Editors’ note: Earlier this year, Selligent Head of Partnerships – Benelux, DACH & Poland, Nicolas Van Vijnckt, sat down with two experts in marketing automation: Alexei Kounine, VP of Product Management & Innovation (Selligent) and Michael Straathof, Commercial Director & Co-founder at Selligent agency partner 100%EMAIL, for a roundtable conversation about the importance of data and how artificial intelligence can assist brands and marketers today. These topics are addressed in a synergistic way, from both a technological point of view (Alexei) and the perspective of the marketeer and marketing consultant (Michael). The conversation is being presented in a multi-part series on our blog, and we hope you find the insights inspiring and practical. Here now is part two of our series.
NVV: Alexei, from a technology point of view, are there recommendations you can make for a brand to successfully deploy an AI project? Do you start with defining everything from the beginning and do you need state-of-the-art technology? Or can you also rely on a kind of phased approach, say a learning process?
AK: If you look at a standard campaign that is representative of the kinds of campaigns people usually do, you can pick any example, but it’s always about starting with segmentation. Or you select the individuals you want to target, then you create the content, and then you launch your campaign at a given time that is optimized based on all the tests you’ve been doing – your favorite day, your favorite time – and you send the campaign. After that, you look at the results via the dashboards and potentially rerun the same campaign a year away from now by optimizing it again. That’s the kind of standard process in which you would not make your decisions in a data-driven way. But actually, in every one of these steps you can optimize things using data and AI.
You can create an email by letting an algorithm select the content that you will place in the email based on the individual who will open it. So every individual will actually have a different email based on their previous behavior. We have a feature that recommends content, which is called Smart Content. That’s one thing.
The second thing is, normally when you do your segmentation, you always use a constraint editor. You’ll select the last people that clicked in a message, for example, over the last 30 days, who didn’t convert on the website. So you want to create a reengagement campaign or an abandoned cart. Again here, what you’re doing is thinking as a marketer, “What is my best target audience?” and then creating the rules to isolate these numbers of individuals in order to target them.
But actually, you can let an algorithm do that for you. For example, we have a feature called Smart Audiences, the goal of which is, based on the content of your message, to select the right individuals that will have the highest propensity of engaging with the message. And just using these two features, one that creates the content automatically or the other one that selects the right moment based on specific content, you can get engagement rates that are five times higher than what you would using a normal segmentation or a manual content selection. But then you can actually add AI all over the place in different features; you can predict the best send time, you can predict the best channel, you can predict the right number of messages to send over a period of time. All of these are key points in which you can use your data to drive your decision.
NVV: Data should really result in action. You talked about segmentation, targeting, and you gave a few examples. Marigold Recommendations is a platform that can make this data actionable. Today, there are a lot of AI solutions on the market, so depending on the goal, the expectation you have, and the investment you want to do and already have done, what is the best? Is it the best-of-breed or an integrated solution making use of the native database of the marketing platform, like, for example, Selligent, and this way maximizing consistency between the different modules of marketing automation philosophy?
AK: To me, the obvious question is always, “When you develop an algorithm, where is the data going to be that feeds this algorithm?” You need the initial data to train your algorithm. I think one of the core advantages or core characteristics of Marigold Recommendations – and by the way, all these features that I talked about, Smart Content, Smart Audience, Send Time Optimization, and so on, this is all Marigold Recommendations. One of the strengths is that you don’t need to think about the data pipelining, data cleaning, getting data from the website and correlating it with email data, ensuring that all the identifiers are the right ones, and so on – because this is all done natively in the platform.
When you think you want to try an AI capability from Cortex, what you have to do in Selligent is just start using the module, create your content or create your audience by click, and all the data-related tasks and all the machine learning tasks are done in the backend for you; you don’t need to think about this. So whether you should go for a best-of-breed solution doing only AI, or find something that is integrated directly with your core database that hosts all the data, to me, it is an obvious answer if you have the choice, but it is obviously dependent on the current setup that you have and the kind of tool that you’re looking for.
If you need to replace all your marketing automation suites by something that rebuilds the channels, collects the data, does the reporting, and adds AI capabilities on top of it, it is quite a big project. So I’m not surprised that some clients would go for just an AI addition to something that they already have today. What my goal is, is to build a platform that helps you grow from a very standard approach to marketing towards a modern marketer approach that uses all these AI-driven and data-driven tools within the same product. And this is what we’re doing, essentially.
MS: There you go, Alexei, you say all the right things! Isn’t this a great space to work in? Because if there is ever a time where you can make a difference as a marketeer, it’s now. If you’re reading this or you joined us in the webinar, when we talked more in-depth about examples, you can really boost your results in the short run by applying these kinds of techniques and the way you look at your customer contact strategy. We live in an amazing world now where the opportunities are in front of you.
NVV: Michael, if you know that Marigold Recommendations is a complete, integrated platform with your marketing automation and your marketing execution platform, or whether it’s a separate standalone tool, does that have a difference on how you look at things, in how you assess or design marketing processes?
MS: The main thing you should look at is how you can fulfill all those site conditions; how you organize all your conditions in order to reach your goals. And what that means is you look at your current staff, how capable and mature your staff is in doing all these kinds of things. You look at your content, how your content is organized, your data and how it’s stored, but also at your marketing technology stack. So, as what Alexei mentioned, is if you already have a marketing stack in place, it’s doing the right things, but you want to add AI and machine learning to your decisioning, then maybe a best-of-breed is the right answer. But if you have the choice to make your life easier, go for a suite solution where it’s all integrated and you don’t have to configure all connections between the data. That’s my opinion, but again, it all depends on how things currently are organized in your company.
NVV: Alexei, you referred to Marigold Recommendations: a native, integrated technology in Marigold Engage. Can you give a short explanation on Marigold Recommendations?
AK: Sure. Marigold Recommendations is the umbrella term for all the machine learning features that we have within Marigold Engage. These are structured into three pillars: we optimize content, we optimize journeys, and we optimize audiences. And by optimizing, I mean using the data that resides already in the platform to make predictions so that the metrics associated with the campaign are improved. With segmentation, you can do automatic segmentation using machine learning, you can get increases of performance by factors of up to five, which was seen by some clients. This is truly incredible if you think about it, that just by using a tool, you can multiply by five the amount of opens of your emails; just because you target the right individuals.
Same for journeys, for sending the message at the right time. We got incredible numbers in terms of increases of open rates, but also of conversions using Send Time Optimization. And for content recommendation, we had clients that experienced upsides, so increases of transformation on their websites and emails by 30%, just because the content was chosen wisely by the algorithm. And all of this fits into Marigold Recommendations – this is what it’s all about.
MS: I can pick this up with client examples as well, Nicolas, that results are increasing dramatically. It’s amazing how you see these results if you’re starting to use the kinds of smart technology in your daily operations. It’s amazing how you see people opening your emails and start interacting with them when they have the right level of hyper-personalization, given the fact that you just have clicked on something, or called a contact center, or ordered some product items. And the next moment you connect with that customer, it needs to have this kind of content in it. “Of course,” you might say, “but it’s so hard without all these techniques in order to manage all that.” With our tools and processes in place, it makes it easy for you.
NVV: We talked about artificial intelligence. Now let’s go back to data again. Today, we hear a lot about CDPs. Alexei, how would you define CDP? And does this also tie into Marigold Engage? Do we have certain CDP capabilities, or do you need a separate CDP platform? How do you see that?
AK: Just like any piece of the marketing tech stack, there are a lot of acronyms and a lot of different components that you make play together. I can give you two examples. I consider Marigold Engage to be two things: a data platform and an execution engine, or omnichannel execution engine. So, coming back to my previous example, if you start investing in your martech stack, you’ll see your company growing, and you need a solid solution to manage all your data in one place – and then use advanced and sophisticated tools to create marketing journeys that really make your clients engaged. In that case, Marigold Engage is the way to go. But we have clients that already have a data platform from another vendor. They need an execution engine to ensure that all their marketing goes flawlessly throughout all the different channels.
And in that case, we have strong API and connector capabilities that allow you to connect to any kind of CDP. So we always see these two types of setups: either Marigold Engage doing the hosting of the data and the execution, or sometimes we only do the execution because the data hosting is done somewhere else. And as I said, if the data hosting is somewhere else, then we still need to synchronize Marigold Engage with this data just to ensure that we have all the proper tools and the proper consumer profiles to make the automation work using machine learning or without using machine learning.
Watch for part three, the final in this series, as Alexei and Michael join us again for more conversation around this critical subject, coming soon to our blog.
Marigold: where relationships take root.