.com deconstructed

http://marketingland.com/a-look-back-at-2011-were-wrong-about-data-2106

A video of me at ad:tech was recently featured on the Yahoo Advertising blog. The full blog post is here or I've embedded the video below for your viewing pleasure.

Wow. It's tough to watch yourself in a video. At least I kept the verbal pauses to a minimum, but to watch myself blink like I have a wooden shard stuck in my eye, now that's just painful.

Beyond my self-deprecation, it does resonate that the idea of horizontal segmentation should also apply to the creative world. It just wasn't achievable at scale in channels like print and TV.

I want to thank the progenitor of Malcolm Gladwell and Smart Ads idea to Pete Kim, the original Smart Ads product manager, who told me the story in the first place.

He goes and describes the act of collecting data for the purposes of advertising as "sinister". First of all, this assumes that users want crappy ads for products they aren't interested in since no data on the user interests equals no targeting. Second of all, no targeting means that advertising revenues go down or even more crappy ads in an attempt to keep revenues up. Either way, the user experience goes in the toilet. This will be less money to pay journalists for these "free" content services. Sinister? Hardly. As someone in the industry, it'd be so much easier if everyone just paid directly for the content they consumed, but everyone wants it for free and most will give up a little data to keep content freely available.

On another note, I believe this journalist is being overly simplistic about why Google felt it needed to compete with Facebook and creating Google+. He said it's because of the growing time spent on Facebook and that means less money for Google who makes the majority of their revenues in advertising. Ok. Where's the insight that social is replacing search in the content discovery game? More time is being spent online in general so there's plenty of advertising revenues to go around. The real issue is that social is a direct attack at the core utility of search: finding content.

Still, Terri Gross is one of the best interviewers on the planet.

Podcast/Article:
NPR: 11-03-2011 Fresh Air

URL:
http://podcastdownload.npr.org/anon.npr-podcasts/podcast/13/142006579/npr_142...

Description:
Stories: 1) The War Between Google, Amazon, Facebook, Apple 2) Kelly Clarkson's Vocals Keep Getting 'Stronger' 3) A Critic To Remember: Pauline Kael At The 'Movies'

This content comes from:
NPR Programs: Fresh Air Podcast

URL:
http://www.npr.org/rss/podcast.php?id=13

I don't think the POV on display ads (or advertising) gets more cynical than this. It's Demetri Martin's drawing of the day. Enjoy:

Demetri Martin on WhoSay

Having recently graduated from Ross School of Business in December 2009, it's still fresh in my mind what it was like to be an MBA student. Having now informally and formally interviewed fresh grads, I was asked to put together a 5 minute podcast, but I figured that it'd be less work to type it all out and is useful information for anyone doing an informational interview, although this advice is geared toward an MBA student calling an alumnus.

The 5 Dos

Do have lots of research on the target company, target functional area, and the person you're speaking with. I'm not talking about doing a full financial analysis on the company. You should go in knowing some recent news, basic product portfolio, current market challenges, etc. It's all data you can get pulled from any company report that is provided through a b-school's career development office. Furthermore, you should know something about what functional area you're interested in. Focus is key. The more detailed you can be, the more someone can help. If there's one thing you take away from this, that'd be it. Finally, use LinkedIn and look up the person you're about to talk to. Figure out whether they're really someone who is going to be able to help you out.

Do ask actionable questions. What do I mean by that? Ask questions that result in an actionable result from either you or the alumnus. Again, it's only in our best interest to help you and see our school's footprint at the company expand. However, if you ask softball questions that don't leave us with a set of actions we can take on your behalf, you're doing yourself a disservice and wasting the alumnus' time. Examples of this would be: "What's the process you used to get your position?" or "I'm interested in X. My research is that there are some openings based on my interests. Are there any key steps you would recommend at positioning myself for success for those openings?" The goal is that the alumnus hangs up with you with a clear set of action items to help you.

Do pointedly ask for help. So, this one is tricky as a lot of recruiters will tell you that during an informational interview, you shouldn't be too pushy to ask for a bunch of stuff, and they're right, but you should ask for something based on how strong the relationship is. For example, on an initial call, asking for the person to vouch for you is unlikely a good move. However, if the conversation goes well, it might be appropriate to ask for an introduction to a hiring manager they know well, which brings me to another do.

Do use an example or real job that you're interested in. This makes life really easy for the alumnus because now we have something to work with. "So, I saw posting X on your careers website and it says that you need 2-3 years of product management experience. Is there a way someone with no product experience but a great MBA position themselves for a position like that?" It makes it much easier because now you're working off a template. I had one person call me up who used a posting as a template that happened to be my posting!

Do follow back up (without being annoying). We're busy out in the real world, so when you don't hear back from someone who you've made a connection with, don't take it personally. The best approach is to follow back up with a nice friendly reminder email.

The 5 Don'ts

Don't ask questions you can answer on your own. I was asked questions that someone could've just looked up the answer. This is silly because it makes you look you didn't do any research (read: lack of interest/waste of time). It also calls into question your competence. If you're making up questions just to create an excuse to call that person, then don't call them.

Don't pester people. If you've sent 2-3 emails and you haven't heard a response, then you're being ignored. Respect that. If you get overly aggressive, you end up turning someone who was neutral about you into a negative. This is important. It's doubtful an alumnus would ever take it upon themselves to proactively tell hiring managers and recruiters that you're annoying, but if asked, they'll be inclined to tell the truth.

Don't assume you've got a second chance, period. Every interaction with a target company should be seen as an interview. The career development people let you know that's especially true when you're interviewing with consulting companies (e.g. the cocktail mixers they hold are ways to informally interview you without the chairs), but guess what? This is true in general. This is highly related to my point about being highly prepped in my first do.

Don't waste people's time. You'll know in the first 5 minutes if the conversation is going well. If it's not going well, do yourself and the alumnus a favor by politely excusing yourself from the call. They'll respect you more and it'll minimize the damage you've done. Not all calls go well, so it's OK, but get out while you can.

Don't be a know-it-all. I've guilty of this one. I read one article about some topic and I think I can outsmart an expert in the field. Well, that's just plain dumb. Don't make this mistake. It's counterproductive to tell an alumnus that you know better than they do. This is silly because then a) why did you ask the question in the first place? and b) no one wants to work with someone they can't stand.

I know, I know. I promised I'd write a three part series on attribution and I will! I'm way overdue and Part 2 is going to be a doozie. I'll get to it, but mind you, I'm in the middle of having a second child, just moved, and it's not a simple topic.

So, let me take a break from that to write a quick post on something that's easy to write about, my day job. I'm at Yahoo! and currently manage the Smart Ads product. For those of you too lazy to click on the link to the product site, it's a dynamic creative product with one gigantic differentiator: we power it with our own data (and we have a lot of it thanks to running the world's largest portal for 15 years).

In our announcement last Wednesday of Smart Ads for Homepage and Smart Ads for Video, there's been fantastic press around it all and I'd like to share some of my favorite quotes, but provide some commentary on why we launched these products to be a more interesting topic rather than it's what advertiesrs want and blah blah blah.

When we were asked how we balance the advertiser's creative needs with the consumer's creative needs on AdExchanger, Dev responded:

We have editorial control over the page, which means these ads don't pop off willy-nilly. They don't control your content consumption experience when you don't want them to. You can do expand on a click or on a hover. There are guidelines that allow us to ensure we don't bastardize the user experience

I wish John had let Dev talk more about this because he didn't really get a chance at discussing the real underlying strategy here. The entire point is making the ad as relevant as possible and that includes the format itself. Therefore, if we do it right, the ad itself will become a part of the overall user experience and you'll no longer be thinking about tradeoffs between the advertiser and the consumer. They will become one in the same and ads become part of the overall content experience.

(For those of you who are cynical about advertising, this isn't about tricking you into thinking that the ad is part of the content. In fact, we go out of our way to make sure you know it's an ad through the AdChoices marker and a really obvious black border around the ad itself.)

On Digiday Daily, Brian Morrissey writes:

Look no further than its bet it can become a leader in the growing market for dynamically generated ad creative. For now, the Web has obsessed -- and poured billions in venture capital and tech resources -- into building a data-based system for delivering ads to audiences efficiently. Great, but what about the ads themselves? Yahoo’s answer: SmartAds, a project it began back in 2007.

1. We're already the leader.

2. Yes. What about the ads themselves? You spend all this time rifling through billions of impressions a day to pick those you really think are the right person (demo/geo targeting), in the right mindset (contextual), and in-market (behavioral/search). Then, you serve them some boring generic ad after going through all that trouble? In all seriousness, studies have shown that 50% of an ad's effectiveness is the creative. If you believe that, then not taking a serious look at any creative technology that purports to deliver awesome results is just plain crazy.

I'm excited about both products as they represent a monumental leap forward in capability. Whether it's the massive scale of the homepage with precision messaging with Smart Ads, or taking an already engaging ad format of video to the next level, it's all about:

• Delivering exceptional lift in ROAS for advertisers
• Executing campaigns that weren't even possible a few years ago
• In-depth campaign insights that reveal surprising and useful take-aways that can be used to power your next buy (even your offline buy)
• And finally, which tends to be forgotten - makes the overall user experience of the web that much more pleasant. Getting ads that actually might pertain to what you care about. Banner blindness - be gone.

Thanks to Pete Kim for reading through this. You can find him at his blog where he hasn't said anything in a while due to the fact he's working 90+ hours a week. His old posts are still very relevant so go visit if you want to read about display optimization or his odd fascination with the United States Postal Service.

I apologize to anyone who was actually looking forward to reading this. After posting that I was going on a multi-part diatribe on attribution, I moved from Silicon Valley to New York. Now that I'm (partially) settled, we return to our regularly scheduled program.

No one in their right mind agrees that last click attribution is the way to go, yet 99% of the advertisers with which I speak use last click. Worst of all, when you ask why they don't use something else, they simply say, "Well, you got a better idea?"

Current Incentive Structures

I don't blame them. The real blight on the war on attribution is the industry's incentives. Search engine marketers live and die by last click. Do search guys really have any reason at all to help you move away from last click when that's what they get paid on? Let's think for a moment what happens if it were discovered that the majority of the value of every "touch" you made with your potential customer didn't come from search, even if that was how that customer ultimately converted. Right. The search guys want you keep thinking the click is almighty.

Now, you've got Dotomi, Criteo, Fetchback, Buysight (fka Permuto), TellApart, on and on and on. All of these companies are doing a form of personalized retargeting, probably display's closest performer to search. Why am I bringing these guys up? They too have no reason for you to move away from last click. All of them get paid on the sheer fact that a site retargeting campaign is likely the last impression you'll see of an advertier's before you click and convert.

All in all, lots of people who have a vested interest in your never moving away from last click attribution. Oh, and let's not forget the strongest incentive of all: laziness. That's right. I might be talking to you. I mean, your site gets a certain number of click-throughs and it just seems so easy to distribute those clicks among all the campaigns you're running. It's simple, easy to implement, and oh so wrong. We'll get into the wrong part in the next post, but c'mon. Can anyone say with good conscience that the one impression that drove the click is 100% responsible for driving the conversion?

Why We Need To Fix This

Why is tackling it so tough? I'm going to do something that no one else does in these damn blog posts. I'm going to talk about the guts of the math involved.

Let's start with the current practices and methods. Mind you, everytime I've asked advertisers and agencies that they're looking beyond the click, it's rarely a solid answer, but again, it's hard to blame them. As I mentioned in my previous post, if you could do attribution with a high-level of confidence, the game would be over and that person would be rich. However, rather than the conversation be in the private corridors of agencies and their advertising customers, I think we need to be public about the current methods, the flaws, and potential ways to solve them.

At the end of the day, it's costing all of us in the display marketplace money, mainly through advertisers fearful of shifting dollars from traditional advertising where these issues don't exist due to mature marketing mix modeling techniques. One clear example of money out of the pocket is when the guard changes at a particular advertiser. We've all seen it happen: Old guard leaves behind their way of attribution. New guard comes in and blows it up. In the meantime, their spend usually tanks (usually because before they figure out the "new" way of doing it, they go to last click as a crutch). This simply doesn't happen in other forms of media becuase the bean counters there use the same methodologies. This bad news continues because people who use marketing mix modeling techniques look for stable, predictable measures, so what do they do about display? Ignore it since it more or less becomes a rounding error.

N.B. This is not to say that marketing mix models aren't flawed themselves as they are riddled with assumptions and plenty of their own "black box" predictive math (read: guessing). Still, CMO's trust this stuff so display severely cripples itself by not being able to plug into that, but that's for another post on another day.

So, let's start with some ground rules. Let's all agree that last click isn't best (and if you don't agree, then contact me and I'll show you hard data on why it's not).

The Ugly Math

The most common method of attribution beyond the click is a weighted average or distribution technique where each touch gets a fraction of a conversion. This method can be represented by the following formula:

where λ of t = attribution value of that "touch", t-1 = number of "touches" except for the one we're measuring, w of n = weighted percentage of each "touch", and Λ = value of the customer (typically lifetime value).

This formula itself makes a few disturbing assumptions like the fact one knows the lifetime value of the customer, but we'll go into that in my following post.

I surveyed four popular companies in the attribution management space and almost every single one talks about assigning proportional credit to each campaign for each conversion. Almost all of them allow you to plug in "customized" attribution models, which is great, but it's not a horribly fantastic feature. If you have a universal tag, then it means you have all the data in the world to apply any attribution model you want. It's just a question of which model you're going to use. Only one hinted at the fact that they have other models from the most simple to the most complex.

As we'll see in my next post though, complexity can sometimes be your undoing. Next week (hopefully), we'll talk about why proportional credit is fatally flawed.