Death of the Digital Media Agency (Redux)

ImageLast year, I wrote that the digital agency was dead. I was mostly talking about how platform technology was going to knock a lot of digital media agencies out of business. In a world where over five trillion banner impressions are available every month, I argued, it was simply too much for humans to navigate through the choices and wring branding effect and performance out of campaigns. Well, digital media agencies are still around—but they continue to lose share to platforms as the amount of programmatically bought media increases. With RTB-based spending estimated to rise at an annualized rate of nearly 60% a year, according to market intelligence firm IDC, we could see as much as $14 billion in spending by 2016, or 27% of total display spending. Looks like the machines are slowly taking over.

Fairfax Cone, the founder of Foote, Cone, and Belding once famously remarked that the problem with the agency business was that “the inventory goes down the elevator at night.” That’s a big problem for an industry that relies on 23 year-old media planners to work long hours grinding on Excel spreadsheets and managing vendors to produce fairly mediocre media plans. Cone was talking about IP—what, exactly is the digital agency’s core intellectual property when the majority of the work seems to be hard labor? Digital creative agencies have no such worry. In this world of ubiquity, where everyone has access to wonderful SaaS-model technology that enables real-time bidding and access to trillions of exchange-based advertising impressions, the one place an agency can make an impact is on the creative side. Agencies that can create the miracle of getting more than 1 in 10,000 people to click on an ad, or watch a :30 pre roll video to completion are considered geniuses. But, what about the media shops? Can they really buy more efficiently than machines? More importantly, can they leverage the right machines to once again own the middle position between the advertiser and his prospect?

As I write in my recent report, looking at the history of display advertising, the future doesn’t favor the agency. In the beginning, agencies’ favored relationships with publishers made them a great way to buy media. Publishers aligned their content with the audiences that advertisers wanted (ESPN for sports enthusiasts), and largely controlled their inventory and audience data. Soon enough, the Network Era told hold, and smart companies like Tacoda started segmenting audiences based on context and behavior. By using technology to understand audiences better than the publishers themselves, they put yet another layer of IP between agencies and audiences. Then the DSP Era started, which further decoupled audiences from media. Agencies scrambled to create new vendor relationships with the MediaMaths of the world—but grew nervous that they would be disintermediated, and formed their own trading desks. This era is now evolving in the DMP Era.

After all of promises of easy audience targeting and automation, advertisers are looking at the same disturbingly low click-through rates, near impossibility of true attribution measurement, and spending waste—and determining that their own data is more valuable than most data that they can buy. Their desire to activate their “first party” data has given rise to the “DMP era.” Andy Monfried, who has brought his company Lotame through this transition, sees it this way, “Agencies are attempting to become technology providers for their clients, and from our perception, clients are hesitant to adopt. The larger agency holding companies have made an attempt at understanding first-party data but have come to be just a solution for clients to leverage third-party data. This is due to the lack of agency technology and lack of trust that clients put in agencies accessing their first-party data in a raw state.”

So, what happens now? Are advertisers simply going to license DMP technology, and build small practice groups for audience segmentation, targeting, and analytics? Or, are agencies going to adopt and learn how to become the centralization point for evaluating and helping clients implement new advertising technology? Media Kitchen digital head Darren Herman thinks the way through the trees is through education: “We are super bullish about teaching our strategists to learn the skills of data scientists. While the average media strategists will probably not have the skills of a robust data scientist with a PhD, from Stanford, an entire organization that learns to embrace data and make it useful will be more powerful than a few data scientists sprinkled [through] many. Knowledge of how to action data must both come from the top down and bottom up and be embraced by all. Building a culture that does this is hard as many people resist, but retooling and finding people who want this type of career is what we’re doing.”

Is your agency ready to hire a data scientist? Looks like the days of agencies hiring armies of English majors is over, and the next MIT recruiting session you see may have a few agency folks in attendance. Are digital media agencies dead? The data says not yet.

This post originally appeared on the EConsultancy blog on 12/13/12.

The RFP is Dead: New Concepts in Audience Discovery

The Programmatic Approach to Media Allocation is Coming Soon to a Platform near You.

Since its inception, advertising has always been about putting the right message in front of the right audience. Back when televisions were really expensive, and people used to gather around them in bars to watch baseball, beer companies started to do a lot of television advertising. While it’s still pretty easy for marketers to find the right beer demographic on sports programming in broadcast, the new world of multiple screens makes finding that audience at scale tougher every day.

The guy who was likely in the pub watching the game back in the 1940s and 1950s is now watching the game at home, but maybe on his iPad. Or perhaps he’s sneaking it in at work on his computer via Slingbox, or following along on his Android phone on the MLB Mobile app. The point is, there’s no easy way to find him, it’s almost impossible to find him at cheaply at scale, and we may have the wrong way of discovering him online.

The traditional method of finding your audience in the digital space is to put together a campaign request for proposal (RFP) that details the nature of your ad campaign, the audience you are looking for, where you want to find them, and the most you expect to pay to reach them. An agency’s trusted inventory suppliers receive and evaluate the RFP, and put together (hopefully) creative strategies that deliver a way to find that audience, and put the agency’s message in front of the user at the right time, in the right place. This approach makes complete sense. Except when it doesn’t.

Here are some ways in which the traditional, single RFP fails:

Multiple Pricing Methodologies: One of the problems in the traditional RFP process is that the agency is often limited to suggesting a single price range they are willing to pay for the media. For example, a typical RFP for a branding campaign looking for contextually relevant, above-the-fold inventory may suggest a price range to publishers of between $8-$12 CPM. This is fine if the proposal is only going to premium publishers with guaranteed inventory. But what if the advertiser is also interested in finding his audience on a cost-per-click basis? Knowing the historical performance of similar past campaigns, he might suggest a range of $1.50 -$3.50 per click. While the agency is comfortable buying using both methodologies (and certainly prefers the latter), the publisher is left wondering how to respond in a way that gives him the best overall price, and best revenue predictability. After evaluating the campaign, he may well decide that he will fare better on the CPC model, but in the absence of the granular past performance data of the demand side client, he will probably opt for the revenue visibility afforded by a CPM campaign.

Markets tend to work most best when both sides of the transaction have access to similar information. That leads to pricing efficiency, which in turn creates long-term sustainable performance results. Unfortunately, the traditional RFP process tends to strongly favor the demand side customer rather than the inventory purveyor. Add in the possibilities of buying on cost-per-lead (CPL) and cost-per-action (CPA), and you have a situation in which the demand side customer has the benefit of greater data visibility, and the supply side opportunity becomes purely speculative, leading to even more pronounced market inequities. These dynamics have largely occurred due to the seemingly unlimited supply of banner inventory (a supply side problem that will be debated in another article), but the fact remains that today’s standard agency RFP process falls far short of accounting for the multiple ways in which digital media can be bought and sold today.

Multiple Buying Methodologies: Along with a new multitude of pricing choices available to both sides, the emergence of real-time-bidding (RTB) makes the traditional RFP process even less relevant in for today’s progressive digital marketer. Say a marketer wants to reach “Upper income men in Connecticut that are in-market for a BMW 5-Series sedan.” That’s a pretty specific target, and I’ll bet that if a marketer could actually identify and find the several dozen guys in Darien, Stamford, and Greenwich that are looking for that specific make and model of car within the time period of the campaign, they might be willing to bid upwards of $500 CPM to reach him. Unfortunately, if you were to restrict the RFP variable to that exact target, you would end up serving a few hundred impressions, and probably fail to even spend $500 altogether. Naturally, the marketer is willing to bid a lot less find all men and women in Connecticut that are in market for a BMW; or just men in Connecticut in market for a car in general; or even just men in Connecticut, whether they need a car or not. Naturally, bids for each segment will vary widely, and can span from single to triple digits. Without a CPM-based pricing cap, it is not uncommon to see bids above $1,000 for certain impressions, although very few of them are won.

Well executed RTB campaigns have multiple segments that bid at different levels, and impressions are won at widely differing prices. While the marketer expects some visibility around what the effective CPM may be for such a campaign, RTB systems work best when agnostic to media cost, and should depend purely on the advertiser’s CPC or CPA goals. While a marketer can be very specific about his ultimate CPA, CPC, or CPL pricing cap, the traditional RFP does not address his tolerance for certain types of risk, his willingness to deploy a large percentage of media budget for data costs, and his willingness to forgo placement and context in exchange for reaching his ultimate demographic targets. This is just one of the reasons that agencies are having difficulty transitioning to the new world of demand side platforms in general.

New Discovery Mechanisms: Finding your audience by creating a well-crafted RFP and working with inventory suppliers to cobble together an effective buying program is still a great way to reach your ultimate goal, mostly because publishers know their audiences really well and have been able to offer new and creative ways to engage them on webpages (and, now, multiple screens). But what if the publisher isn’t really in control of his audience? What if the content an advertiser wants to be associated with migrates and changes constantly, based on user behavior and activity? I am talking, of course, about user generated content. Companies like Buzz Logic measure the “conversational density” around a topic and find where people are talking about, say, “organic food.” You can’t find that audience with a traditional RFP. The prevalence (or downright dominance) of social media outlets has created an explosion of UGC that is creating content almost faster than marketers can discover it. And that those new content areas are highly desirable to advertisers looking to engage consumers in contextually relevant activities. Those audiences are found via technology. How about finding people through the products they own (OwnerIQ) or even based on their occupation (Bizo)?

RTB and data make finding very granular audiences an intriguing option for marketers, but the traditional RFP process makes it hard to describe a marketers willingness to mix traditional, contextual audience buying (finding fantasy football fans on ESPN, for example) from some of the new audience discovery options (finding college students online based on their ownership of mini refrigerators, for example). Both are possible, and probably great to deploy over the course of a single campaign, but the traditional RFP process doesn’t really address this well.

Allocation: In my mind, the most important aspect missing from the traditional RFP process is that it doesn’t bring the demand and supply sides together effectively to suggest proper budget allocation for a campaign. If you have a $100,000 budget, and suggest $10,000 per publisher, every publisher is going to suggest $10,000 in media—regardless of whether or not they have it available. Moreover, you are going to alienate some publishers that may have larger minimums. The real problem is that the traditional RFP process doesn’t easily allow budget allocation across multiple media types (guaranteed display, real-time bidded display, mobile, video, search, and social) or take into account historical performance data. Essentially, the RFP makes a crude guess at budget allocation, with the marketer using his gut and some past performance data (“well, the $40,000 I spent with Pandora last time performed pretty well, so I’ll do that again”). Although the amount of choices today’s digital marketer has have expanded greatly, his form of communicating specific campaign needs is still an essay-length Word document or form-based technology with limited fields that do not capture the breadth of choices available.

So, what is the answer? New platform technologies are helping marketers expand the way they describe their campaign needs-and their willingness to deploy differing pricing and buying methodologies to reach their intended audience. Real time bidding systems are also giving end users hundreds of different levers to control the types of bids they are willing to make, based on the granularity of the audience, and performance of the inventory they purchased. In coming months, technology will not only expand a digital marketer’s ability to better describe his goals, but also use past performance data to suggest more effective media allocations in the beginning—and during—a campaign. Based on granular campaign attributes, knowledge of price points where certain real-time bids are won, and historical campaign performance, systems will be able to tell the marketer: “Allocate this percentage to SEM, this percentage to guaranteed display, and this much to real-time display” while suggesting the most effective bids to place. This three-dimensional discovery technique is where we are headed. While we are getting ready for its arrival, marketers should start thinking outside the traditional RFP box, and begin configuring new ways to ask inventory partners to find their desired audiences.

[This post originally appeared in eConsultancy on 8/19/11]

Epic FAIL

This is why agencies buy direct.

Much has been written about the notorious “logo vomit” map of famed internet banker Terence Kawaja. I reference his handy charts on my blog, and often his “Display LUMAscape” as a reference point for thinking about the digital display business, and what will happen to it. Many have tried to navigate through the various categories and dissect what may be “happening” in the space, which is a favorite pastime of company executives trying to raise money for many of the identified advertising technology outfits referenced within. Nobody ever really tries to explain the whole thing, though. It’s just too complicated, I guess. Allow me to try:

 “A few years ago, people started to figure out that you could use technology to target advertising to people on the Web. Ever since then, 250 companies have placed themselves in the middle of the transaction between the advertiser and the inventory, confusing everyone. Now, most of them are running out of money and will sell cheap, get acquired, or go out of business.”

Perhaps that oversimplifies things slightly, but the reality is that there are many companies in the space that are primed for one of those three scenarios. Unfortunately, most of them will sell for less than their investment, or go out of business. Here are the three big reasons we have gotten here:

It was a Bad Idea

The whole point of most of the companies on the Kawaja map is to help advertisers use data to find exactly the right audience at the right time, serve them the right ad, and maybe find something out about them that helps drive branding or sales. In the past, most advertisers used to do that contextually (putting ads for shoes in Vogue, for example) and it seemed to work pretty well. When that Internet thing came along, publishers could get something nearing their print CPMs for “site sponsorships” and premium banner advertising alongside good content. Sooner or later, however, publishers decided to put banners ads on all of their pages, creating the advertising largest inventory glut known to man. That created a big problem.

All of that banner space needed to be monetized somehow, and publishers were quickly discovering that it was hard to make money on the trillions of monthly advertising impressions they had created. But nobody wanted to buy $10 CPM banner ads on message board pages, and the “contact us” page. So, in order to “solve” this problem, exchanges popped up and allowed publishers to “monetize” this space by having various parties bid on the inventory. Things got even better when data companies came in, and were able to layer some demographic data atop those impressions, making audience buying possible for the first time. The venture money flowed, as smart young technologists created fast-moving software companies to help marketers exploit this trend as they sought a way to help reduce industry average CPMs from $20 to $2.

Mission accomplished! In the last 10 years, average CPMs have been drastically reduced, 100% of a publishers inventory is being “monetized” (often by 10 or more companies), and you can target an ad down to one’s shoe size.  So, what’s the problem? Hasn’t turning advertising from an art into a science worked?

The answer is: Yes, but not for all of the companies on that map. People visit three sites a day, and one of them is Facebook. If you want audience targeting, why not just find exactly what you want from a social network? They are the ones with the real audience data. They are also the ones with the audience scale, having about 5 times as many “profiles” as the next largest data company. The problem with all the companies trying to sell you audience targeting and ad technology is that it only works when you have audience scale (they don’t) and deep audience data (they don’t have that either).

Facebook, Google, and LinkedIn (and the next company that people are willing to share their private information with) are going to win the audience targeting game. When you are talking about audience buying at scale, social media IS digital media.

It’s Still about Art

If you believe that the average web user visits only two sites a day besides Facebook, then you better find them on those sites—and give them a really amazing experience with your banner ad. That thing should play video, games, talk to you, and almost pay you to look at it. Since only three out of every 10,000 people will click on it, you had better make sure the creative really tells a terrific story and gets your brand message across too.

That means standard sized banners that work with exchange-based buying are pretty much irrelevant, since they have a hard time doing any of the above. It also means that context has to accompany placement. It is not enough to reach a “35 year old woman in-market for shoes.” You have to reach her when she is on her favorite fashion site, or otherwise psychologically engaged in shoe consideration. The ad should be in a brand-safe environment that engenders trust—and compliments the creative in question. That sounds suspiciously like premium display advertising…the stuff that was being sold 10 years ago!

In a certain sense, we have almost come back full-circle to guaranteed, premium advertising. And that means an emphasis on the creative itself. If you look at the map, it’s clear that creative isn’t a part of the picture…but it might be the most important thing driving the future of the digital display advertising business.

It’s Confusing

Even if agencies and advertisers wanted to take advantage of a few of the of companies cluttering the “landscape,” they would need to log into and learn multiple systems. As a marketer looking to reach women, am I really going to log into Blue Kai and bid on demographic “stamps” from Nielsen, log into AppNexus and apply those to a real-time exchange buy, constantly log into my DART account to check ad pacing and performance, periodically log into my Aperture account to download audience data, and then log into my Advantage account every month to bill my clients? Maybe—but that’s exactly the reason why digital media agencies are making 3% margins lately. Most of these technologies are really great on their own, but string together too many of them and you start to get lost in the data, and are unable to react to it.

For digital marketing to be effective, a set of standards need to be created that enables systems to work together and share information. Basic B-school dogma teaches you that effectiveness starts to break down when a manager has more than 5 direct reports. If you believe that, then it’s not hard to imagine the effectiveness of a 22-year old media planner managing 5 logins on behalf of his agency.  It’s not just confusing, but impossible.

We have built an industry ripe for aggregation, and the Googles, Adobes, and IBMs of the world will not disappoint us! So, what companies will succeed in this ecosystem?

– Social Scalers: If you agree that all reach advertising targeting audiences will eventually be on social networks, then you should look to work with companies that are making social advertising scale effectively. Doing Facebook advertising is incredibly easy—but doing it right is hard. Doing it properly requires extreme multivariate creative optimization and, more importantly, knowing what to do with the mounds of truly actionable audience data that Facebook and other social networks will hand you. Companies like XA.net that are doing this are EPIC WIN.

 – Creative enablers: Since the conversation is coming back to the creative, how can technology help make great creative even better—and help advertisers understand how that creative is being engaged with?  The click is a dead metric to most seasoned advertisers, who are spending more time with branding measurement tools (Vizu) and creative ad analytics startups (Moat) that are well positioned to “science-ify” the truly important part of advertising: the creative itself. Companies doing that well are also going to be EPIC WIN.

 – Standard Bearers: With all of the logins out there, it is inevitable that one company is going to try and create the technology stack for next generation media buying that puts all the pieces together seamlessly. There are a number of companies trying to do this right now (full disclosure: I work for one of them), and I believe there will be a lot of advertisers and agencies relieved to log into a single platform, and be able to access all of their vendor relationships in one dashboard.  This will take some time, but the companies that enable standardization across technology providers will also WIN big.

[This post originally appeared 7/20/11 on eConsultancy blog]

The Agency of Procrustes

Is Your Media Shop the Right Fit for the Digital Age?

Nassim Taleb’s marvelous book of aphorisms is called The Bed of Procrustes, named after the myth of Procrustes, a cruel owner of a roadside estate between Athens and Eleusis in ancient Greece. According to Taleb,

He abducted travelers, provided them with a generous dinner, then invited them to spend the night in a rather special bed. He wanted the bed to fit the traveler to perfection. Those who were too tall had their legs chopped off with a sharp hatchet; those who were too short were stretched.

Taleb’s point is that we humans tend to “squeeze the world into crisp, commoditized ideas.” In short, we try and fit things we don’t understand into our particular worldview. But, what if the new things don’t fit?

As a digital media agency owner faced with keeping up with the times and (more importantly) earning margins from notoriously labor intensive digital campaigns, it is tempting to fall back on time-worn models. If you think about the tried and true “agency” model, it is exactly what the dictionary says it is: “a consensual fiduciary relationship in which one party acts on behalf of and under the control of another in dealing with third parties.” In other words, the client can do the work himself, but would rather stick to making widgets or selling plane tickets than have 300 different media and technology relationships to contend with.

The problem? That’s not enough anymore. What clients want—and an increasing number of them expect, is a different definition of “agency.” Maybe even a legal understanding of the term: the person or thing through which power is exerted or an end is achieved. Is your digital agency exerting true power on behalf of your clients, or are you just buying media? I believe that, in a world where technology enables most agencies to have ubiquitous access to media and software tools, the modern digital agency needs to go beyond traditional notions of “agency” and provide their clients with unique expertise.

The traditional agency “bed” is still rather misshapen for the world of emerging technology. Most shops still don’t have a cohesive social strategy (beyond Facebook); the technology to properly target audiences through exchanges; or the ability to leverage technology to wring performance from digital creative. Some do, and are leveraging relationships with social technology providers, DSPs, and creative optimization companies. The problem here is that many of those technology providers are going directly to your clients as well.  So, how do you defend against disintermediation and start building proprietary expertise to enable you to win and retain digital business in the future?

  • Data: Create it, analyze it, tie into your clients’ data, and make it actionable. I know an agency in upstate New York that only gets paid every time its client performs an oil change. The agency is tied into their client’s POS system, and gets a true end-to-end view of attribution. They know how they are getting people to the business, when, and how they are getting them to return. I know other agencies that, through tools like Datran’s Aperture, are getting a household-level view of who is converting on their online campaigns, and using online data to go offline to seek new customers and reengage them. If you are not leveraging the data you currently have—and seeking to partner with your client to create or get access to new streams of data, then you are not being an extension of power to your client.
  • Technology: How is your shop leveraging available technology to gain efficiency? Media platforms like Transis, Facilitate, and TRAFFIQ (disclosure: I work for TRAFFIQ) offer agencies the ability to let workflow technology handle the blocking and tackling of digital media (RFPs, AdOps, billing, etc) so agencies can work on things that have value (strategy, creative execution, data analysis).  What about real time bidding technology that uses machine learning to auto-optimize campaigns based on performance data? If you are not leveraging technologies like these, then you are already in danger of becoming extinct.
  • People: If you are in fact going to leverage data and technology to transform your agency business, then you are going to necessarily need different people. In the good old days, you could hire a 22-year old for $25,000 and bill them out at $40,000. Unfortunately, the 22 year old wants $35,000 these days, and by the time you train them to be a “digital media expert,” a larger shop will pay them $50,000 to take advantage of the free training you gave them, and start billing them out at $75,000. Also, that 22 year old media person who used to good at collating spreadsheets and ignoring publisher e-mails is not the person who is going to transform your business. Someone who can dive into data to determine media placements—or someone who is passionate about the social space and understands the new social technology ecosystem are the folks that are going to make a difference (and profit) for your agency now.

In the end, Procrustes faced poetic justice. One of his guests was the mighty Theseus, of Minotaur-slaying fame. Theseus invited Procrustes to lie in his own bed and, seeing it slightly too small for his frame, decapitated him to create the perfect fit. Your agency may not currently be the right fit for clients that need advanced digital agency help. The answer, however, is to make your bed fit your clients better, rather than shrink them down so they fit into your legacy paradigm.

[This article originally appeared in Adotas 3/16/11]