|
|
Real-time Campaign Optimization & Beyond
|
|
 |
|
Exclusive Interview with Hitendra Wadhwa, Founder, Paramark
|
 |
|
Very few current attempts at campaign optimization qualify as real-time. Most barely even qualify as optimization, at all. Riddled with slowness and computational inaccuracy, current optimization procedures appear dramatically out of synch with the fleetingly quick campaign cycles and infamous accuracy of measurement that characterize Internet Advertising.
The crux of the trouble is that campaign optimization is, as things now stand, mostly a manual endeavor, wherein some person - typically working on behalf of the advertiser - is responsible for manually sifting through the mounds of data that are produced by a campaign, manually analyzing it, and manually making mid-course campaign adjustments based on this analysis. By the time campaign adjustments are finally made, so much time has elapsed, and so much critical data has been either overlooked or misinterpreted, that the optimization process yields campaign performance boosts that are routinely minute.
Enter real-time optimization. The endeavor of real-time optimization, currently being pursued by a number of firms (Paramark, Advertising.com, and Poindexter Systems, among them), is to enable the effective harnessing of incoming campaign data, in true real-time. Whereas, with manual optimization, there are substantial limitations on how much and how quickly campaign data can be put to use to boost campaign performance, with real-time optimization - in theory, anyway - all incoming data is acted upon to boost campaign performance, immediately upon its availability.
Traditionally, optimization has played a relatively isolated role in Internet Advertising campaigns, and has been thought of mainly as a way of impacting campaign performance, on a campaign-by-campaign basis. While real-time optimization is able to reliably improve the performance of any single campaign by 10% to 250%, avant|marketer believes that the advent of real-time optimization is likely to have far wider impacts than merely this. As the following interview brings forth, real-time optimization has the potential to fundamentally alter the way in which ad targeting and personalization are implemented in the online environment.
In what follows, avant|marketer Editor, Ajay Segal engages Hitendra Wadhwa, Founder of leading real-time optimization technology firm, Paramark, makers of the PILOT real-time optimization platform, in a revealing and important discussion about the implications of real-time optimization for the future of Internet Advertising.
In the course of the discussion, in-depth consideration is given to the nature of real-time optimization; how real-time will supplant current manual optimization methods; the significant impacts real-time optimization is likely to have on ad targeting practices and campaign performance; the real-time optimization of Internet Branding campaigns, and the automated real-time optimization of the total media mix; and the "bigger picture" role of automation and real-time technology applications in the future landscape of Internet Advertising.
avant|marketer: What exactly is real-time optimization?
Hitendra Wadhwa: Real-time optimization is a particular process that allows interactive marketers to extract the best possible performance from direct response campaigns.
Basically, real-time optimization technologies are designed to be deployed to analyze the enormous piles of data that are collected through interactive channels during the course of a campaign, learn from these data piles, and translate that learning into action, by automatically adjusting interactive campaigns in real-time - as soon as those learnings become available to the system.
avant|marketer: Give us a synopsis of how campaign optimization is currently done. What are the specific areas in which the current optimization tactics are most lacking, and why do you believe the industry really needs real-time optimization?
Hitendra Wadhwa: Well, the most common way of doing optimization today is really not doing optimization at all. Certainly, in most campaigns, there is the appearance or feeling that optimization is being done: Lot's of data is being collected by advertisers, and via a whole variety of interactive channels - through banner, email, ecommerce servers, and so on. But, most of this data just sits there, and remains untapped.
The more diligent interactive marketers out there, who do engage in some actual optimization, might review campaign performance data periodically - typically, we're talking about these data reviews happening once every one to two weeks - and then, adjust their campaigns accordingly. This is essentially a manual optimization process, which typifies what goes on in the industry right now.
But, there are really a number of things significantly lacking in the manual optimization processes.
The first problem is that manual optimization is a very slow process, which means that, even if it does uncover important information, it does so after that information is useful, and, therefore, doesn't put the advertiser in a position to react to data swiftly enough to make any significant difference to campaign performance.
Manual procedures are also not done in a statistically rigorous fashion, and so, most often, doing optimization manually involves making misinformed judgments, based on overly simplistic rules of thumb.
I would say that the third critical problem is that manual optimization is not comprehensive in its analysis of the incoming campaign data. Therefore it's unable to identify and target different customer segments properly, even though a true, comprehensive data analysis would show that certain customer segments are statistically important, and could impact campaign performance significantly.
So real-time optimization, by basically automating the process of optimization and moving it to a truly real-time basis, in essence, is able to overcome these challenges, and allow advertisers to extract a much higher level of performance from Online Advertising.
avant|marketer: Now, real-time optimization inherently involves the testing of ad creative. Often we're talking about the testing of one set of creative against tens of others. Does this mean that the advertising creative itself will have to be developed in real-time? Poindexter Systems, in particular, has been much lauded for developing technology that enables advertisers to change the parameters of a set of creative, in real-time, by allowing the advertiser to pull a wide range of data into that set of creative from an external database. Is this sort of technology going to be required to do real time optimization effectively?
Hitendra Wadhwa: This is a very interesting issue.
If you recall, Mediaplex began doing just this sort of thing a while ago, with their so-called MOJO technology, which involved pulling data in real-time out of say an airlines tickets pricing database, or a Ticketmaster-type ticket database, into an ad to inform that ad of inventory availability, and that sort of thing. While there were certainly a few seductive stories about how this technology was used and could be applied, by-and-large, one has not seen advertisers leveraging this type of technology to very substantially improve performance. So, in my estimation, there are only a limited number of cases in which this pulling of data in real-time into an ad would benefit the performance of a campaign.
From a technological standpoint, though, there is no reason that this couldn't be supported for advertisers that want or need it...
avant|marketer: So this technology will play no significant role? In so far as the process of real-time optimization requires fresh creative to be continually inputted into the mix, it would seem that the real-time construction of ad creative ala Poindexter Systems or Mediaplex has a role to play - and can play that role inexpensively, at that.
Hitendra Wadhwa: I think what has to be looked at with regards to this type of technology is - very practically speaking - what situations there are where, in order for an advertiser to provide compelling creative that results in significant performance boosts, the advertiser needs to pull into that creative some real-time information, from some database.
If you look at it commonsensically, there are relatively few cases in which the advertiser needs to pull into the ad creative information that was not already available in some advanced form when that creative was originally developed.
Still, there could be a few scenarios in which this kind of data would be useful to the advertiser.
One situation that might be compelling is a scenario in which there is some excess product inventory that a company needs to get rid of, at the very last minute. Here they might want to be able to make sure their campaign is able to access an inventory database in real-time, and present remaining product items to prospects in an up-to-the-minute fashion.
But, generally, I think that - incrementally, for the consumer - the addition of real-time data to creative produces creative that has proved not to be anymore compelling than creative lacking this data.
avant|marketer: What then, is the relationship going to be between real-time optimization and ad creative? If real-time optimization is not going to necessitate new forms of ad creative altogether, one would at least expect that it would furnish advertisers with some insight regarding their existing ad creative?
Hitendra Wadhwa: From the advertiser's standpoint, one of the most powerful things about the Internet is its ability to allow them to generate insights as to what works and what doesn't work in a very structured, very rapid fashion. Jupiter[Media Metrix] used to call this "structured experimentation." I think that real-time optimization can actually be used to significantly increase the quality and pace of advertisers' structured experimentation.
As you're alluding to, part of what real-time optimization offers goes well beyond just allowing marketers to get a boost in campaign performance, on a campaign-by-campaign basis. It goes all the way to gathering ongoing insights over the course of multiple campaigns, about what aspects of an advertiser's campaigns are truly driving performace. And, many of these insights are important qualitative insights regarding the ad creative.
The insights regarding ad creative from real-time optimization come at a variety of levels of analysis.
At a high-level of analysis, the advertiser might discover which general elements of their creative have impacts on their campaigns' performance. It might be determined, for instance, that the color or template style that an advertiser is using doesn't matter to the performance of their campaigns, but, by contrast, the type of message they were using within their advertising, was driving performance in a very dramatic and significant way.
At a lower level of analysis, a specific retailer, say, might determine that discount-type messages perform very poorly for them, while humorous advertising messages perform extremely well for them.
With this sort of information, an advertiser is able to home in on what kind of creative it should be focused on developing and using. The advertiser is basically able to say to itself , "people are responding very favorably to funny messages of such and such a color and style, so we should focus on developing more of these."
avant|marketer: These qualitative insights on ad creative, are these company-specific, category-specific, or what? How widely do you believe these insights can be generalized? In other words, will it soon be possible based on this data for advertisers and media planners to say, "we're advertising a financial services product, so here's how we need to design our ad creative"?
Hitendra Wadhwa: Certainly, not all of the qualitative insights we're talking about will be able to be extrapolated liberally from one context to another.
But, I do anticipate that there will be some clear trends in the data that will allow us to draw conclusions about what kinds of creative work, at least at a category-by-category level. So the scenario you're referring to with financial services will eventually be possible.
avant|marketer: If that's the case, then is the availability of this aggregated data on what works in certain categories going impact the way in which ads are targeted online? What do you foresee happening with regards to the impacts of this data on online ad targeting?
Hitendra Wadhwa: Right now in the industry we have the components of a campaign being described in terms that are not very actionable. Media planners are forced to talk about campaigns in terms of "banners A, B , and C" and "[advertising] channels A, B, and C."
What is going to happen over the next couple of years, I believe, is that with the availability of real-time optimization systems and the qualitative insights that they bring to advertisers, advertisers are going to start to be able to talk about the attributes of a campaign in a much more actionable, much more intuitive way.
Instead of talking about banners A, B, and C, what's going to happen is that - using real-time optimization technologies [as a go between] - advertisers and media planners are going to start to be able speak to banner ad servers, commerce servers, email servers, etc. in the language of the actual, common sense attributes of the products being advertised, and the common sense attributes of the creative that is going to be the advertising vehicle for those products.
The trouble is that, right now, data is not even being stored by these systems - these various servers - in a way that would allow this to be possible. The first step in getting us there is getting the data stored in these actionable terms, and this is one of the things that Paramark is seeking to enable.
Ultimately - after this is done - media buyers will be able to slice and dice data in terms of red-colored banners, blue-colored banners, orange-colored banners, rather than in terms of banners A, B, and C, and so on.
avant|marketer: So, the flow of data in and out of ad serving systems will be done in common sense terms? Basically, advertisers will be able to say to a delivery system, "I'm uploading a set blue banners, and a set of red banners," and then, after the campaign ends, the system will provide figures to the effect that the blue banners performed better than the red banners, or vice versa. Is that the critical leap forward, here?
Hitendra Wadhwa: That's only part of it.
The critical leap forward comes in how this will alter the way in which ad targeting is done.
Once data begins to flow in these terms, advertisers will be able to say to the ad server, through the optimization system, that they are putting in a set of creative with certain attributes - color attributes, text-style attributes (Is the ad a plain text ad or not?) - and not only will the ad server be able to recognize the creative, it will be able to apply the entire aggregated history of that very type of creative - all of those generalized learnings, we've been talking about - to the targeting of that specific creative set.
avant|marketer: So, in effect, real-time optimization processes are going to involve two distinct layers of optimization, one which involves the application of campaign-specific rapid feedback data, and the other which involves the application of aggregated historical data, which will be used to assist in targeting ads from the get-go?
Hitendra Wadhwa: Right. That's absolutely right.
If you really think about it, all optimization efforts ultimately seek to boost performance.
So optimization is ideally done by blending whatever relevant historical data is available, with the rapid feedback provided by the various interactive channels that a campaign is being delivered through.
The more relevant the historical data that is available to the advertiser from the get-go, the less investment the advertiser has to make in the rapid feedback, real-time portion of the optimization process. So, the build-up of this historical data is likely, over time, to bring the costs of campaign optimization down for advertisers, and is also likely to considerably lessen the amount of time it takes for a campaign to reach its optimal performance level.
avant|marketer: What's emerging from what you've said is a picture of ad targeting that is vastly different from the one we have today. Right now, ad targeting is often achieved based on the advertiser leveraging a whole host of disparate pieces of often prohibitively costly data: Data contained on cookies, opted-in user data (typically brought in from the publisher at a substantial premium), and so on.
With the introduction of real-time optimization into the equation, does all of this data become obsolete, from a targeting perspective? And, does real-time optimization improve the cost efficiency of ad targeting then, by eliminating the advertiser's need for using all this costly data?
Hitendra Wadhwa: I think it's fair to say that ad targeting has really, so far, been an unfulfilled dream.
There's been a lot of promise and, to some extent, hype about the one-to-one personalization, targeting, etc., etc. capabilities of our medium. Certainly, a few companies have tried to do some very ambitious things in the ad targeting arena. But, overall, these efforts have received a very lukewarm reception from advertisers and publishers. And the reason for this, I think, is mostly that the lifts generated from these efforts have been too minimal to justify the costs of involvement from either side - publishers or advertisers. So, thus far, ad targeting has failed to become institutionalized as a serious and powerful practice.
What I think real-time optimization will do to move ad targeting forward is allow the advertiser to lay out a road map for the progressive targeting of a campaign that starts from a very basic level, in a way that doesn't involve a whole lot of costly, complicated data going into the equation, but just involves a few simple data-driven parameters like time of day and geographical location. Used in conjunction with rapid feedback real-time optimization algorithms and the historial data, this very basic data set alone will allow the advertiser to reap significant boosts in performance - so there will be no longer an initial requirement that the advertiser bring masses of outside targeting data to the table - opt-in data, and so on - in order to do ad targeting, and get results from it.
The other thing that real-time optimization will do to with respect to all of this targeting data is to allow it to be used much more efficiently. After the initial phase of real-time optimization is complete, and the initial performance boost has happened, real-time optimization systems will allow the advertiser to start to incorporate other data streams, like opt-in data, where this is available, and to figure out, just like with the creative, precisely which data streams are leading to a significant boost in campaign performance and which aren't, and eliminate the use of the data streams that are not adding to boosts in performance.
So, I would not as much claim that real-time optimization makes the traditional data obsolete, as that it allows the advertiser to incorporate this data at a later stage, allowing the advertiser to then even further improve targeting. Where such additional, more robust data is available - opt-in data, and all of that - real-time optimization will, essentially, make it's use much more efficient, which will, in turn, favorably impact cost, because the advertiser is not paying to incorporate data streams that the optimization process demonstrates are not impacting campaign performance.
avant|marketer: So far, most discussion we've had has focussed on the separate optimization of individual Internet marketing channels - Email, Banners, what have you. The Holy Grail here would seem to be the simultaneous real-time optimization of all the channels that comprise a given Internet media mix. Are we ever going to be seeing total media mix optimization, as such, in which the metric optimized for is overall campaign ROI?
Hitendra Wadhwa: It's important in looking at this question that we differentiate between two different types of media, if you will. One of these involves the marketers interaction with customers in channels and on media that the marketer owns - house file email lists, their own web site and in-house banner inventory, and so on. Here we're mostly talking about retention channels. The other involves interactions with customers through media that the advertiser has bought on partner sites, third party email lists, etc.
In the first case, where you're talking about the advertiser's own real estate, we see the advertiser's ability through real-time optimization to simultaneously optimize their campaigns across all of their internal customer touch points, if you will, as something that will soon be a reality, and something that is coming very quickly.
In the second case, when you're talking about bought third-party media, where the situation is that I want to establish a media mix that involves essentially buying customers by paying money to third-party media properties, whether those properties include a Yahoo!, a New York Times Digital, or an eUniverse (for email delivery), and try to do an integrated optimization across those properties, I see it as being less practical for real-time optimization to happen at the advertiser level.
I think in these cases, it is going to prove far more practical and powerful that real-time optimization happen from within the media outlets themselves. This is something that I think will be an attractive and valuable option for particularly the media outlets that have a lot of scale, a lot of advertisers, and a lot of variation in the types media channels that they offer - a New York Times Digital, for instance.
In these cases, you can imagine then, that the real-time optimization tools are likely to influence how the media outlets actually package their media for their advertising customers, and will allow large media outlets to more significantly drive financial value for their advertisers, by enabling the advertiser to start with a media mix that is spread out broadly across the various types of media that they offer, but then, through rapid feedback real-time optimization, allow the advertiser to home in on the sweet spot of media that is providing the highest ROI in their specific case. So what happens is that media outlets begin to segment their media properties not as they do now, in terms of say, customer profiles or demographics, but rather in terms of the media itself.
avant|marketer: Paramark claims PILOT can real-time optimize for any metric. However, at the outset, you were quick to link real-time optimization to direct response. Yet, firms such as Dynamic Logic are taking big strides toward make the branding impact of Internet Ad campaigns measurable, in an automated fashion. Are we going to see real-time optimization systems then, begin to optimize campaigns based on the real-time branding feedback delivered by say, Dynamic Logic's AdIndex system?
Hitendra Wadhwa: We have actually already been approached by a worldwide brand in the consumer packaged goods arena that has asked Paramark to develop a branding optimization solution along the lines of what you're alluding to. When the solution is done, the vision is that the advertiser will be able to identify what is the optimal story-boarding strategy in terms of what sequence and what frequency which ads should be shown to which segments of customers, in order to maximize the impact on the specific branding metrics that are important to the advertiser.
But, to do this is a big leap from the direct response scenarios that we have been discussing. Because, when it comes to Branding, what one is trying to do is to gradually shift the psychological state of a consumer through multiple advertising impressions, over a very extended period of time. This contrasts significantly with the impulse-oriented nature of direct response marketing, where what we look at often has to do with cause and effect connections happening over the matter of a few seconds. So to do branding optimization in real-time does require some changes in the approach one takes to real-time optimization. And these are changes we're currently conducting research into.
avant|marketer: So we will soon see real-time branding optimization algorithms be made to work with the data streams that are coming off of say, Dynamic Logic's technology, in a standardized fashion?
Hitendra Wadhwa: I think we are going to see something like this happen. But, it's not going to happen until there is some significant level of agreement that advertisers come to over which branding measurement providers are using measurement methodologies and metrics that are useful and correct. Until you have you have this agreement, you really can't have accountable brand advertising happening in the online world, in the first place. And any real-time optimization process, of course, is only as good as the data you feed into it. So branding optimization is a really a second step, that will come after all of this has been ironed-out.
Dynamic Logic has certainly been working hard to innovate and work towards the standardization of certain metrics. And more and more important companies are gravitating toward, and getting comfortable with their metrics. The more we see companies like Dynamic Logic being able to evaluate online branding investments on the basis of these metrics, and put these investments on an equal footing with offline branding initiatives - which is key - the closer we are to bringing in real-time branding optimization, and doing some very powerful things here.
avant|marketer: Right now, media planners and buyers expend significant efforts manually managing campaign optimization efforts. If automated real-time optimization supplants manual efforts at campaign optimization, what becomes the job of media planners and buyers? How will real-time optimization change Internet media buying and planning?
Hitendra Wadhwa: Well, obviously, real-time optimization, if it takes root across the industry, will be a significant driver in shifting online media buyers and planners away from the drudgery of manual analysis.
How will it more broadly change their role?
Naturally, media buyers will still play the important role of defining the advertiser's overall media strategy, based on the advertiser's objectives, identifying the right kinds of media buys to do for the advertiser, and negotiating those buys on the advertiser's behalf.
That said, I think the broader, more fundamental change we'll see in the roles of media buyers, brought about by real-time optimization, is that it will allow the buyers and sellers of interactive media to act in a much more collaborative way, because both sides, through the use of this technology, will be able to seek and extract the best performance from campaigns - in terms of say, converting impressions into actions, on the back end - in a much more structured, transparent way.
avant|marketer: Finally, think beyond real-time optimization alone: What do you believe the next two to three years in the Internet Advertising space are going to bring, particularly from the standpoint of automation and technology - which are two of Paramark's main themes?
Hitendra Wadhwa: Well, first I expect that there will be very dramatic steps forward in the area of Rich Media, leading ultimately to a new paradigms in how we engage with consumers on the Internet.
I'm particularly very fascinated by the distinction that we typically make between an active consumer (we think of consumers on the Internet as being in an "active" mode), and a passive consumer (which the way in which usually think of consumers when they are viewing TV), and how, as you add more and more Rich Media to the Internet, the consumer's state of engagement can be made to vacillate back-and-forth between an active and a passive state. I think, ultimately, this ability to make a consumers state vacillate between the active and passive modes will mean new rules for how we market to consumers on the Internet. And, I think this is going to be an area in which we'll see huge innovation over the next two or three years.
On the automation side, I believe we're going to see a rapidly increasing integration of customer profiling across multiple online and offline channels. In part, this is going to be driven by the eCRM initiatives that Fortune 500 companies are already engaging in, and, in part, by the renewed realization, which happened over the last year or so, that, ultimately, online is part of a larger mix of channels that should be viewed in unity with the others, in many ways. What's going to happen is this integration of profiling and targeting across these multiple channels is going to be actually supported by optimization, to allow marketers to leverage their full knowledge of the customers they're marketing to across all marketing touch points, in as close to real-time as possible.
|
|
|
|
