m
Our Mission Statement
This is Photoshop's version of Loremer Ipsn gravida nibh vel velit auctoregorie sam alquet.Aenean sollicitudin, lorem quis bibendum auci elit consequat ipsutis sem nibh id elit.
Follow Us
Top
3 Key Points from our webinar Preparing for a New Programmatic Adtech World with TripleLift - VoltDB
8958
post-template-default,single,single-post,postid-8958,single-format-standard,mkd-core-1.0,highrise-ver-1.0,,mkd-smooth-page-transitions,mkd-ajax,mkd-grid-1300,mkd-blog-installed,mkd-header-standard,mkd-sticky-header-on-scroll-up,mkd-default-mobile-header,mkd-sticky-up-mobile-header,mkd-dropdown-slide-from-bottom,mkd-dark-header,mkd-header-style-on-scroll,mkd-full-width-wide-menu,mkd-header-standard-in-grid-shadow-disable,mkd-search-dropdown,mkd-side-menu-slide-from-right,wpb-js-composer js-comp-ver-5.1.1,vc_responsive
VoltDB / Consistency  / 3 Key Points from our webinar Preparing for a New Programmatic Adtech World with TripleLift

Blog

3 Key Points from our webinar Preparing for a New Programmatic Adtech World with TripleLift

One of the reasons I love working for a leading-edge database company is that I get to talk to and work with many different kinds of companies. Data is everywhere, and companies across the spectrum are trying to harness that data to improve their businesses, whether that means accelerating their growth through more accurate targeting of prospects, increasing customer retention by building 360° views of customers so they can improve customer satisfaction, or decreasing costs by automating inefficient labor-intensive processes. Adtech is an area where data is having a transformational effect.

(You can watch a replay of the webinar here).

1. Data is changing every business, including (particularly?) Adtech.

This is an obvious one, but it sets the stage for why we, as an in-memory database company, were hosting a webinar about Adtech with a cool leading edge company, TripleLift.

I used this diagram in my webinar slides because I think it shows the transformation in advertising very well.

Data is allowing advertisers finer granularity in the people they are reaching with their ads. Instead of broadcasting the same content to a very large audience and hoping that enough target people are reached, data such as rich user profiles allows content to be hyper-personalized and displayed to individuals who are being targeted. We’ve all noticed the effects of retargeting. Say you are looking to buy a new car and you go to the BMW website to check out the new models. For the next week, you are amazed at how many times you see ads for BMW or maybe Audi pop up, sometimes on websites that have nothing to do with cars, like CNN or Vanity Fair. You are being retargeted — the adtech software knows you showed recent interest in luxury performance cars, and that fact makes you a more valuable target for advertisers who are bidding on ad space on web pages you visit.

2. “Real time” is subjective — but “faster” is not

This is an important point to help companies cut through the hype. We at VoltDB talk about real-time, but often real-time is used incorrectly or inaccurately, so it has become a very imprecise phrase. You could certainly get into the physical argument that everything takes some certain amount of time. Grace Hopper, the famous US Navy rear admiral and computer scientist, was known to carry around a piece of wire that represented a nanosecond– that is, it was a length of wire that electricity could traverse in 1 nanosecond – which was 11.8 inches long, showing that even moving data from one computer node to another in the same rack takes some time, albeit a very small amount.

We’ve tried to put it in a bit more perspective and say that “near real time” isn’t good enough for some applications. However, if “real time” is an abused phrase, then “near real time” has been beaten to a pulp and is completely meaningless. Perhaps the best way to say it is that many companies are trying to do things faster than they can do them now.

I came to VoltDB from the analytics world where not too long ago, predictive models were built through a rather slow, inefficient iterative process that often took days or even weeks. This process involved the data scientist extracting some data from the data warehouse to his/her laptop, using analytics software to examine the data to create a predictive model. The model was then deployed against some test data, which gave the data scientist direction for tweaking the model, so more data may then have been extracted and the model refined, retested, and so on until a satisfactory model was deployed into production. My company at the time enabled customers to build those models more efficiently within the data warehouse itself, eliminating many repeated steps. This meant our customers could create each model faster and thus create more models in the same amount of time. When a customer could create 10x-20x the number of models or do model refinement 10x-20x faster, it changed the way those customers did business, letting them – and their customers – get returns faster.

3. Adtech companies have been doing things faster, but sometimes that meant sacrificing important things like tracking and transparency

Online advertising is constrained by time. Web page load speed will impact user defection/page abandonment– the longer it takes to fully render a page, the more likely it is that the requestor of that page will kill the window or try a different site. You might have a super-rich user profile collection and highly pertinent ads to place to a particular viewer, but if it takes you 10 seconds to make that decision, you won’t win any ad auctions. Supply Side Platforms and Demand Side Platforms are keenly aware of the need for speed and they drive very tight time limits of about 200 milliseconds for ad auctions (the blink of an eye takes about 100-400 ms). A lot of things need to happen within that 200 milliseconds, so adtech companies are pushing to make everything in that process faster.

Ad targeting began as a rather crude and imprecise thing, since a rich user profile wasn’t available or couldn’t be accessed reliably within ad auction time constraints. But network speeds and compute have both gotten faster, so more data can be accessed more quickly, and ad targeting has gotten more accurate.

But things like logging and reporting have sometimes been neglected in the race for speed. If NoSQL platforms are used, it might be a case where all the nodes don’t agree on a bit of data within the required SLAs, so anything that comes back from such a cluster is “approximate” at best…which could also be considered “wrong”. How much of the customer’s ad budget has been spent down already? Is there enough left for bidding on one more ad? How many times have we shown this ad to a particular user? Did the ad actually get rendered so it was viewable? Lots of different things that should be tracked and logged and reported aren’t because they can’t be easily with some types of software. This is why “transparency” has become a hot topic within Adtech.

Those are some of my takeaways from our webinar. Check out TripleLift if you are looking for the ability to do native programmatic ads, and check out VoltDB if you are trying to do more with your data and you want to do it faster.