How to Boost Your Marketing Strategy With AI

With more data available than ever before, knowing what you can do with it is half the battle. When you add AI into the mix, you can amplify the impact of all the promotions you already use to make them that much more successful.

Join us for this webinar, “How to Boost Your Marketing Strategy with AI” to learn more about ways that you can use technology to improve interactions with your customers.

We will share how you can:

  • Use machine learning to discover buying clusters and craft the right promos for your guests
  • Cross-sell and make smart recommendations based on customer data
  • Stay on top of AI trends to know what is coming next

Ready to take the next step?

Schedule a Demo Browse More Resources

Read the Webinar Transcript

Hello, and welcome to today's webinar, How to Boost Your Marketing Strategy with AI.

And to introduce myself, my name is Jessica. I'll be your presenter today. I work on the content marketing team here at Paytronix specializing in the restaurant industry.

Paytronix serves more than 350 brands and the restaurants base, and that number is growing all the time. We provide tools and services for restaurants and convenience stores to engage with our guests through loyalty programs, gift card distribution and redemption, mobile application development, messaging services, data analytics, and ordering and delivery. Shown here is just a sampling of our customers.

So with that out of the way, let's talk about today's agenda. We're going to talk about how to use machine learning to discover buying clusters, and craft the right promotions for your guests. We'll talk about cross selling and making smart recommendations with data. And staying on top of AI trends to know what is coming next.

So, what is AI? Artificial intelligence makes it possible for machines to learn from experience, adjust to new inputs and perform human like tasks. Most of the AI examples that you hear about today, from chess playing computers to self driving cars, those rely heavily on deep learning and natural language processing.

Using these technologies, computers can be trained to accomplish specific tasks, by processing large amounts of data, and recognizing patterns in the data. So you probably hear several terms that are thrown around between AI, machine learning, and deep learning, and they're all AI. So think of AI as kind of the all encompassing bubble. That includes any system that leverages human capability, human capacities for learning perception and interaction that are all much more complex than our own abilities.

Then, within artificial intelligence, is machine learning, a subset of AI, that involves programming systems to involve to perform a specific task without having to code rule based instructions.

And then within machine learning, is deep learning. So that's a further subset where systems can learn hidden patterns from data by themselves, combine them together, and build much more efficient decision rules, and that’s how our restaurants are using AI. 

So, you know, if you're only thinking of AI as far as robots and self automation, you're doing it a bit of a disservice. Because there's a lot of ways that using AI can improve day-to-day functions and make them a lot more efficient. 

So one way is through AI menu boards. McDonald's made a splash in the spring when they purchased the company dynamic yield, and by May, they were starting to roll out menu boards using dynamic yields, AI technology, which makes different suggestions based on the time of day, weather and trending menu items.

And the more data the system gets through millions of daily customer interactions, the smarter these recommendations can become. Also this year Chick-fil-A built an in-house tool that monitors social media mentions that could indicate a customer might have gotten sick after eating at one of its restaurants. It screens data every 10 minutes looking for 500 relevant keywords, and then is passed through a natural language processor that checks the comments.

And they find that this approach has a 78% accuracy rate and helps Chick-fil-A track down the source of any foodborne illness quickly to prevent more customers from getting sick.

Beyond using AI to personalize offers, and promotions, Starbucks has used its AI efforts to also manage inventory and optimize labor. It's reduced the amount of administrative tasks inside restaurants, freeing up employees to talk to customers. Starbucks believes that having staff available to build that customer relationship will drive repeat visits and has a positive impact on their sales. And they're also using it to manage equipment. They have some new espresso machines that can pull three shots of espresso instead of two, which reduces the time necessary to create certain beverages. And since the new machines also have Internet of Things, sensors in them, they're able to see if their machines need tuning or maintenance. To not only improve the quality of the espresso shot, but also to perform maintenance before a machine actually breaks.

And so while there are a lot of great ways to use AI in restaurants really touching every department from finance to operations, for marketing one great way to think about it is that we want to turn AI into IA or individual action. So how can you use these insights and intelligence that we can gain from machine learning algorithms to really make a difference in sales?

And the first way we're going to talk about turning AI into IA is through K means clustering. So if you're not familiar with the term, you've probably encountered this in your daily life, when you shop at Amazon or are browsing Netflix, it's what is used to make recommendations to you, based on what people like you like. 

So sometimes you might see those things and think, wow, this is amazingly accurate. So a personal example for me is when Netflix first told me that I might like 18th century British period pieces based on literature, I thought, wow, that's incredibly specific. And then I also thought, yes, that's actually precisely what I like to watch. So the first step is really determining who the people like you are before you can make recommendations.

This is one type of machine learning algorithm used to create clusters of data. So you can imagine that this is your customer data. And you want to see what clusters exists within the data. First, you select your K, which indicates the number of clusters created. If you select too low of a number for K, you might miss out on potential clusters, But if you select too high of a K, your clusters may not have meaningful differences from each other. Theoretically, you could continue adding more clusters until only 1 or 2 data points were in each group.

But if you wouldn't interact with two of your clusters differently, it's possible that they're really just one large group.

So, to explain the algorithm, simply, this is how it works. So at random K number of points called centroids are inserted into the mapped out data. So I'm going to use stars to differentiate between the centroid than the data points. All the data points closest to each centroid are then assigned to its cluster. So we'll colorize them into those three colors there. So then the mean of all points, which in each cluster is calculated and the centroid moves to that new point.

Then the data points are re-assigned to the new centroid that might be closest to them.

So there are a handful of points that are going to change colors here and get re-clustered into new groups. So this continues through several iterations, until there's no more significant movement between the groups. And this is a basic example that can be plotted on just a X and Y axis because there's only two variables. But as you add more and more complexity to the situation, it becomes harder to visualize and much more dependent on those algorithms and machine computing power.

So how can you turn this kind of insight from artificial intelligence into individual action? And let's look at a clustering example.

So for this example, I'm going to walk through an example from one of our convenience stores who discovered clusters within their customer data.

Well, this example focuses on a uniquely convenient starter issue, getting people inside from the pump. You can apply this same strategy to a restaurant setting. So if you're a casual dining restaurant, you likely have cost clusters of customers who come in maybe for date night. So you see them purchasing to Andres and maybe alcoholic drinks. You might have families of two adults and maybe at least one kids’ menu or you might have people that never buy appetizers or dessert.

If you're a coffee chain, you might have people who are coffee only, tea only, maybe coffee and a pastry, coffee and a hot sandwich, et cetera. 

So it's easy to see how this solution will translate, but let’s walk through this convenience store example. This convenience store has people who come and never enter the store. They only get gas, and that's one cluster. Then they have people whose visits are driven by coffee. And they only get gas when necessary. They have visits that are driven by cigarettes and gas. And then people who come into the store for snacks, and then the final group, that's still gas driven, but kind of a mixed bag otherwise.

The thing is, that once you have these clusters determined, you can tailor your communication and offers to them accordingly.

So that first group, that only ever comes for gas and never comes into the store. Once you know that's a cluster, you can send them promotions to get them inside. So it could be a free dispense beverage. Could be $5 off anything in the store, or something like spend $5 in store to get 10% off for people that are driven by coffee, and occasionally get gas. Those are people you want to expand the basket beyond coffee. So maybe it's buy one, get one snack, et cetera.

It's kind of similar promotions as you go down the clusters. But once you know what those clusters are, you can craft really relevant promotion. So if you think back to those restaurant specific examples before, if you think about a coffee shop that has a coffee only segment, and maybe they have grab and go lunch segment that grabs a sandwich from a refrigerated case, you can send those sandwich driven people promotions that will get them to add a drink. Or you can send your coffee only purchasers, a promotion that encourages them to get any food item whether it's a pastry or something grab and go.

And another application of machine learning is through 1 to 1 visit challenges when thinking about driving visits and spend. You want to make sure you aren't cannibalizing sales, and that you're achieving increases. That are incremental. So, take this graph, as an example of the guests you might visit in a month. You have a lot of members, who maybe aren't visiting in a given month, a good amount that visit once a month, and then it decreases down the line.

And if you wanted to run a visit challenge, and you could only run just one, where do you set the threshold?

If you set a visit challenge at five visits, you can reach the accounts that are slightly below the level needed for a reward, but who would still find five visits achievable?

So, basically, those who are already making 2 to 4 visits a month, but you can't send the challenge to everyone else because they're already visiting that frequently.

If you bump up the challenge, you can’t reach these guests, but now you miss out on people with a lower frequency, who are actually a pretty large portion of your member base, but would require a pretty dramatic and maybe unlikely shift in behavior in order to complete the challenge. 

The great thing about 1 to 1 visit challenges is that you can set challenges that will reach all of these guests to see incremental visits and sends. Currently, from our customers who have used 1 to 1 visit challenges, we've seen 117% increase in reach, and nearly five times the sales lift.

So, if you can reach each person and push each visit to be incremental, then you're driving those increases and you can modify the promotion, so the best promotion is going to go out to each guest.

So maybe for some guests, the best promotion is to visit five times during July and receive 100 points.

But maybe for others, it might be to visit six times in maybe two weeks to get a free coffee. And through machine learning and AI, that can be really helpful here and figuring out which combination of these variables is going to result in the highest ROI.

The next thing we're going to talk, about is cross selling, so rather than offering the same things to everyone, online ordering platforms can learn over time which items are frequently purchased together. And this brings us back to another use for K means clustering. So, if your system learns that people who order a pepperoni pizza also likes to order garlic knots making that recommendation is likely to be more well received. 

So while a prompt that suggests a drink or a side salad to everyone might be effective, it's not very effective if people already have those items in their cart. So someone that orders an entree side salad might not respond well if they're prompted to add another salad as a side, but they might react well to an addition, let's say, of garlic bread or something else that they don't already have in their cart.

And one thing that we're working toward is to move from beyond an approach of cluster recommendations to move to a much more individualized recommendation algorithm and have it make suggestions on an individual basis and on an individual's ordering history.

So, what's next as far as staying on top of AI trends.

So, one thing to think about is customers are going to determine whether sharing their data in order to receive these highly targeted offers is worth it to them. Many states are either passing or considering privacy legislation, and customers will soon have the option to opt out of having their data sold or shared. So any value that they receive for letting brands use their data is going to need to outweigh any of their privacy concerns. So making sure that offers are really individualized and relevant is more important than ever.

More and more algorithms will be able to determine the right promotions in order to get results that you need to see. The more data the algorithms have to learn from, the more accurate they can be, in determining which decisions will result in the action you want as well. And then location based learning is also going to continue to grow. Especially as ordering technology becomes more prevalent within cars. 

You can imagine a situation where knowing your estimated time to get home with traffic is approximately 30 minutes. Maybe you're prompted to place your usual Thursday order now, so that dinner is there by the time you get home.

So to summarize, turning artificial intelligence into individual action brings your marketing campaigns that you're already running to the next level. By determining your clusters based on purchase data, you can reach customers and those like them with targeted offers.

After optimizing your visit challenges can boost your ROI by varying different variables, like time, length, and the reward, which is a huge impact on your business. And cross selling items that customers frequently purchased together, rather than just blind select suggestions, will also land well with your guests.

Artificial Intelligence And Future Concept Picture Id1034901762

Choosing the Loyalty Program That's Right For You

Download e-book
Annual Loyalty Report: 2021
Loyalty programs shone through as one of the key survival tools

The first annual Paytronix Loyalty Report examines trends across the loyalty landscape in both restaurants and convenience stores throughout 2020