Originally appeared at Martech Advisor March 2020.
Here’s something we often hear from advertisers:
“What makes you different? And please don’t say ‘machine learning and AI’… we get dozens of people telling us that’s what they do, and it’s rarely true.”
Companies know what buzzwords prospective clients want to hear, and they’ll use them regardless of whether they truly relate to their offering. This not only damages the client’s trust; it also harms the entire ecosystem’s push towards a safer and more efficient industry.
So what can advertisers to do best decipher whether or not a company is actually data-based?
1. Use LinkedIn
One of the simplest, yet most effective ways to determine whether a company is AI-based, is to use LinkedIn as a research tool. A data-driven company will have quite a few R&D employees, analysts, and data scientists on board. I’ve seen countless companies that define themselves as technology-driven or AI-based, but all of their employees on LinkedIn are account managers or in business development. One data scientist isn’t enough, either, therefore the smaller the company, the bigger the percentage of tech employees it should employ.
2. An automatic rule engine isn’t necessarily AI!
Clients exploring potential partnerships will often hear something like, “Our system automatically blacklists publishers.” Clients should make the important distinction between automatic rules engines and machine learning. Automatic rules engines can be very effective tools; however, they can be very limiting if the engine comprises only a set of simple rules that any campaign manager can generate. This type of engine might determine that a publisher should be blacklisted if it has X spend with zero installs, but this misses the bigger picture – which is where machine learning comes into play.
Machine learning models factor in thousands of different features and the interactions between them, continuously improving themselves so they can perform optimally based on the most updated data. The result is a much more detailed picture that can allow the media buying process to be based on truly informed choices.
3. Check for Professional Articles
Companies with a deep understanding of machine learning often publish professional articles on these subjects in addition to case studies and other resources, which can usually be found on the company’s website. Search their website, read the articles and evaluate the company’s knowledge and expertise and if you can, go the extra mile and have an analyst or data scientist read the article and give their opinion on it. Keep in mind, however, that the lack of these types of articles on a company’s website doesn’t imply that the company is not tech-oriented, but having such articles implies that a company is.
4. Data Sharing & Getting Insights
A company that deals seriously with data should be able to give you insights right off the bat, even before going live with a campaign. They should be able to do so either from their experience or from analyzing the data you sent them about your users. For example, if you send a data-driven company a user list, they should be able to offer new insights about some or all of your users, such as the users’ interests, how active are they in places outside your app and more. It also helps to check whether the company’s insights align with your experience and knowledge, as you can supply them with your data, ask for a churn prediction of your user and then check whether their prediction matches yours. Remember that you’ll need to share your data if you want to generate high-quality, tailored results for you, as even the giants (Google and Facebook) require that clients share their data in order to build the best profiles possible and address the correct audiences.
5. Focus and Expertise
Collecting and analyzing data is a difficult, strenuous process that demands focus and time, while the learning process is gradual and time-consuming. You need to start very narrow and expand from there, understanding that each expansion requires many adjustments. Remember that “learning” is half of the term “machine learning,” so when working with data companies, you’ll probably find that they will want to start relatively small—with one to a few geos, a relatively small volume, specific verticals, etc. If a company promises volume, reach and results from day one, where is the learning part of the equation?
Taking Time to Find the Right Fit
Companies whose services are genuinely based on AI and machine learning have typically put in a great amount of work to build a powerful data-based product. With this, it’s worthwhile to take the measures needed to successfully identify these companies, versus those that use the terms “AI” and “machine learning” without really having the expertise or offerings to back them up.