Everyone gets a bidder

We had Shane Shevlin from the new customizable DSP Bedrock on the pod this week and it got me thinking about why people want their own ad tech stacks, and what are some good use cases. So I wrote about that. Also, Pubmatic came on to talk about all their new products.

Vendor interview: PubMatic

We all know PubMatic, or at least I thought that before interviewing Nishant Khatri, their EVP of Product.

PubMatic has been busy, building products like:

• Connect: Their data curation tool;

• Activate: A way for buyers to execute PG and PMPs without DSPs;

• Convert: Tools for commerce companies including sponsored PLAs;

• Open Wrap: An open source prebid wrapper.

Watch the interview with Ari. Free for a week before going behind the paywall.

Podcast: Shane Shevlin on customizable ad tech and unwittingly building a dating app

Shave Shevlin has seen it all. In his years at industry stalwart IPONWEB he fielded every crazy dreamer looking to create their own tech stack. Now he’s back at it with a new company called Bedrock, that promises traders an edge based on custom bidding technology.

We also hear about the origins of BidSwitch, about Carrickmacross, the small Irish town that is the epicenter of ad tech in Europe (“the Baltimore of Ireland”), and how ad tech algos accidentally powered a dating app.

Listen to the pod now:

• Spotify

• Apple

• Anywhere you listen

You get a bidder! You get a bidder! Everyone gets a bidder!

Having spent five-plus years in the DSP mines, I have a good feel for what people want out of a bidder. But even then, I sometimes wonder whether there is a real need for custom bidding technology, or whether the customers are more likely feeding their egos than creating real differentiation.

This week’s podcast with Shane Shevlin of Bedrock got me thinking again, so I’m putting down some thoughts.

In this piece I’m using a lot of shorthand. When I say “custom tech” I’m painting with a broad brush to cover options ranging from:

• Fully custom build

• Licensed, hosted software like that offered by IPONWEB

• The hybrid cloud solution offered by Beeswax

• SaaS solutions with deep APIs like The Trade Desk and Xandr

There are use cases that come up frequently as reasons why you might need a custom bidder. Below are my takes on what the common use cases are and what your best options are.

Custom bidding algorithms

There are a lot of smart people in ad tech, and many of them think they can write better algorithms and get better results than the default optimization systems in most DSPs. There are even vendors like Chalice.ai that promise custom algos that exceed DSP capabilities.

But the reality is that many platforms, including TTD, Beeswax, and Microsoft (Xandr) already give sophisticated users really powerful tools for controlling the bidding algorithms without building an entire new stack. If you spend time with the vendor APIs you can get 99% of your custom bidding algo with much less work.

Verdict: Build on an existing platforms using APIs.

Unique data

Companies that have unique data assets to monetize are often compelled to build their own bidders to express that data. These data assets might be difficult to express using the DSPs’ data store, or there could be other factors to consider that prompt the need for custom tech. Some of the use cases I’ve seen:

• Geography based data, especially custom polygons

• Probabilistic identity data that isn’t in the bidstream (e.g. need to fingerprint on the fly before bidding)

• Real-time data lookups or scoring

• URL-based or other supply-based lookups

There are a couple of things to consider when architecting a unique data solution. First, are you executing a media-based business model (i.e. selling ads) or would you prefer to allow others to sell the ads. Until recently you didn’t have much of a choice, but with the rise of curation tools offered by the major SSPs there may be a shorter path to monetization.

A second consideration is whether existing DSPs will allow you to use your unique data in their bid requests with an “augmentor” (sometimes called a pre-auction). In this model you would still need to stand-up a real-time data stack, but could avoid much of the custom bidding logic. To my knowledge, Beeswax is the most flexible in supporting this, but Xandr (and maybe others) can as well for larger customers.

Verdict: You probably need to invest in tech, but be smart and learn what the platforms can support for you.

Supply Relationships (SPO and more)

When you’re working with a DSP you are giving up some measure of control over which supply you have available to your bidder. This can be both a business problem and an execution problem.

Many ad networks utilize a bidder to buy their own supply. In this model, the network has publisher relationships and wants to run campaigns on a hybrid of both their own supply and open web auctions. This can be difficult to execute if your platform of choice does not want to listen to the auctions you control.

Another challenge that can arise when working with a DSP is the desire to stay on the cutting-edge of innovation. You read an article in AdExchanger (or listen to our podcast) and hear about a new company running auctions inside consumers’s brains using Neuralink. How do you get access to that in-brain supply when your DSP says it is “on the roadmap”.

Or, more prosaically, you may just want financially beneficial relationships with your SSP of choice, and that may be at odds with your DSP’s priorities.

Verdict: Controlling supply is a very good reason to go down the route of custom bidding technology, including a full tech build out.

Cost control

While DSPs will offer lower rates as volume increases, the cost picture when you have your own technology is very different. When you spin up your own tech you are going to have higher fixed costs, and much lower variable costs. The result should be that you get better economies as you scale your business.

Shifting your cost model from percent of media to a hybrid of fixed and percent will have varying shaped curves depending on the platform and business model. For example, while TTD is a very customizable DSP, their business model is almost exclusively percent of media, whereas Beeswax and IPONWEB/Bidswitch have flat components tied to percentages.

Verdict: Costs shouldn’t be your primary reason for investing in custom bidders, but it can be a compelling one. Do your homework and shop around. 

“Ownership”

I put this one in quotes, since it’s a vibe. If you’re a venture funded company, or looking for a high multiple exit, you may get feedback that owning the stack is important to your valuation. Fair enough. But if you told those same financial wizards that you were building your own data center instead of using a cloud service, they would think you were crazy. So where to draw the line?

There have been many, many companies with high multiple exits that were built on Beeswax, IPONWEB, or Xandr. But I also know of some cases where the reliance on one of these vendors was seen as a red flag in diligence.

Verdict: Do what’s right for your business and not what the investors say.

The bottom line is that there’s isn’t one solution that works for everyone. The leading vendors continue to make their platforms more flexible and expand their APIs. But some businesses still require their own tech.

Reading list

• Sir Martin says “Turkeys don’t vote for Christmas,” referring to the future of agency holding company staff members about to be eliminated by AI. See also, Ari’s Linkedin post making the same argument.

• Alex Kantrowitz breaks the story of what exactly Google’s news AI product claims to do

• NBCU using gen AI to map storylines to ads, seeing 49% boost in performance

• Disney connects with Google and TTD directly, and here’s the inevitable Twitter fight about it

• WSJ reports that Vizio data may dry up after acquisition by Walmart

• The Apple antitrust case, what’s in it and what’s not (advertising, commissions)

• MARKETECTURE EXCLUSIVE: We got a copy of Michael Kassan’s expense report!

More Marketecture.tv…