 7 Days
● Example actions: increase budget on high-ROI campaigns, refine targeting, tweak ad
creatives, adjust bidding strategies, test new lookalike audiences.
Apply Top N Analysis
● Identify the top-performing pieces of content, audience segments, and ads.
● Use findings to scale successful strategies and pause underperforming ones.
Refine Lookalike Audiences
● Test 1%, 3%, and 5% lookalike audiences for customer acquisition.
● Segment based on website visitors, video watchers, lead form submitters.
Update Success Tracker
● Document performance improvements in a Success Tracker spreadsheet.
● Include before-and-after metrics to measure optimization impact.
Beginning Optimization
The more complex your campaigns, the more time it takes to optimize them. The more ads
you’ve created, the more “weeding” your garden needs. So don’t go crazy making a ton of ads
at once – just 5-10 per campaign within the 3 campaigns.
That’s why we advocate the simple three-layer campaign strategy of Awareness,
Consideration, and Conversion. These campaigns are “always on”, so they’ll continue to pass
your traffic through these three levels automatically.
When you have only a handful of ads in a campaign, you can quickly see which ones to kill off
and which ones to create more variations from. This is a modified version of “winner stays on”,
where you keep cloning the top performer with similar audiences, but slightly different.
Remember to create ad groups of one or multiple interest filters. Most of our ads have 3-4
filters each. Some might have 7-8 filters on them. Having multiple combinations is called “onion
targeting”.
For a new set of campaigns, you should check in 2-3 times a week, spending only 15 minutes
each time. Better to do a few adjustments over time than try to create a ton of ads at once and
have only one cycle of optimization.
You want to work with just a few ads per campaign, allowing you to nimbly pause a couple ads
(don’t hit “delete”) and create a couple new ones.
This allows you to quickly get to statistical significance – a minimum number of clicks to be able
to tell whether the difference in performance between a group of ads is a real difference or just
random noise. A general rule of thumb for people who forget what they learned in stats class –
get 10,000 impressions or at least 20 clicks for any ad. Any less and it’s noise – you can’t really
tell.
Once your pilot is successful
Once you reach a point where you’ve been able to prove your campaigns are successful, then
you can scale up the daily budgets. That also means you can readjust your audience sizes to
reflect the larger budgets.
At this point, you’ll be creating more and more audience targeting combinations to include
custom audiences.
You then clone your three campaigns so you now have two of each– two Awareness, two
Consideration, and two Conversion. Except you have one “test” campaign and one
“production” campaign for each type of campaign.
The test campaign now becomes what you’ve been experimenting with, while the production
campaign is much larger. When you find a winner in the test campaign, you copy it over to the
production campaign.
You wouldn’t test out new ads in the production budget since it could risk eating up the larger
budget on a larger audience. For those who are veterans of Google Ads, this is called the
“paste and stick” method.
Super pro tips
Keep in mind that if the audiences are small for an ad, it would cap out on the audience well
before it capped out on the daily campaign budget. Budgets are set at the campaign level, not
the ad level, so you still need to be careful.
Facebook doesn’t have frequency capping or ad rotation features natively in Power Editor, the
regular ads interface, or the API – but we hear it’s coming. Meanwhile, that means you have to
watch your frequency per ad carefully, as well as look at Frequency at the campaign and
account level.
If you’re doing a great job at onion targeting, some of your users are seeing multiple ads from
you–perhaps not the same piece of content, but different News Feed posts, dark posts, and
sponsored stories.
You know that reach x frequency = impressions. Facebook shows reach and frequency, but not
impressions, but you can calculate it easily. They choose to show reach and frequency instead
because it’s more actionable. Your reach tells you who are exposed to your ad, while
frequency is how often those folks see it.
If you have a reach of 10,000 people and an average frequency of 10, then you’ve served
100,000 impressions. A synonym for reach is unique impressions – or unique users who have
seen at least one impression.
In general, the News Feed placements (desktop News Feed and mobile News Feed) are better
for awareness and consideration campaigns. For conversion campaigns, sometimes the RHS
placement performs better – better CPC or better CPA. You just have to test. It may be that
non-social businesses (think of things like hemorrhoid cream, industrial plastic extrusion
machinery, or personal injury attorneys) produce content that people wouldn’t want to actually
react to, even if they’re interested.
We optimize using our Facebook Optimization Checklist, but with specific adaptions for the
Warriors:
- Make sure campaigns are turned on/off depending on the games they are promoting.
- Find the best performing page posts and boost them before, during, and after each game.
- Monitor the mix between remarketing, lookalike and interest-based ad sets.
Use our Top N and MAA approach to optimize ads. - Focus on campaigns, ad sets, and ads with the biggest budget and conversions.
- Look at metrics (CTR, CPM, Conversion Rate).
- Create an analysis (Why is CTR high/low?).
- Take action (Change the ad, ad set, or targeting options.
Boost Optimization
Kill 50% If after 7 days of a dollar a day ($7 spent total), we don’t see
results (meaning the cost per engagement or average watch
time isn’t good), then just let it die. Do this 90% of the time- don’t
waste your money.
Expand 5% If it’s pretty good, we can then add another $30 for 30 days.
Perhaps 5% of the time you’ll find a worthy post- usually a video
that gets above 20 seconds in average watch time or a regular
post with over 10% engagement (interactions divided by
impressions).
Switch-boost 3% If it’s a winner, change the audience we’re targeting in the same
boost, right from the timeline or Pages Manager. Usually, we’ll
initially boost to generate high traffic (BuzzFeed audience to a
post on BuzzFeed), switching to a media workplace audience a
few days later. Or we’ll switch to whatever audience we currently
want to incept– perhaps we’re speaking at the Social Media
Marketing World conference, so we’d switch a bunch of posts to
that audience temporarily. Some posts live forever- they keep
producing for months or even years, often triggered by custom
audiences.
Stack-boost 2% If it’s a “unicorn” (off the charts performance), duplicate the ad
set inside Ads Manager so that we have another ad set running
concurrently. When something is doing really well, we want to
have multiple audiences seeing a post, which means we need
multiple ad sets. This is not possible inside the timeline, but only
takes a minute inside Ads Manager or Power Editor. The risk of a
unicorn is that if you run it too hard, it will burn out. We’ve seen a
few companies accidentally kill their unicorns by literally
spending millions on them– derivative pieces of content so you
can get a similar effect from similar posts.