Timothy Whitfield, director, technical operations at GroupM, weighs in on which ad tech terms are useful and which ones should be avoided at all costs.

There are currently over 5,000 known ad tech companies in the global landscape and that number is changing daily. It’s very hard to keep up with all the companies that are coming and going. It feels like every day that a new company emerges or gets purchased by another company. It’s easy to get “FOMO” when worrying about what ad tech your company is NOT yet using.

Therefore, it’s important that each ad tech company really tries to stand out in the crowd. They need to ensure that they (a) talking about their technology in an open and truthful way, whilst (b) ensuring that their audience understands what the hell they are selling!

Below is a list of keywords that are appearing more often in sales decks these days. Some of these keywords are helping these ad tech companies, but other keywords should be avoided like the plague.

Here is a list of keywords that your ad tech company either should or shouldn’t use…

1. Artificial Intelligence – NOT HOT

Artificial Intelligence is such a broad term that it doesn’t really mean anything any more. This was really confusing so I turned to some universities for help. They explained that the word no longer really has a defined meaning. It’s a generic term like “maths” or “calculation” and therefore has become meaningless. The only people still using it is out-dated ad tech, CEOs that think it sounds good on a sales deck.

2. Machine Learning – HOT

Machine Learning is a field of mathematics/computer science whereby you input a large dataset to a computer and tell that computer that you definitively know what that dataset is about. Then you ask the computer to tell you something about that dataset. A really good example is this: Upload 1,000,000 images of a car to a computer and tell the computer “This is a car”. Machine Learning techniques will then break down each image into its constituent components (wheels, windows, shiny metal, curved edges etc…) and the computer will then know what a car looks like. Then upload one more image of a car that the computer has not seen before. With any luck and some “machine learning” the computer should know that this is a car because it has wheels, windows and shiny metal surfaces.

Ad tech companies are starting to use machine learning to help with brand safety, contextual targeting and many other features. It’s very hot right now. But don’t overuse it.

3. Deep Learning – LUKEWARM

Deep Learning is a sub-set of machine learning. It’s almost the same as the idea above with machine learning but in deep learning you con’t tell the computer what the output is going to be. For instance, you upload 1,000,000 images of a car but you don’t tell the computer that these are all cars. Instead, you ask the computer “Tell me something about these images?” and a deep learning computer should say “Well, these all have 4 x wheels, windows and have shiny metal surfaces.”

Ad tech companies that are saying that they are using deep learning at the moment are few and far between. They may say that they use this technique in their sales decks but they need to be careful as it’s very hard to prove.


The word “algorithm” gets thrown around the industry like a hot potato. An algorithm is just a simple set of instructions used in calculations. We use them all the time in our daily lives. Consider this very basic example: Imaging sitting on the lounge after a long day at the office and viewing the EPG (Electronic Programme Guide). Here is a series of steps.

  • Input: List all of the available shows to watch
  • Calculate: Score each show between 1 and 100 based on which you like the most
  • Modifier: Game of Thrones just started
  • Output: Buy chips and chocolate cause we are watching the new GOT episode

Naturally, in ad tech the concepts are about cookies, users, databases and decisions. However the logic of the process is similar.

Don’t be overly impressed by companies that put too much emphasis on “The Holy Algorithm”. In ad tech companies that are Mangoes then the algorithm is simply a piece of code that works well.

“Some ad tech companies that use the term “The Algorithm” really mean “cheap labour that sit in an office offshore that manually optimise campaigns.”

5. Match Rates – VERY HOT

While this isn’t a really sexy subject, it is a very important one. A Match Rate is the percentage of unique users that can be matched between two different systems. For instance; there may be 1,000,000 people in your CRM system but what percentage of these people can you accurately target with online marketing?

There are ad tech companies out there that claim to have a 100 per cent match rate, which is of course impossible. There was one ad tech company that claimed to have an 80 per cent match rate but when tested only had an 8 per cent match rate.

The way match-rates are calculated is by connecting two different systems together that normally don’t talk to each other. The best match-rate that I’ve seen is created by the Rewards division of the Food chains. They send out reward emails almost daily and when the user clicks on these emails it forces the user to open a new page in a browser which then can match the user’s email address with their cookie ID. A very smart move. A good Match Rate could the be success or failure of an ad tech company.


It’s simply too hard for any ad tech company really to talk about data. In the beginning the first sales decks had a slide based on the amount of ‘transactions per month’ that that company processed. However, other ad tech companies quickly saw this as a weakness and updated their own sales decks to show that they are more data.

Then sales decks changed and talked about cookies. Each company seemed to have more cookies than any other company. Then this all changed with programmatic media. Everybody was suddenly doing cookie synching with everybody else. As soon as that happened then everybody has the same amount of cookies as everybody else and the point became moot.

Now sales decks are talking about data in the sheer volume of data processed. For instance one sales deck talks about 1,000 PetaBytes of data processed monthly. Does that actually mean anything to anyone? Would you be just as impressed if it was only 999 PetaBytes?

There are also sales decks which are now talking about GigaFlops of processing power. A FLOP is a Floating Point Operation per Second and is a measurement of how quickly a computer can process information. If you knew that your ad tech provider could process at the rate of 1,000 MegaFlops then would that make you feel like your investment was safer than an ad tech company that could only process at 999 MegaFlops?

7. Latency – VERY HOT (just not sexy)

This is an interesting one, especially for anybody that agrees with my point that there is a JavaScript bubble forming in the internet. Websites simply have too much code currently deployed and this is starting to impact on performance. After reviewing many ad tech it was clear that not many of them are optimised well for latency.

For example: one major global publisher has over 100 pieces of ad tech on their web site and it takes over one minute for a video ad to finish loading properly.

If you are the CEO of an ad tech company and you are using a Content Delivery Network (CDN) then you are one of the good guys/girls and you should be talking about it. If you have optimised your network to respond quickly and have ultra low latency with 3rd party ad-servers then you should be shouting about it. If you have decided to pay the extra cost and deploy servers in the major cities outside of the US and EU then you should be “sounding your barbaric yawp from the rooftops of the world” as that’s something to be proud of.

8. Server-Side – HOT

Server-Side programming is where some of the technology used to perform marketing tasks has been moved away from the client’s browser to the server for processing. One good example of this is Server-Side Ad-Insertion (aka Ad-Stitching). In this technology a server pre-fetches a digital advert on behalf of the consumer and “stitches” this video ad into the live stream of video that the use is consuming. The ad is then played seamlessly in the stream of content and there is no more pressure/stress on the browser to make this advertising experience happen. Rather than Client-Side JavaScript which is a relatively “hacky” experience featuring the “spinning wheel of latency death”, Server-Side tech is smooth to the user as the heavy lifting is done offline by a bank of servers in a data-centre.

Ad tech companies should look to use “Server-Side” processing where possible to maximise the user experience. Another good example of this is Automated Guaranteed.

9. Device-Graph – NOT HOT

There are a number of ad tech companies that make a big deal about their device graph. A device graph is a way to link together all the various digital IDs that a person has together into a single persona. For example; John Smith has a laptop, mobile phone, tablet and smart TV. Each of these devices is connected to the internet and has a different “digital footprint”. Each device is given a different ID. A device graph is a “sales tech” way of bringing these devices together under one user ID specifically for John Smith. Therefore, when somebody wants to target John with an ad then they can link all these IDs together to ensure that he is not shown the same ad too many times.

The problem with device graphs is that either (a) marketers believe that simply “should already exist” and therefore they become irrelevant or (b) marketers get too confused, don’t understand what the ad tech is saying, mutely nod as the sales presentation drones on. Either way device graphs are definitely not hot right now.

10. Blockchain – EXTREMELY HOT

Saving the best for last. Blockchain has the potential to change the industry. In fact, it has the ability to change the world (depending on how it gets used). Blockchain is the ability to store large amounts of information in a distributed, yet safe way. For instance; say that you have one user profile for “John Smith” that you wanted to store. That profile may include age, gender, basic geo, basic likes/dislikes etc… That information could easily be stored in a small database. Hell, you could even store that in a basic XLS file if you wanted. However, what happens when you have 1,000 profiles. Then you probably really need a database. What happens when you have 1,000,000 profiles then you probably need a distributed database. What about 1 billion profiles? What about 1 billion profiles that need to be openly read/updated securely by every ad tech company on the planet? A single database, no matter how powerful couldn’t manage it, not even Facebook could do that! That’s where blockchain would come in. It’s the ability to store very large amounts of data in “blocks” in a distributed manner across many computers (even yours!) and then “chain” this information together.

Soon, people will be talking about blockchain in the same way that people started talking about “the cloud”. People will say, “Just update the profile on blockchain” in the same way that people currently say “Just upload that photo to ‘the cloud'”.

“Tomorrow people will be talking about blockchain like they talk about the cloud today.”

If your ad tech company is taking advantage of Blockchain then you should certainly be talking about it.

In summary: The world of advertising technology is a multi-billion dollar industry and sometimes it’s hard to keep up. Even for the ad tech companies themselves. I hope that this list of Hot vs Not has helped with some of the key terms of what is trending well in the industry now.

Previous post

Customer impact should drive IT prioritisation, says PagerDuty CEO

Next post

Alibaba Cloud brings start-up comp to Australia

Join the digital transformation discussion and sign up for the Which-50 Irregular Insights newsletter.