Marketing Land » Antony Chen

Why Display’s Moneyball Event Hasn’t Happened…Yet

Antony Chen — Mon, 27 Feb 2012 14:30:23 +0000

I recently went onto Quora to notice someone had asked the question that is the bane of my existence: “What are the barriers to widespread adoption of dynamic display ad platforms and formats?” Folks provided great answers and I’ve hinted to a few solutions in my previous columns, but it’s important to really attack those barriers given the results we’ve seen like:

Data-driven dynamic creative can increase return on advertising spend (ROAS) anywhere between 2 to 20 times
Almost all indications of user sentiment to personalized/relevant ads is very positive (minus a few incidents of annoying ads that won’t go away like here and here)

So, in summary, we’re delivering both better user experience and better performance for the advertiser. Anyone rational outside of our industry would ask, “So, everyone’s doing this, right?” Alas, this is not so.

In the best-selling book, Moneyball, Oakland A’s general manager Billy Beane was in a tough spot. The team had one of the lowest payrolls in Major League Baseball and was already destined to be near last place, so he made the controversial decision to apply a special form of statistics to valuing players. From 2000 to 2003, the A’s made it to the playoffs despite their anemic payroll. Today, Billy’s methodologies for valuing players are par for the course.

I draw a lot of parallels between Billy’s situation and the display ad landscape today. Billy was able to produce significantly better results with a disruptive technology or method. In Billy’s world, once everyone saw it work, everyone jumped on it to keep competitive. You’d think the same would hold true given the results we’ve seen with data-driven dynamic creative, so what gives?

Left Brain, Right Brain Problem

For dynamic creative to be designed properly, proper data utilization is key to drive results. Whether your creative is rules-based (e.g. “Show the closest store if the user is within 10 miles”) or dynamically optimized (e.g. “Explore and exploit best performing based on audience data in real-time”), this requires some technical knowledge of the data available and its application. This tends to be the realm of quantitative market researchers, marketing analysts, etc. — the “left brain”.

On the flip side, good creative agencies hire creative people. This is not to say that creative people can’t also be technical, but unless that graphics designer or front-end user experience designer has specifically received technical training, data-powered dynamic creative won’t make sense. They’re “right brain” people.

In essence, we’ve given creative agencies an incredible tool to drive the strategic direction of a campaign with little to no instruction on how that tool can be used. Most of us (myself included) in the position to educate and inform have done a pretty poor job of putting forth resources.

N.B. Most of the mythology around the right hemisphere of the brain as the creative side and left hemisphere being the analytical has been debunked, but it still serves as a useful analogy for my purposes.

The Broken Workflow

In the current workflow, the media agency designs the media plan. The media plan goes to the creative agency to design creative for the audiences selected on the media plan and so on. Unfortunately, designing good dynamic creative should be done while the media plan is constructed. Dynamic creative opens up possibilities that aren’t available otherwise.

As an example, if you’re a quick serve restaurant with thousands of locations and you want to run an ad that shows the closest location, is this possible without dynamic creative? Not really. No one in their right mind is going to design a media plan with thousands of lines. (In this example, localizing the ad should create a discussion around the granularity and quality of the available user geo data, but this points to my left brain right brain problem above where the creative agency needs to have the technical knowledge to address the issue properly.) The creative agency should be at the table at the same time while the media plan is discussed.

Dynamic Creative Setup is Fragmented and Slow

Several dynamic creative players have touted their self-service interfaces and ability to quickly launch campaigns. However, with speed, you usually lose flexibility. Even if the interface allows for flexibility, then you’re back to the speed problem. I’ve seen a campaign using non-dynamic creative assets take as little as two days to launch.

Today, launching a data-powered dynamic creative campaign in that short a period is not possible unless it’s reusing pre-built components or conforms to a template’s tight parameters. The good news is that dynamic creative companies are trying to solve this problem, right now, because it’s a clear issue to adoption. By being slow or inflexible, it automatically locks you out of a fair amount of available budget.

When we get past these three big issues, I believe we will see dynamic creative disrupt display at a massive scale. It’s a fact that consumers change preferences to brands and products through the influence of a variety of factors. In display, we have the ability to evolve the campaign with these changes, but it needs to happen at all layers of the campaign: targeting, bidding, creative, and measurement.

We’re already making great progress at disrupting the targeting and bidding world with real-time bidding, but are only getting started with the others. Dynamic creative is real-time creative. We’re on the brink. It took over 100 years of professional baseball for Billy Beane to come along. As an industry, we can work together for our disruptive event without waiting 100 years.

My Hope For 2012: Taking Data And Creative Seriously

Antony Chen — Mon, 30 Jan 2012 14:00:11 +0000

I went on the attack against the incorrect use of data in my previous column. Early in my career, a mentor taught me that when you tell people about problems, you should also suggest a potential solution. Oherwise, you’re just a complainer. That’s great advice, so let’s talk about how we could potentially fix these issues.

The ultimate promise of the billions invested in the ad ecosystem — and why we’re willing to put up with a value chain that looks like this – is to get the right message in front of the right person at the right time. I’ve said this mantra myself, and I’ve heard it from competitors and from partners. Yet we’re failing to deliver on this in an end-to-end fashion. Except for personalized retargeters who are focusing on just a sliver of the solution, we’re relatively nascent at getting this to work. Before we start a discussion on how to start fixing this, it’s a good idea to agree on what we’re actually trying to fix. Let’s examine what “right message, right person, right time” should mean.

Working backward, “right time” has two factors:

Right Time

Showing the ad when the person is in the right context (contextualization).
Showing the ad to the person when they’re temporally ready to hear your message.

Right Person

“Right person” seems obvious, but it’s not. While audience targeting is a beautiful thing, its complexity has absolutely boggled even the most sophisticated. At its highest level, this is matching the campaign goals to people who would be

Interested in the marketer’s product/service.
Are properly positioned in the sales cycle to be receptive.

Right Message

N.B. Many of you may be thinking that “showing the ad to a person when they’re temporally ready to hear your message” sounds the same as “are they properly positioned in the sales cycle to be receptive”. They’re similar, but not the same. An extreme example of someone who is temporally ready would be a pizza delivery company targeting people with promotions on delivery when it’s raining outside. Day-parting is an example of a feature that focuses on temporal readiness. Properly positioned in the sales cycle refers to the fact that you’ve developed brand awareness and a fair amount of brand engagement before showing your best offer for product trial, as an example.

And, last but not least, “right message” is about getting the most effective creative in front of this right person at this right time. This is ultimately about

Tying the right part of the brand story to who the user is.
Using intent data properly to drive the user to their most likely conversion path (whether it be showing exactly the right product that you know they’re perfect for or knowing they don’t buy online and therefore showing them the store two miles away where they can purchase).

While it seems that “right person” is tied directly to targeting, it is not. If you’re really looking to get this right, all of the six factors I’ve described above should lead to decisions made in targeting, creative, and measurement.

Why do we want to go through the hard work of tying this all together? Because if we do, we’ll have a real shot at reversing the inertia of spending in traditional marketing. The reason why most marketers spend only 10-15% of their budget in a medium that represent 40%+ of consumer time spent is simple: We haven’t given CMOs enough benefits to overcome the switching costs of shifting media budget and the downsides of display advertising. But there is one huge unrealized benefit: In display advertising, we can deliver audiences at the scale of above the line (ATL) campaigns, but with the effectiveness of below the line (BTL) campaigns.

Display can be the first to deliver on the promise of what agencies and advertisers have been calling integrated marketing, integrated communications, through-the-line (TTL), etc. It’s all marketing-speak for delivering broad reach, but having extremely personalized communications at the individual level. The first folks that started delivering part of this promise is are direct mailers (which, in large part, is why this industry is still growing despite the fact that the United States Postal Service is in the process of imploding).

Before we get into how we get all the way to the promise of truly integrated marketing — where we’ve integrated cross-channel and cross-campaign with integrated targeting and messaging — we need to start with the basics, which is integrating targeting and messaging within a campaign. This sounds extraordinarily basic, but it’s not. Conversations about what we do after targeting don’t seem to happen all that often.

First Steps

So, let’s take the first step to getting to a world where we can generate substantial benefits in display versus other channels. And, before we open back up the data can of worms I mentioned in my previous post, we should set up a process where we can use data more effectively once we get our grubby hands on it. Here’s what we do:

Stop the complacency around last click. Yes, I know you and your company have been tracking it for years and if you stop looking at it, you’ll have no year-over-year results any more, but if they were all directionally wrong to begin with, then shouldn’t that be OK? Plus, who is saying you can’t continue to use it while developing a more sophisticated attribution model? Until you start looking at measures that offer a more direct proxy for sales lift, all of this becomes moot.
Get your creative guy/gal in the room the first day you start talking about a campaign. While it’d be great to get into creative personalization/dynamic creative right off the bat, having that person in the room is the first real start to having a tightly integrated plan where media buying decisions coincide with creative decisions. And, have him/her start thinking about ad format right away. There’s a right time to use rich media, in-stream video, etc. This person should be on the hook for consulting on ad format.
Integrate brand and performance, or at least have projects where both sides are involved. In the new world, the budgets of these two worlds will be fungible. How do I know that? Because there are brand effects to performance campaigns and direct response effects to brand campaigns. Right now, we ignore those and we can’t do that forever. Plus, what is a brand campaign except a direct response campaign with a much longer time horizon? Brand favorability is just a synonym for likelihood to purchase over an extended period of time. I’m not saying we slam the budgets together tomorrow and deal with the chaos of it all, but at least having a few projects where both teams are involved will start the dialogue that’s necessary for these two teams to see how similar their worlds really are.

Maybe by getting these three things underway, we’ll have a process where — back to my original point about data — we can have an interesting discussion around what data we need, how much of it, where to use it, and when to use it. That’ll be the world when display advertising could show its true promise against traditional marketing channels.

A Look Back At 2011: We’re Wrong About Data

Antony Chen — Mon, 26 Dec 2011 14:00:31 +0000

As we toast the New Year, we shouldn’t pat ourselves on the back when it comes to making smart marketing decisions based on data in the display marketplace. In fact, I’ll go as far as to say that much of the industry has focused on unsustainable business models, some risking their brands for short-term payoffs. On top of that, we’re just dabbling. We’re not truly serious about data as a pillar of marketing and, where it’s being used, much of what’s being done is wrong.

In talking to direct marketers and agencies alike, almost all data is coming from one of two places:

the marketer’s own website or
a third-party data provider.

Site Retargeting: Essential But Not The Sole Solution

In the case of the marketer’s own website, we’re talking about site retargeting, or more recently, personalized retargeting. This is when pixels are dropped on the marketer’s e-commerce site that gather up clickstream behavior on a user. When that user abandons the marketer’s e-commerce site and goes elsewhere that’s offering up their impressions for bid, an ad is shown that is personally relevant (based on the products she was viewing). (Disclosure: I launched personalized retargeting at Yahoo! in November 2010, but I’ll do my best to present an unbiased POV here.)

First, let me say that if you run any site where there’s an online conversion path and you’re not doing personalized retargeting, do it. It is foundational to any e-commerce site’s marketing strategy. Even if you don’t have online conversion goals, personalized retargeting can be used for branding — it’s a powerful tool for marketing to people who’ve initiated interest in your brand. (By the way, most folks aren’t thinking about it this way, missing out in a great opportunity to re-engage with “brand intenders”, but that’s for another post.)

There are essentially three problems with personalized retargeting:

Marketers end up using it as a crutch because it performs so well, but it cannot be the sole strategy. I’ve seen it happen time and time again where marketers start shutting down publisher run-of-site lines only to watch their site traffic plummet. Personalized retargeting only works when there are folks coming to the site in the first place — what some of us affectionately term “feeding the beast”.However, we in product in the display ecosystem also need to acknowledge that while it’s much too easy to make money from retargeting, our advertising customers demand and deserve high performing products in the upper funnel as well. Personalized retargeting is boring. Marketers need more. (And, we need top stop the insanity of non-integrated personalized retargeting campaigns whose brand effects are completely unrealized.)
Today, many of the companies offering personalized retargeting as their core product have marketers paying based on an arcane business model popularized by advertising.com more than a decade ago. For the last two years, everyone has been saying how the ad networks are dead. Not so. Tons of retargeting players have cropped up offering end-to-end retargeting solutions.That’s great except for the fact that they’re gathering up the marketer’s valuable e-commerce data and buying remnant impressions at pennies — with the result being that marketers are buying back their own online data at several multiples above the impression’s original value. I’m not saying the service isn’t valuable. The issue is what real value is being delivered versus what’s being charged to marketers.
Ninety percent of all retail purchases are still done in-store, so why only focus on the online channel? Yes, I know the online sales channel is growing like crazy, but even in the future, some goods and services will still require an offline purchase path. High consideration items, for example, are likely to be purchased offline. Even for items that are eventually purchased online, they often had an offline component. Think about the iPhone. You’d never buy one without playing with one. This is why Apple was brilliant to put stores all over the place. They don’t care if you walk out of their store after playing with the phone, because they’re fairly certain you’ll go home, noodle on it a bit, and buy one at 2am while your spouse is sleeping.

The point here is simple: What is the online world doing to leverage the other 90% out there? There are a handful of products out there that are attempting to address this problem like Yahoo!’s Audience Match and TARGUSinfo’s AdAdvisor, but the much of the industry is just waking up to this point. Think about it: store retargeting. In this world, offline actions drive marketing activities to the online channel where 40% of our time is spent.

There’s a subtle point here that if high consideration items require some offline touchpoints, then it leads one to realize that low consideration items are the ones being sold online. While this is a gross generalization, it’s an important point that demonstrates why current uses of retargeting fall short as a marketing strategy. Retargeting companies are focused on your online audience, a segment that is possibly your least profitable customer segment.

Using Data Very Very Carefully

The biggest gating factor to using that offline data is that it’s somewhere between “really hard” and “impossible” for most companies to activate that data. At the highest level, the online advertising industry is currently under the legislative microscope of Congress, which has threatened to heavily regulate the industry, so the usage of Personally Identifiable Information (PII) is a touchy subject. Then, you need a large, addressable, verifiable pool of online profiles (cookies) to match with known offline data, so it’s no mistake that Yahoo! and TARGUSinfo are the leaders in this space. Retargeting companies who only see a finite amount of online activity, without full user profiles, can’t offer access to all your customers. Finally, the marketer needs to get their customer data to a safe place where it can be matched against the addressable cookie pool. In thinking about creating marketing programs using customer data, really tough questions come up like:

Why am I doing this? As a marketer, my customer data are my crown jewels. What are the goals for creating this online program; what will this mean for my relationship with these customers?
If I do this, do I strip out some data because it’s too sensitive, or am I undercutting my goals by not providing that data? As an example, shopping frequency data could be highly useful for optimizing marketing messaging, but it is highly sensitive.
What assurances do I have on data leakage? How can I be sure that my data segments, once running out in the online environment via DSPs, Exchanges or Publishers direct, won’t be sniffed out by other data modeling companies and compromised?
And, probably the most frightening: This means I have to talk to IT. They tell me my CRM data lives in ten places and every time I ask for it in one place, the IT guy tells me it’s nearly impossible. Do I really want to deal with this? (Note: My link to the Nick Burns character is by no means a disparagement of IT at companies. I used to be one of these guys, but this character is only funny because it’s a fairly accurate perception.)

When marketers develop answers to using their online and offline data to shape their relationships with their customers, then they will truly have the keys to the kingdom. By using both sets of data, they can grow customer relationships by having meaningful, personalized conversations. And, best of all, the results will not only generate short-term performance goals, but help build the marketer’s dream of brand equity.

(I would like to thank Natalie Kubitz from Cognitive Match for her contributions and edits.)