What You Need to Know About Real-Time Optimization

It wasn't that long ago that direct marketers were a breed apart: data-obsessed spreadsheet jockeys who were constantly tweaking the knobs and dials of campaigns to yield incrementally better results.

Now, of course, everybody's data-obsessed; everybody's busy tweaking the knobs and dials. If you're marketing digitally, you're a direct marketer -- period.

Think about some of the hot trends in digital marketing -- from behavioral retargeting to real-time bidding on ad exchanges -- and they're all about direct marketing to individual consumers. Even image-burnishing branding campaigns that don't have an e-commerce component (i.e., they're not specifically designed to prompt a consumer to click through and make a purchase) are deployed using cookie-based data to target, in real time, consumers as they surf the web.

What direct marketers have known for years is that more data means better targeting, and better targeting means better results. And the best results are all about optimization. Of course, "optimization" means different things to different people. That's the problem with industry buzzwords: they tend to get diluted and distorted. So let's start with a couple of conventional dictionary definitions:



1. Making the best of anything.

2. A mathematical technique for finding a maximum or minimum value of a function of several variables, subject to a set of constraints, as linear programming or systems analysis.

When most companies talk marketing optimization, they mean the former (basically, Let's give it our best shot!). But the "mathematical technique" approach to optimization isn't necessarily complicated either, conceptually speaking. For instance, in key marketing areas such as email- or website-optimization, the most common techniques used to optimize are A/B testing and multivariate testing.

With A/B testing, a marketer will deploy two versions of an email, or two versions of a web landing page, and watch how each performs. The difference in metrics -- e.g., open rates, click-through rates, conversion rates, etc. -- might be subtle, or they might be dramatic, but either way the goal is to pick a winner and then keep on testing with new A/B sets.

Multivariate testing simply expands the number of elements that can be monitored at once. Essentially, though, A/B and multivariate testing are sort of a general version of "making the best of anything." They're old-school, see-the-forest-not-the-trees approaches in that they look at consumers as relatively monolithic groups; the underlying characteristics of individual consumers are ignored. The goal may be to try to get as many trees as possible in the consumer forest to sway a particular way, but the focus isn't on any individual trees within that forest.

The traditional response to the underlying weaknesses in A/B and multivariate testing approaches has been to customize content based on individual data elements - e.g., location, age, sex, etc. - or by delivering different content versions based on statistically determined segments. But in no case has specific content been statistically and mathematically optimized for each specific consumer according to their relationship at that point in time (i.e, "in real time") with a brand.

Optimization only gets really interesting when individual consumers are regarded as, well, individuals. Consider, for instance, a 34-year-old working mother, an existing customer of a brand, who visits a brand's website. Cookie-based data can tell an incredibly rich story about her that can allow the brand marketer to, you might say, hyper-optimize. The idea is to "see" the website visitor as a specific consumer with a relevant past - meaning she has a transactional history with the brand consisting of all her past purchases, as well as a track record of reactions to marketing campaigns (e.g., emails opened, clicked, etc.).

Wouldn't it be cool if the next email sent by the brand contained references to the content she's just consumed, as well as suggestions about a next logical purchase based on past purchase history? Imagine if the email also contained messaging specific to her psychographic characteristics. And then suppose an offer included in the email had been tailored specifically to her needs and desires based on real-time analytics.

That's the potential of optimization. When optimized campaigns work, they can be incredibly powerful for brands, and heartening for the consumer, who no longer feels like just another face in the crowd.

5 comments about "What You Need to Know About Real-Time Optimization ".
Check to receive email when comments are posted.
  1. Lisa Fernow from Fernow Consulting, LLC, May 18, 2011 at 1:53 p.m.

    I think you raise an excellent point - the gold standard should be to know each customer (as well as they would like us to!) and serve them as they would like to be served on that particular occasion. Back in the late 70's, I worked at an insurance company which leveraged information from their database to direct market to their customers using personalized information for each customer (along the lines of "you told us you're over 65 and last time you used the hospital you had to pay $ out of pocket because you didn't have the right supplemental insurance, so we are recommending you buy this particular product"). We saw response rates of over 10% v. an industry average of 2%. Customers were willing to share their information with us because we were endorsed by organizations they trusted. It's exciting to see the possibilities we have for doing an even better job today, now that we have more data sources.

  2. John Grono from GAP Research, May 18, 2011 at 5:54 p.m.

    A conceptually strong piece Bob.

    But to follow your example of the 34-year olf working mother who visits the brand's website, and the "rich story" that cookie-based data tells about her. Let's also consider that she visited that brand's website from the family computer in the study. That is also the same computer that her 7-year old son, 5-year old daughter and 36-year old husband surf the internet on.

    How, when you rely on cookies, do you know who is who in this household? Just how sharp is cookie-based targeting? Optimising targeting based on cookies out in the real world is a very blunt instrument.

  3. Ken Mallon from Ken Mallon Advisory Services, May 18, 2011 at 9:37 p.m.

    Nice article. Agree with John Grono that cookie-based targeting isn't perfect but it's much closer to individual-based targeting than not using it. What you describe, Bob, is essentially available now through Yahoo! SmartAds.

  4. Kelsey Mulligan from Convert, May 19, 2011 at 9:26 a.m.

    Very interesting point, John. It is very important to know the consumer as well as possible, however it doesn't work out too smoothly all the time.

    Convert's apps can do these things right now. If you'd want to talk about it more, you can visit to find out about what Convert can do.

  5. Jonathan Betts from MediaCom, May 26, 2011 at 8:47 a.m.

    My wife recently received exactly the email described here from Amazon with recommendations based on her browsing habits on the site.

    Her reaction was one of horror and anger that the site owner was tracking and targeting her in this way.

    I think we should be aware that the reaction of people to this kind of targeting may not be what we initially expect.

Next story loading loading..