Imagine for a moment that you are in a room full of slot machines.

Each slot machine costs a different price to play – there are ones as cheap as 10 cents and ones as expensive as $60.

Each slot machine has a different probability of paying out.  The probability is somewhere between 0% and 100%

The payout for every machine is the same.  Let’s say it’s $100.

You have a fixed budget – let’s say you walk in with $10,000 and you have 8 hours to maximize your earnings.  You must spend all $10,000 in that time.

What would your strategy be?  Just walk in and start pulling levers?

Let’s make this a little more interesting – a few more details about this hypothetical situation:

• The Casino is very, very large, has thousands of rooms, and millions of machines
• The Casino can add or remove slot machines whenever they want
• Adding new slot machines can lower the payout % of the other machines
• More than 50% of the machines have a 0% probability of paying out – paying for them is completely wasted money.
• About 20% of the machines not only have a 0% probability of paying out – but they also have a mechanism where when you pull the lever, the front of the machine opens up and a boxing glove on a spring punches you in the face.

Seems like a scene out of a Mr. Beast challenge.

Now, what if I told you that I had a device that could:

• Help you avoid the machines that punch you in the face
• Help you avoid the machines that pay out at 0%
• and plus or minus 10% points predict the payout probability of the rest of the machines

How much would you be willing to pay for that device?  Let’s say it costs $2,000 (20% of your budget).  Would you buy it?

Now what if you were presented with an alternative device that was much cheaper, but it was made by the Casino.

Which one would you choose?

Now let’s say that there is a salesperson in the casino that tells you the device is overpriced and you should focus your lever-pulls in just 1 room of the casino that only has 100 machines in it (and ignore the rest).

Would you take that deal?

How much more would you pay if the device was accurate within 2% points (vs. 10% points)?

The theoretical maximum you could win in this casino (if you buy the device, pull only the cheapest machines that cost $0.10, and they pay at ~90%) is $7,200,000. (If the device costs $2,000, you have $8,000 → 80,000 pulls × 0.9 × $100 = $7.2M). This isn’t a perfect analogy because there is a correlation between cost and probability – but even if you only pulled the most expensive machines ($60 cost) you’d still take home $12,000 (ROI positive).

If you pull randomly (assuming fairly generous probabilities: ~1% average payout and since you’re pulling randomly you’ll hit a mix of the machines — say $30 average cost) your expected value would be around $333 (which would mean your return on investment was -$9,667), and you’d get punched in the face about 66 times.

The example is a little silly, but this, in effect, is the case for objective decisioning — the kind you can trust precisely because the house didn’t build it and a salesperson isn’t getting paid based on your selection — and the massive advantage it produces if done correctly.

Thursday night I found myself sitting in the Las Vegas airport, stranded with a few thousand other CES go-ers thanks to a delayed flight home. As I do sometimes, I pulled out my phone and opened Facebook—just to pass a few minutes with some mindless scrolling. 

A few minutes in, I had a genuine “ah hah” moment. 

As I scrolled past AI-generated images, melodramatic headlines, and suspiciously verbose copy, a question clicked into place: 

How much of Facebook’s content can now be technically classified as MFA? 

A quick refresher on MFA 

A few years back, Chris Kane made waves by coining the term MFA — Made for Advertising. It gave language to something many of us already recognized on the open web. 

At a high level: 

Made for Advertising (MFA) content is content created primarily to generate ad revenue rather than to inform, entertain, or provide genuine value to users. It is typically produced cheaply and at scale, optimized for clicks and impressions, and often characterized by sensational headlines, low editorial standards, and minimal original insight. 

At the time, MFA was framed as an open web problem. Something buyers filtered out. Something publishers were warned away from. Something the ecosystem broadly agreed was low quality. 

Look at your Facebook Feed: 

Pull out your phone. Open the Facebook app. Look at the first ten stories in your feed. 

When I did this experiment, 7 of the first 10 pieces of content fell into one of two buckets: 

1. Ads 

2. Click-driven, low-effort content optimized for engagement, not value 

Here are real examples from my feed — not cherry-picked, just… there: 

• “Rome’s Arena Secret: The Fate of Female Prisoners After the Games” 
  (accompanied by an obviously AI-generated image and a breathless, cinematic opening written to keep you scrolling) 

• “This Is How a Woman’s Skull Looked After a Complex Facial Reconstruction Following a Car Accident” 
  (graphic imagery, minimal context, maximum emotional pull) 

• “LOOK: This Dad Tattooed His Baby’s Heart Surgery Scars on His Own Chest — So His Child Will Never Feel Alone” 
  (manufactured sentiment, reposted endlessly across copycat pages) 

• “BREAKING: Possibly the First Appearance in Its 50-Year Flying History, the ‘Doomsday Plane’ Appeared at LAX” 
  (urgent tone, speculative framing, recycled aviation trivia) 

• “This Is the Part of Roman History They Do Not Teach in School” 
  (a sentence designed entirely to trigger curiosity, not convey insight) 

Almost all of it is clearly engineered for clicks, reactions, and dwell time — not reader value. 

What was left? 

Out of the first ten items: 

• Two legitimate news stories 

• One sports highlight video 

That’s it. 

This is obviously a small sample size. But anecdotally, it matches what I’ve been experiencing for months. Conservatively, it wouldn’t surprise me if 30%-40%+ of Facebook’s feed is now effectively MFA — content that would be flagged as low quality or avoided entirely in the open web buying ecosystem. 

The rules are different here 

This is where things get interesting. 

On the open web, MFA is treated as a systemic problem: 

• Buyers block it 

• Publishers avoid becoming it 

• Platforms publicly commit to rooting it out 

On Facebook, however, the platform itself increasingly is the MFA surface. 

And advertisers are still pouring tens of billions of dollars into it. 

It’s also getting worse over time 

If you plotted Facebook content quality over the last few years, the trend line would be unmistakable: 

• More AI-generated images 

• More long-winded, pseudo-literary copy 

• More shock, fear, outrage, and sentimentality 

• Less editorial accountability 

• Less obvious human intent 

The feed increasingly feels optimized for scroll endurance, not insight. 

The new chum box 

There’s a deep irony here. 

For years, the open web relied on bottom-of-article recommendation widgets — the infamous “chum boxes” from companies like Taboola or Outbrain. Over time, the industry largely decided that model was too low quality, too spammy, and too corrosive to trust. 

So folks started moving away from it – Taboola and Outbrain have diversified their businesses and you see many fewer chum box ads that you did 10 years ago.  

In a strange twist, Facebook now feels like the largest chum box in the world — except it’s not at the bottom of the article. 

It is the article. 

Same mechanics. 
Same incentives. 
ust embedded directly into the feed. 

An uncomfortable question 

Maybe this is just the inevitable result of optimizing relentlessly for engagement at global scale. Maybe it’s what happens when cheap content generation meets powerful distribution. 

But I can’t shake the question I had sitting in that airport: 

If Facebook were part of the open web, how much of its own inventory would advertisers actually be willing to buy? 

And what does it say about the ecosystem that the answer might be… not very much. 

Interviewing product managers is uniquely difficult. You’re not just interviewing for skills, or experience, or even raw intelligence—though all of those matter. You’re also interviewing for judgement. And judgement is notoriously hard to assess in the artificial confines of a one-hour conversation.

To this day, the core of my product management interview comes from something I learned in 2015 from Jason Lynn—co-founder of mParticle and one of the best product thinkers in ad tech. When Jason first told me how he evaluates PMs, I recognized his genius immediately. I scribbled down every word, and for the last ten years, I’ve used a version of his method to hire and develop truly outstanding product teams.

Most PM interviews tend to cover the predictable categories: domain expertise, root-cause analysis, use-case construction, segmentation, strategy, frameworks. Good candidates generally come prepared with polished answers to those questions and examples from their experience.

But Jason’s interview goes farther and asks a core question that forces a candidate out of scripted answers and into the uncomfortable terrain where real product judgement lives.

It goes like this:

You’re two weeks away from launching a major product release that your engineering team has been working on for six months. A salesperson corners you in the hallway and says they’re working on a deal worth $10 million—if we deliver Feature X in the next two weeks. Feature X directly conflicts with the release you’ve been preparing.
What do you do?

Most people, even experienced PMs, approach this like a math problem:

• How many points of work is Feature X?
• How many points are left in the major release?
• What happens if we push the date back?
• Should we prioritize the release or the $10 million?

They frame it as a zero-sum tradeoff, as if the only two choices are A or B.

But product judgment rarely operates in binary.

PMs who’ve actually lived through this scenario many times will pause, smile, and immediately start asking different questions:

• How well do I know this salesperson?
• Who’s the client?
• What exactly is the remaining scope of the release?
• What’s the precise scope of the new feature?
• Can I speak with the client directly?

In other words: dealing in ambiguity is part of the job responsibility; PMs are constantly trying to improve their information orientation while at the same time understanding the optimal stop point to make a decision.  “Judgment” being the shortcut term for this.

In real life, the job isn’t to choose between two bad options. The job is to unlock a better option that no one realized existed.

The best PMs instinctively create new paths—by connecting the client, the salesperson, and engineering; by distinguishing between what’s truly required and what’s merely assumed; by reframing constraints; by discovering that Feature X, properly scoped, may not conflict at all; or by finding a path to deliver enough value to close the deal without blowing up six months of work.

This is the difference between PMs that are more administrative in their function, and PMs that truly have world class judgement.

Some PMs are excellent at keeping lists and organizing feature requests. They are reliable, structured, and efficient. Give them 5 units of effort and they’ll return 5 units of value. Nothing wrong with that. Many organizations run on this type of PM.

But the best PMs?

They see the world through the eyes of customers, engineers, salespeople, and executives simultaneously. They understand what actually creates value. They know when to bend a roadmap and when to protect it. They cultivate direct relationships with every stakeholder that matters. They produce 5 units of value with 3 units of effort because they’re optimizing from first principles, not just from a spreadsheet.

That’s the lesson of Jason Lynn and it’s still the best way I know for evaluating product judgement—because it reveals how a PM thinks when the rules are unclear, the stakes are high, and the right answer isn’t in any textbook.

Eric Roza once told me:  

“The only way to build a first-class technology product is with an empowered, incentive aligned Product Leader with direct CEO level accountability; if it happens any other way, it’s by chance.”  

This coaching hit home for me as I led a large and growing product management organization and started acting more like a Product Functional Leader than a Product Leader.  

Roza’s ultimate teaching was that both roles are important (and they’re a bit like Peter Drucker’s Leader vs. Manager dichotomy), but in product management – while the Product Functional Leader is helpful, the Product Leader is critical.  

Here is the dichotomy:  

The Product Leader vs. The Product Functional Leader  

• The Product Functional Leader is obsessed with the systems and processes that solve customers’ problems.  
• The Product Leader knows the customer by name and can see the world through their eyes.  

• The Product Functional Leader sweats the resource allocation across the portfolio.  
• The Product Leader sweats the pixels on the UX mocks.  

• The Product Functional Leader cares deeply about how many engineers are working on what.  
• The Product Leader cares deeply about who the engineers are that are working with them.  

• The Product Functional Leader has a spreadsheet of every product the company must ship this year.  
• The Product Leader has a spreadsheet of the features needed for the next release.  

• The Product Functional Leader ensures the organization is structured so great products can be created repeatedly.  
• The Product Leader gets the product into customers’ hands and watches how they use it.  

• The Product Functional Leader creates the roadmap, release plan, and sequencing for all products.  
• The Product Leader delivers an MVP and iterates quickly.  

• The Product Functional Leader stays up late coaching a Product Leader through a tough problem.  
• The Product Leader stays up late testing the product for themselves.  

• The Product Functional Leader does what is appropriate.  
• The Product Leader does what is necessary.  

• The Product Functional Leader designs the processes that keep the whole machine running.  
• The Product Leader will break the process if it helps accomplish their goals.  

• The Product Functional Leader owns the success of the team.  
• The Product Leader owns the success of the product.  

• The Product Functional Leader will coordinate with product marketing, sales, and field teams to rally the broader market.  
• The Product Leader will find the first customers, write the deck, and sign the first deals if she has to.  

• The Product Functional Leader bets on people.  
• The Product Leader bets on their product.  

• The Product Functional Leader spends a long time writing thoughtful performance reviews. 
• The Product Leader gives feedback in the moment.  

• The Product Functional Leader manages the product portfolio like a stock portfolio.  
• The Product Leader manages their product like it’s their baby. 

• The Product Functional Leader succeeds when their product ships on time and on budget with acceptable quality.   
• The Product Leader succeeds only when the product achieves its revenue and profitability goals. 

Over the past several years I’ve had the privilege of working with and benefiting from the mentorship of Eric Roza – former founder and CEO of DataLogix sold to Oracle in 2014 – and former CEO and owner of Crossfit.  This post was inspired by lessons I learned working closely with him.  

In 2012 Ben Horowitz published a list of good product manager behaviors. For the last 15 years or so I’ve been keeping my own list inspired by his original work, my long time mentor Dwight Porter, and the dozens of excellent product managers I’ve worked with over the years. Each one of these 17 bullets holds a story – many of them hard earned lessons. For folks who have worked with me before, some of these may be familiar. Enjoy!

• Makes decisions based on principles and data NOT based on beliefs and perception.

• Delivers products into the arms of waiting customers NOT builds products and then tries to find a customer.

• Goes to the source of the market feedback (line workers) to understand the full color NOT reads high level aggregated/summarized bullets from managers.

• Treats field team as an input into decision making NOT delegates decisions to the field team.

• Writes clear and concise wiki documentation NOT spends all their time answering individual one-off questions from the field.

• Seeks always to learn and understand first NOT proposing solutions first.

• Always understands the underlying problem we’re solving NOT executes without having full context.

• Builds products that create value for our customer’s business (and our customer’s customer’s business) NOT builds products that create value only for our business.

• Gets buy in early and often from key stakeholders and executive sponsors NOT working in a vacuum.

• Partners with engineering teams for input, feedback, and ideas NOT just for execution and delivery.

• Has a deep understanding of the entire “system” from customer problem to technical complexity to marketing and support NOT focuses only on one part of the system.

• Passionately evangelizes a vision that inspires others NOT expects others to do what you say.

• Builds relationships with functional leaders using data-backed research to influence decisions NOT dictates what functional leaders should do.

• Takes responsibility for 200% of the things that go wrong NOT point the finger at others.

• Surfaces data and options for functional leaders that makes the right decision obvious NOT makes decisions for functional leaders.

• In chaos, clings to problems, outcomes and data NOT just to what someone says should be done.

• Respectfully explains the obvious NOT assumes everyone knows what you’re talking about.

This might trend toward controversial territory, given that Prebid.js has been the foundation of how the open internet (certainly the web portion) has monetized itself for nearly a decade. But it’s worth re-examining what actually made Prebid successful — because the answer isn’t as simple as I (and perhaps others) once thought. 

When Prebid was first released, two things were undeniably true: 

1. Prebid made publishers more money. 

2. No one trusted Google. 

At the time, most people assumed those two facts were connected — that Prebid made publishers more money because Google’s ad server (then called DFP, now called GAM) wasn’t fairly allocating impressions. That assumption was true. The DOJ antitrust trial has recently revealed that Google AdX’s take rate was roughly 5 percentage points higher than the market average (20% vs. 15%). So, in effect, Prebid erased that hidden “Google tax” and put (at least) 5% back in publishers’ pockets (and prevented Google from influencing the market in other ways as well). 

But there was something else happening under the hood. 

Before header bidding, the browser would render a page, call one ad server (almost always DFP), and then that ad server would sequentially request bids from SSPs or networks. Prebid flipped that logic on its head: it let the browser call all SSPs simultaneously before the page loaded, then after all bids came back, an ad was selected. 

That subtle change — disassociating impression requests from actual impressions on the page — created a powerful new behavior in the ecosystem: bid request duplication (which I wrote about last week). 

And here’s the controversial bit: 

Prebid made publishers more money for two reasons: 

1. It removed Google’s bias. 

2. It multiplied bid requests. 

The question that no one can fully answer is: which factor mattered more? 
Did publishers’ revenue increase primarily because Google’s bias was eliminated — or because the same impression opportunity was duplicated across parallel bid requests? 

It’s difficult to prove this objectively, but there’s a strong case that bid request duplication was the larger driver. It created a massive short-term revenue spike for publishers — and a powerful incentive for everyone in ad tech to adopt Prebid as fast as possible. 

Some people recognized this even at the time. Jonathan Bellack (then at Google) debated Tom Shields about the long-term consequences of header bidding. I (and perhaps many of you?) was in the audience. Bellack’s argument — that Prebid’s efficiency gains were offset by waste and duplication — was rational, even prophetic. But it didn’t matter. 

Google had already lost the trust of the industry. And once trust is gone, logic doesn’t matter. 

I’ve read a few articles lately—mostly in and around the Transaction ID (TID) debate—that argue some version of the following point: 

“Impression duplication is good for publishers because it makes them more money.” 

On the surface, this sounds like a good argument – and in the short term it does. But, in my experience, there’s a nuanced story behind this statement and a strong argument that bid duplication is actually bad for everyone. 

The Rational Behavior Behind Duplication 

It’s absolutely true that impression duplication increases demand for a publisher. 

Chris Kane has articulated this better than anyone: publishers are behaving rationally when they implement every SSP, reseller, format, and bid duplication feature or trick that can help them inflate their bid request volume.  

In the short term, this strategy works. The first publisher to adopt a new duplication tactic—whether by adding another SSP, another reseller, or another format—sees a real lift in monetization. They’re taking demand from every other publisher. It’s the balloon-squeeze effect: push one side in, another side pops out. 

But here’s the problem: the advantage is relative, not absolute. It doesn’t create new demand—it just redirects the demand that already exists. 

The Balloon Squeeze 

The first publisher to duplicate impressions sees a big bump. The second publisher to copy them gets almost the same benefit, but slightly less. The third, fourth, and fifth publishers each get a little less again. Eventually, once more than half the market is duplicating impressions, the first movers begin to see declines. The balloon has nowhere left to go. 

At that point, the only way to maintain an advantage is to increase the level of duplication—add more resellers, more duplicative requests. It’s like a drug whose effects dull over time: to get the same high, you need a higher dose. 

The Dystopian Market 

When everyone duplicates, no one wins. 

Once the market has become saturated with a specific duplication tactic, every publisher that uses that tactic ends up seeing roughly the same demand they had before the tactic was introduced—only now, the market is flooded with redundant requests which makes it harder for DSPs to do their job creating performance for advertisers. 

Advertisers see worse performance. 
Spend shifts back to walled gardens. 
The open internet loses. 

And the irony is that this dynamic hurts everyone, including the folks who build and maintain the duplication tech in the first place. SSPs and Publishers face rising infrastructure costs from pumping inflated impression volumes through the pipes. Those extra queries have real server cost—and real climate impact. And once the duplication tech exists, they have to keep maintaining it and building onto it. Whole teams of engineers and operators end up working just to preserve impression duplication tactics that no longer produce any benefit.  Most SSP product leaders I know, bemoan the current state of the market (and in fact, here is Matt Sattel of OpenX last week on Marketcture discussing how duplication hurts them) – they have to spend precious engineering resources maintaining an inelegant system.  They would rather use those resources on real, durable value-added product features. 

Why the Sell Side Can’t Fix It Alone 

It’s structurally impossible for any one publisher or SSP to stop impression duplication. The moment someone on the sell side turns off a duplicative pathway, their revenue drops. They become the “skinny part” of the balloon and their money shifts to other pathways. 

This is why the problem won’t solve itself. As Chris Kane points out – the incentives are upside down. 

The only way out is a market reset—a shift where everyone stops duplicating at once and returns to a clean model where impressions are accurately represented and described. 

A Better Future for Everyone 

If the industry can move toward transparent auctions without duplication  — buyers will get better performance. Publishers will also win, because when advertisers get better performance, money comes out of the walled gardens and flows into the open internet.  

The kicker here is – I believe that SSPs will also win.  They will be free to focus on value-add activities: yield optimization, creative innovation, better user experiences vs. spending   resources building and maintaining duplication tech that, at best, provides short term benefit.   

In that world, buyers win.   Publishers win.   And everyone can stop wasting energy, engineering resources, and server costs on duplication and focus on building real market value that pulls spend out of walled gardens and helps grow the open internet. 

If you mix real Cheerios with knock off Cheerios (at a ratio of 1:1) – and close your eyes, you can barely taste the difference – and you can save $20 per month!

This is one of the lessons I learned starting out in NYC at the age of 22.  Accounting for my starting media planner salary and bare necessities, I had exactly forty dollars per month of discretionary income (half of which was due to the Cheerios hack).  Needless to say, I didn’t have enough money to do anything besides work — which was ok, because I was completely obsessed with trying to understand how media buying and selling actually worked.

On weekends, when no one was around, I’d come into the office alone (at the old FCB office at 100 W 33rd street), spread out across the floor with stacks of printouts, maps of the media landscape, and a dog-eared copy of Media Planning and Buying by Jack Sissors and Roger Baron. My goal was to create a unified theory — one framework that explained how advertising value was created, priced, and traded.

Here are a real photo of me doing this ~17 years ago (complete with original iPhone photo resolution).

One concept tripped me up more than any other: GRPs — Gross Rating Points. They were, and, in some places, still are, the foundation of how media is transacted. GRPs are calculated as reach × frequency, a deceptively simple equation with one fatal assumption: all ad impressions are worth the same amount.

Even then, that didn’t make sense to me.

The impact of a billboard you walk past in Times Square is not the same as the impact of a billboard you drive by at the holland tunnel.  The impact of a 15-second TV commercial at 2am is not the same as a 30-second TV commercial at 1pm. A digital video ad isn’t the same as a display ad. A viewable banner ad is not the same as one that never appeared on screen.

In my weekend scrawlings, I tried to rewrite the GRP formula. I added a third variable: impact. I even concepted a wearable device that could be outfitted on a panel of people, recording everything they saw across every media type— an omnichannel measure of advertising exposure and impact. (This was 2008, so wildly unrealistic at the time. Today, it’s probably possible.)

What I was reaching for was a system that priced attention and impact, not just exposure.

Back then, the media world was still largely IO-based. But today, in the era of programmatic and real-time bidding, we finally have something that can achieve what I was sketching on those floor maps: price discovery.

Price discovery is the mechanism by which each impression finds its own fair value — where buyers and sellers, through transparency, accurate descriptions, and an open auction, determine what an ad is actually worth based on its context, visibility, engagement, and outcomes.

In a sense, price discovery is the modern evolution of the GRP. It’s a more elegant, dynamic, and empirical way to measure value. Instead of assuming all impressions are equal, price discovery lets the market decide which impressions truly matter.

Bad things happen when there’s no price discovery.

When prices are opaque or disconnected from quality, good impressions get lumped in with bad ones. Incentives drift away quality and toward volume and obfuscation instead of impact. This is one of the most persistent problems facing the open internet today.

Using my Cheerios hack as an analogy… a GRP based system without price discovery, incentivizes sellers to put more and more fake Cheerios into the bowl because quality is not accurately valued or rewarded.

If we want the open internet to compete on a level playing field with walled gardens, buyers, armed with advanced DSPs, should be able to value each impression at exactly what it is worth to them.

That’s how we move from a world of GRPs and “lowest common denominators” to one of efficient and accurate valuation — and finally give the open web the economic foundation it deserves.

In his 2010 book Cognitive Surplus, Clay Shirky made the argument that Humans watched too much low quality TV, which was a waste of time, and that new technology (mobile phones, social networking) will unlock that wasted time and turn it into positive productivity.  In a particularly memorable analogy, he compared the habit of modern sit-com watching to the 18th century habit of sniffing gin-soaked rags—numbing the brain with passive consumption.

I think it’s fair to say the media landscape has evolved quite differently than Shirkey predicted. If watching sit-coms is like sniffing gin rags, then doom-scrolling social media is like taking fentanyl. That might sound extreme, but you get the point. A softer analogy is food: user generated content is often like dessert, while news and high-quality content are more like vegetables. The problem is that right now, our collective media diets are unhealthy.

And that imbalance matters. It’s not just about how much time we spend scrolling (and what that does to our brains or mental health); it’s about how our perspectives narrow when we only consume content that reflects back what we already think.

The View From Home

I see this play out every day with my three kids. Like all parents, I want them to grow up healthy—in body, mind, and spirit. But in today’s digital environment, that’s hard.

They love watching Minecraft videos and MrBeast challenges, and I’m fine with that in moderation. But I also see the darker side: even the best social platforms can’t fully prevent inappropriate or harmful content from slipping in. Roblox is a perfect example—it’s designed for creativity, but has had significant safety issues.

As a parent, it’s terrifying. And as someone who has worked in media and advertising for nearly two decades, it’s clear to me that the system itself is off balance. Right now, advertising dollars are flowing disproportionately to “desserts” and “drugs”—the addictive, empty-calorie parts of the internet—while the vegetables and fruits of the open internet barely scrape by.

Why the Open Internet Matters

The “open internet” is the decentralized network of independent publishers, creators, and content producers who get paid to make high-quality content. These are the journalists, artists, educators, and storytellers who enrich culture and deepen our understanding of the world.

But the open internet is not healthy right now. It’s competing against walled gardens that optimize for engagement at all costs. And when the ad dollars don’t flow to the open internet, the quality of the diet we all consume suffers.

So the big question is: how do we create a healthier, more balanced digital diet for the next generation?

Why I Joined The Trade Desk

The Trade Desk is our best shot at making the internet meaningfully better.  It comes down to three things:

1. Mission. TTD is fully committed to building and strengthening the open internet—not as a side project, but as its core purpose.
   
2. Founder Leadership. In industries that move this fast, only founder-led companies have the conviction and courage to make big, disruptive bets.
   
3. Data at the Core. To reshape the internet, we need analytics, transparency, and data-driven decision making at the heart of the system—exactly the kind of approach The Trade Desk has pioneered (and companies like Sincera have championed).

That’s why I’m joining The Trade Desk. It’s the best shot we have at rebalancing the internet, creating a system where quality content can thrive, and ensuring that the digital world we pass down to our kids is one worth inheriting.

I’m excited to join Mike O’Sullivan, Jeff Green, Samantha Jacobson, and the entire TTD team on their mission to transform media for the benefit of humankind.

After seven and a half wonderful years at AppNexus / Xandr (which was the best home and family I ever could have asked for) I’ve made the exciting decision to join the TripleLift team as SVP, Product.  Official coverage here from MediaPost.

What is TripleLift?

TripleLift got its start as the leading programmatic native advertising platform (think – advertising that is integrated seamlessly into webpages – not banner ads), but is quickly emerging as a leader in the native category and investing in adjacent areas.

Why did I decide to join the TripleLift Team?

There are 5 key reasons why I think TripleLift is the next breakout star in ad tech.

1)    Native ads are better for advertisers, publishers, users, and the internet.

Some ad tech insiders have grown skeptical of native advertising due to “content recommendation widget” companies showing predominately low quality “native” ads tacked onto the bottom of websites.  However, TripleLift is different. TripleLift’s core business facilitates real in-feed native advertising that is integrated seamlessly into the content of a website (similar to Facebook in-feed ads).  These native ads perform better for advertisers, generate more revenue for publishers, and provide a better user experience than standard banner ads. It’s no wonder that native advertising (and particularly native in-feed advertising) is one of the fastest growing segments of the advertising market.

2)    Investing in cutting edge tech

Because native advertising involves serving ads into odd shape ad slots with mis-matched creative assets – TripleLift has become expert at Computer Vision and image scanning technology to accurately crop, resize, and render native ads programmatically.  This technology extends naturally into new verticals like OTT, messaging, VR/AR, social platforms, and more, where TripleLift is making major investments. Also, since TripleLift always renders its native ads (rather than serving a 3rd party tag), it’s immune to a lot of the malware, ad stuffing, and poor-quality ad problems that plague other ad tech companies.

3)    TripleLift is hiring, profitable, and relatively bootstrapped

TripleLift has achieved scale, is growing quickly, hiring, profitable, and, having only raised $16M, relatively bootstrapped compared to a lot of its ad tech peers.  As an employment brand, it’s one of the brightest spots in a quickly maturing industry.

4)    Solid technical foundation

The founders of TripleLift are 3 ex-AppNexus employees (and my former colleagues).  AppNexus has become synonymous with excellent technology and the TripleLift leadership team has maintained that tech-first mindset.

5)    Diversifying outside of core ad tech

TripleLift has successfully launched and commercialized a branded content product (Content Dial) that makes branded content easy to scale for brands and publishers through an automated marketplace.  This “non-ad tech” line of business helps diversity the company into new markets.

For these reasons I’m thrilled to make TripleLift my next home and I look forward to an exciting adventure in this next chapter.