Getting the most out of big platform optimization
Big ad platforms allow you to choose between automated and manual campaign optimization. Deciding which to use is complex, and the stakes can be high.
Automated vs. manual optimization
Platforms like Facebook, Google, and LinkedIn are essential to most digital campaigns. Each of them offers a solution for campaign optimization: you tell them which creative assets to run, and they will automatically find the “best” audience exposures to drive your main KPIs.
For example, I could run a Facebook campaign advertising a new restaurant with images showing burgers or hotdogs, and tell the platform to achieve as many conversions as possible for a given budget. Behind the scenes, Facebook leverages an economy of scale, first-party data, and sophisticated experimentation and optimization technology to determine the right creative to show to each audience in order to fulfill my criteria. This is how most agencies operate on these platforms for every campaign, and often yield excellent results.
But click-throughs are not restaurant visits. Suppose I have my own research telling me that people who own motorcycles are much more likely to buy hamburgers at my restaurant. Most platforms also offer the option to make that targeting decision manually: I can dedicate a fixed budget on a campaign showing images of hamburgers to people interested in motorcycles. In the right circumstances, Known has seen strategies like this outperform automatic optimization by margins of 10–15%.
So which to choose? Known’s media strategists and buyer-scientists make these calls based on our client’s goals, products, and audience intelligence. We use Skeptic, our proprietary science-driven media buying platform, to enable manual campaign optimization at scale when we believe it’s the right strategy.
Here are some things we think about when we’re making this decision.
- What are our goals?
Automated solutions work best when the ad platform can directly measure the KPIs we care about. When we run awareness campaigns for clients who want website visits, cost-per-click is a critical KPI that LinkedIn or Google can optimize directly. But for clients with long-timeline sales funnels, or where ad exposure is not directly tied to purchase, automated solutions run the risk of finding audiences who preferentially click but don’t necessarily buy.
2. How specifically defined is our audience?
Automated solutions work well for broad audiences because the big platforms can leverage scale to learn quickly. They’re also useful when a client knows less about who their audience might be, which is somewhat typical for new brands. But the more our strategists and buyer-scientists understand the audience, or the narrower the audience is, the more likely we are to target manually.
3. How critical is audience-creative pairing?
Often certain creative executions resonate better with specific audiences. Automated optimizers will usually find those audience-creative pairs over time, but they don’t allow the option of testing specific pairs. If we have a reason to think that certain audience-creative pairs will over-perform, or if we want to test them to make strategic campaign decisions, we’ll strongly consider manual optimization.
4. How important is the why?
Sometimes we’re after the most performant campaign possible, come what may. This is where automated optimization shines. It’s a “black box” that gives you a (hopefully) good result without an explanation. Sometimes we’re willing to sacrifice some performance to get insights that will fuel future campaigns. “How do my ads perform with women vs. men? Is it better to include images of green shoes or white shoes?” In those cases, we want manual control to set sample sizes and audience-creative pairs to test specific hypotheses. Most times we have a combination of these goals, and we need a mixture of the two strategies.
5. How long are we willing to wait?
Automated optimization is always associated with a “learning period” where a platform’s algorithms figure out who the right audience is. We’ve discussed this previously in the context of A/B testing and Multi-Arm Bandits [link this], but it’s always true of test-and-learn optimization. If we believe we can learn faster through manual optimization or if we need results right away to make campaign decisions, we prefer manual testing.
Why don’t more agencies consider custom optimization?
Many factors go into our decision between manual and automatic optimization for campaign, and we often use a mixed strategy that leverages the strengths of both. But this is uncommon for media buying agencies, who overwhelmingly prefer to use automated optimization. We think this happens for three main reasons.
- Automatic is the default, and it works a lot of the time. Representatives from these platforms prefer that agencies use their in-house tools, and tout their platform’s proprietary automated optimization as “best practice.” It allows them to leverage their economies of scale and perform arbitrage to extract margin. It also often leads to better campaign performance. Being able to make the right trade-offs and push back against platform recommendations when needed requires significant technical and media experience.
- Manual optimization is BYOM (bring your own model). To optimize manually, we need to know the value of a click or conversion, which is different for each campaign and changes over time. If agencies want to use this option effectively they need analysts, scientists, and engineers on staff to build and maintain the models. Not to mention, manual entry of campaign parameters can be slow to do at scale without the right technical infrastructure.
- Lack of data or audience intelligence. In order to “beat the house” at their own game, agencies need to bring something special to the table. Known is often armed with customer segmentations from our Strategy team, integrated first-party data from our clients, or third-party data from our partners to unearth consumer behavior that Google, LinkedIn, or Facebook can’t find through their analytics alone. Without these tools— and the strategists and scientists to use them well—agencies can’t hope to do better than the big platforms.
In-platform optimization at Known
Automated optimization at large platforms is a powerful tool that we commonly use to run performant digital media campaigns. Besides being a “default” option, it often makes sense to allow these platforms to use their first-party data and massive scale to find the right audiences.
But when we have more information than the platforms on client KPIs, audience intelligence, or creative performance, there are opportunities for improvement. Known‘s buyer scientists and media strategists work together to identify these opportunities. By leveraging Skeptic, our proprietary marketing science platform, to enable the speed and scale necessary to beat the platforms and deliver the best results possible.
We’ve listed a few of the criteria we use to make this decision, but there’s no simple formula. Platform policies around manual campaigns continue to evolve, and every client’s situation is unique. We work closely with their marketing teams to choose the best tool for the job, ensuring that their goals are met.
Read more from Known on media buying.
- Want more efficient performance campaigns? Sometimes less is more
- Using surveys to test creative assets? You’re not doing enough
- The importance of data-driven creative
Reach out to learn about our testing capabilities.
