"Data-Driven Thinking" is written by members of the media community and contains fresh ideas on the digital revolution in media.
Today’s column is written by Tom Manvydas, vice president of advertising strategy and solutions at Experian Marketing Services.
Many marketers believe that programmatic media-buying platforms will optimize their ad campaigns but, in reality, nothing could be further from the truth.
Part of the problem is that the role of the technology platform in the media-buying process is often misunderstood. While many marketers use third-party platforms for programmatic media buying – bid on the right impression, at the right time, place and price – the objectives of their ad campaigns are often at odds with the financial health of the platform.
Third-party platforms are multi-tenant, network-centric businesses. Therefore, the health and performance of the platform takes priority over the performance of any single advertising campaign. There is nothing fundamentally wrong with this – it is a business after all, but marketers need to understand this very important concept of network health trumping ad campaign performance.
Data-Driven Tactics Give Marketers More Control
The main benefit of platforms for programmatic media buying is the ease of executing digital ad programs. While many marketers use these platforms for performance-based marketing, actual performance benefits are often tied to the cost of media or the capabilities of the marketer, and not with the platform. In order for marketers to achieve consistent performance and improvements in their programmatic media buys, they must take control over how their ad campaigns are deployed and managed on programmatic platforms. First and foremost, marketers need to deploy the right data-driven practices that enable them to buy desirable ad impressions and filter out undesirable ones.
There are three underutilized data tactics that every marketer should consider for improving the results of their programmatic media buys:
1. Apply Large-Scale Analytics To Standard Media-Optimization Processes
Marketers cannot prevent fraud or low-quality media within their programmatic media buys but they can limit their impact. Traditional media optimization practices that reinvest buys based on performance aren’t feasible within programmatic. With millions of potential media sources, only systematic algorithmic processes can tackle the media optimization for programmatic advertising. Even the most basic form of this practice will give marketers a 10% to 30% improvement, right out of the gate. Marketers can start by building an analytics process around the 80/20 rule:
- Sort and rank all source domains, such as websites, top-level domains or networks – whatever is available at the most granular level – by the amount of spending from your ad program. Separate the top domains that represent 80% of your spending. This will be a surprisingly short list.
- Re-sort these source domains based on your desired campaign KPI, such as CPA, CPC or time spent, and then blacklist the bottom 10-20% from your ad campaign.
- Keep the other domains that only got 20% of ad spending (this will be really long list). Many will get additional spending since we just removed several domains that received a significant amount of spending.
- Repeat this process monthly.
This type of media optimization process should only be done with strong data analytics support. Marketers need to ensure that they are working with a solid computational mathematician who understands probability and relative value applications.
Over time, you will identify domains that should be blacklisted for every campaign vs. domains that should only be blacklisted for a specific campaign. By repeating this process, you will identify the ever-increasing number of nonperforming domains before they bankrupt your marketing budget. Some marketers attempt to solve this problem by whitelisting “safe” domains. This practice will rarely work since it will materially impact scale and performance. Larger media domains perform inconsistently from one campaign to another or not as well when significant spending is allocated.
2. Apply Relative Value Measurements To Your Audience Segments
Audience segments are generally based on some set of demographics. However, segments like “soccer moms” are diverse and hard to target effectively, even for brand advertising. A better method is to use the 80/20 rule again, applying it to your customer data:
- Group your customers into segments based on your preferred demographic definitions, such as young male urban professionals, wealthy retirees or Midwest blue-collar families. Sort and rank these segments based on their actual economic value to your product or brand based on your customer data, such as sales, profits, lifetime value and so on.
- Avoid targeting the bottom 20%, and for the remaining 80%, group these in two to four ad groups, such as “average” and “best.”
- Calculate the relative value between your “average” and “best” group. For example, your “best” customers may generate three times the profit as your “average” customers.
- Auto-allocate your advertising budgets across both of these ad groups but set your bid for the “best” customers to be three times higher than your “average” customers. You can also try three times the frequency or a combination of setting bids and frequency caps, based on relative value calculations.
The most important thing is to reassess your ad groups every one to two weeks based on the actual performance of the ad campaign. Over time, this tactic allows marketers to gather enough data to refine their ad groups and improve their audience targeting.
3. Use Segmentation Methods Based On Population Scoring Models
This approach can be particularly useful for performance marketers who have hard-to-define customer segments, or for those who want to focus on acquiring new customers. Population scoring is an algorithmic process to assign a unique economic value score to every member of your target population. This could be applied to the 80% of the customers we previously defined or to lesser-known prospect segments.
As long as you have a consistent scoring method and the ability to group the results, this quantitative process will add tremendous consistency and incremental performance to your ad programs. Again, use the 80/20 rule as the framework for this strategy:
- Sort and rank the results of the economic value score by value and group them into thirds, quartiles, quintiles – whatever is most appropriate for your marketing program. You end up with a set of segments defined by your economic scoring process, such as high-value prospects that look like high-value customers. If you examine your top quartile, you might find blue-collar workers and wealthy retirees.
- Put these segments into the desired number of ad groups and deploy. Refine this tactic further by running distribution analysis across your scored segments to help define more granular sub-segments and their relative values.
Don’t overcomplicate this. Start with the simplest thing you can handle and you will get immediate results. Ideas to improve this process will come faster than you can implement them.