iMedia Connection

The metrics you should be using

Jarvis Mak

"You can't manage what you don't measure," is an old management adage. However, by the same token, just because you can measure something, does not mean you should manage by it.

The metrics you should be using

I authored a previous iMedia article about why clicks are the wrong metric. In that article, I alluded to the view-through effectiveness of display advertising despite the absence of clicks. In the spirit of that conversation, I want to discuss the relevance of other metrics in determining campaign success.

Use: Brand measures like lift in awareness or consideration

You can't click on a billboard, radio, or TV ad but we know they impact awareness, consideration, even purchase intent. Of course, the same is true of online ads. If the campaign has a branding objective, then optimize to it, ideally with a real-time data-driven approach.

Brand marketers want solutions that demonstrate improvement over time, transfer learnings across campaigns from top of the funnel to mid-funnel, and finally, result in sales. Everyone loves search for converting in-market folks but don't always remember to widen the funnel. As a marketer, if you rely on offline media (namely, TV) to be your lone awareness vehicle, then you're subject to over-saturation in television and missing out on cross-media optimized exposure. Complement offline with a digital awareness initiative.

Use: Ideal customer composition

Most advertisers know their customers inside and out, or at least they think they do: They have detailed customer profiles, even personas of Jim the Technophile or Susan the Savvy Bargain Shopper. However, when it comes to reaching those audiences in media delivery, the richness of the Jim and Susan personas are often lost to become coarse gender and age descriptions of "men 25 to 54" and "women 18 to 34," respectively.

Publishers that have personas that perfectly match advertisers' personas are one step ahead of the game. For advertisers, you trust that the targets those publishers claim to have are good proxies for your ideal customers. If there are no publisher persona targets, then we've seen agencies often default to the age and gender target as the next best proxy.

Why use a proxy though? Optimize segment composition in real-time. How? With a survey-based approach. We tried this with a couple CPG brands, with excellent results. The first advertiser wanted to reach previous brand customers. Utilizing real-time survey input to optimize segments, targeting of previous customers increased by 50 percent over the course of the campaign. The second CPG advertiser wanted to reach frequent users, and the same technique resulted in the target composition of this audience growing by 40 percent during the campaign. Results were validated with an offline sales impact study from Nielsen with a 250 percent return on ad spend.

Use: LTV (Lifetime value) of acquired customers

For direct response advertisers, especially in finance and retail, customers acquired cheaply are often low-value customers, especially when promotions are involved. The value of more expensive leads (as measured by account balances or subsequent basket size and purchase frequency) ends up much higher than cheaper direct response leads. An agency that is maniacally focused on low CPA without regard to LTV is doing their advertiser a disservice.

To advertisers, this means arming your agencies with visibility into these LTV measures. It means you need to do homework about customer behavior and transactions so that agencies can also pass information about customer value along to publisher partners, even anecdotally. We once executed a campaign for a bank but we didn't have access to back-end metrics like deposit amounts. As the campaign slowly auto-optimized toward the metric we had, CPA, the agency provided us feedback that we should go back to what we had been doing before day 45 because after that point, although CPA improved, the average starting balance amounts started shrinking considerably. With that bit of feedback, we were able to rejigger settings to pre-day 45 and go back to acquiring higher value customers.

Ignore: CTR

See the aforementioned iMedia article. We almost always provide data about how optimal segments are negatively correlated with CTR. For many agencies, they are glad to get this analysis to provide back to their advertisers as part of ongoing education.

For example, we demonstrated to a customer that pursuing segments that demonstrated higher CTR for the campaign would have led to incorrect optimization. Trying to maximize CTR would have led to a lower quality audience, since "clickier" segments were not the ones with the highest target composition.

Ignore: Ad interaction rate

It is not universally true that ad interaction rate doesn't matter. When the rich media unit has rich functionality like lead capture or submit, some other transactional feature, or a product catalog based on recently shopped items, then interaction is highly relevant. This is usually accomplished with dynamic creative and an XML feed.

However, when ad interactions are really opportunities for an expanded canvas that tell a more elaborate brand story, the benefit is dubious at best. At Rocket Fuel, we repeatedly see that the simpler ads are, the better. In fact, we have run static ads for a campaign alongside the long-looping Flash versions and find that performance is 50 to 300 percent better with simpler creative.

If advertisers want to understand impact of ad interactions, work with your publisher partners to instrument their pixels in ad unit expansion frames. Get ahead of the game and work with your creative partners before the assets are finished. Then the publishers can do the same type of correlation analysis with a brand metric, as well as optimize in-flight if there really is a positive correlation.  

Ignore: Video completion rate

This is in the context of in-banner video. It is a natural extension of the other two "unimportant" metrics. Does it matter that a video plays all the way through to completion? How much does auto-start inflate the metrics? It really doesn't matter what percentage of ads went all the way to the end unless it was all reaching the desired customers. As a marketer, I'd be happy with static ads shown to the right audience over a 75 percent video completion rate to the wrong audience.

A very notable exception to this is when video completions are well-documented to correlate directly with sales, like movie trailers for theatrical releases.


In general, we almost always see that metrics about online ad engagement or interaction (alone) not translate to bottom line results as effectively as lifetime value, ideal customer targets, and changes in brand measures. Ad metrics can definitely be a part of the overall story of the campaign; however, I wouldn't recommend them in isolation. It happens too often that an agency measures (and eliminates) partners along a single dimension like ad interaction rate.

Instead, the advertiser and agency should compare partners along a multitude of dimensions. Better yet, go the extra step and attribute value along all these dimensions. Here are some thoughts to get you started on attribution. It is important to review and compare partners but measure in ways that give you a comprehensive view of their overall campaign impact.

Jarvis Mak  is VP of analytics for Rocket Fuel.

On Twitter? Follow iMedia Connection at @iMediaTweet.

"Successful people are standing" image and "Social media, communication in" image via Shutterstock.