Optimization Vs. Attribution: How They're Different, And Why This Matters
As online marketers become more sophisticated in how they measure and plan media buys across various online channels -- from display and search advertising to video and social marketing campaigns -- they’re also sometimes confused by the different ways industry pundits, peers and vendors talk about new terms and trends. Take, for example, two of the most popular topics on the minds of online marketers today: attribution and optimization. While they’re complementary, they play very different roles in helping advertisers improve their investments in online and offline media.
Before we get into how they’re different, let’s talk about one important common trait that both ad optimization and attribution share: the need to rely on technology automation. To be most accurate, advertisers must account for and understand the impact that their entire stream of media has on conversions. Those cross-media interactions are too complex to be analyzed and managed manually. Technology automation helps brands view different media channels holistically so they can pinpoint the components that are actually driving conversions.
While both attribution and optimization need to be automated, they also serve very different purposes in helping brands to improve their online marketing campaigns. Attribution provides the deep insights advertisers need to improve and plan future campaigns. Optimization produces on-the-spot recommendations on the incremental actions advertisers can take to improve media buys for campaigns that are already running. As we shall demonstrate below, two different methodologies are required for attribution and optimization.
Attribution is calculated using average metrics over a historical period. Optimization, on the other hand, relies on a methodology based on predictive models that track the non-linear relationships between specific campaign goals and spend levels. But brands are often misled to think the two can be used interchangeably. Such an assumption could be doing more harm than good for their campaigns. Let’s look at a simple example to illustrate how optimization works and why using attribution to optimize a campaign could produce the wrong recommendations.
Assume an advertiser has spent $100,000 so far this month across a number of sites and media channels and has generated 20,000 conversions (for a cost per action of $5). This $5 CPA is the average amount the advertiser has paid so far for the life of this particular campaign. Now, let’s assume that advertiser wants to increase the number of conversions for this campaign before month end. Using the average $5 CPA metric to make incremental optimization decisions would lead to the wrong conclusions. Here’s why.
Based on the above, the advertiser would infer that increasing spend by 10% (to $110,000) among these same channels would increase conversions by 2,000, or 10% (generating the same $5 CPA). But the reality is almost always different for campaigns that are already running.
For example, if there is low saturation (i.e., not all of the targeted audiences have been reached the maximum number of times needed to spur conversions) among the sites and channels, the advertiser may find that the additional, incremental spend generates more than 2,000 extra conversions (as the cumulative effect is additive toward conversion.) If there is high or over-saturation, the advertiser may find that the additional incremental spend generates fewer than 2,000 extra conversions (as users are overexposed to the media, reducing incremental conversions.) In either case, the rate of conversions isn’t constant over different spend levels of the campaign.
Therefore, to correctly calculate optimization recommendations, you need a methodology that produces recommendations based on the incremental improvements in conversions. This means that an accurate predictive model that depicts the causative or contributing relationships between the various in-play events is needed to determine the incremental changes in conversions based on various budget-related constraints -- not the aggregate weighted averages that are generated by attribution.
Attribution, on the other hand, serves a different purpose. It’s used to conduct deep analyses of past campaigns over time to figure out how to shift or allocate investments for future media planning purposes. As a result, a probabilistic model based on averages is most appropriate rather than a model looking to make incremental improvements. Let’s look at how attribution could be used in the example case mentioned above.
As the campaign wraps up, the same advertiser, using attribution analysis, identifies the parts of the campaign and creative that did a better job driving conversions at different stages of the purchase funnel. Using the average attribution metrics over the entire campaign, the advertiser is able to establish revised strategies and tactics for the next campaign, based on the underlying conversion insights that attribution provides. This allows the advertiser to build better campaigns that drive much higher aggregate results and conversions.
While both attribution and optimization are powerful and effective tools for helping online marketers measure, improve and predict the performance of their multichannel advertising campaigns, it’s important to understand what each is designed to do so that you can properly leverage them to improve campaigns. If you do, you’re well on your way to making sure you maximize your investments in online and offline media.