How Monocle Uses AI to Revolutionize Ecommerce Discounts

How Monocle Uses AI to Revolutionize Ecommerce Discounts
Smiling face
  • Twitter
  • Facebook
  • Linkedin
  • Email

In this discussion with Noam Szpiro and Mark Lotman from Monocle, we explore how their platform is revolutionizing promotional strategies for ecommerce brands. Discover the science behind smarter promotional decisions and how Monocle empowers brands to optimize for success.

1. Customizable discount strategies tailored to user behavior

2. Rapid integration and easy onboarding for marketers

3. Advanced analytics to verify and optimize promotions

4. Options to optimize based on revenue or gross profit

5. Proven success with consistent lifts in gross margin for top brands

Read on how Monocle is shaping the future of eCommerce promotions.

Wiehan  Britz: Welcome to another episode of Inbox Invaders. Today I'm joined by two very special guests. I'm joined by Noam Szpiro and Mark Lotman. They are the guys at Monocle. And I'll have them quickly introduce themselves and then we're going to be talking about the platform and the problems they're solving and how you can use the tool to replace your current static accounts with AI powered incentives. I'm looking forward to digging in.

Mark Lotman: Thank you for having us. Yeah. So I'm Mark, co-founder, COO of Monocle, based here in New York.

Noam Szpiro: I'm Noam, co-founder and CEO of Monocle, mostly based in Tel Aviv, sometimes in New York as well.

Wiehan  Britz: Okay, I think let's start.

Mark Lotman: Yeah.

Wiehan  Britz: Problems you guys are solving for ecommerce brands.

Noam Szpiro: Yeah, we'll just give you maybe a quick high level overview. So the way I like to think about it is we're a promotion operating system. We help brands improve their promotional spend by making it much more efficient. So a little bit of background I come from, I worked for Lyft for a long time. So a big ride sharing company in the US, and then Instacart, a big e commerce company also in the US.

And what I did there at Lyft at least, was we basically built a very sophisticated team that was in charge of figuring out how to allocate promotional budget to all of our user base. And initially we did what every e commerce company does, which is let's give everyone who hasn't taken a ride in the last few weeks a coupon. Same thing they do in a card abandonment flow.

Noam Szpiro: And that worked really well. People still come back when you give them coupons. But what was surprising to us at least, is we saw the moment we add some sophistication around that and machine learning that really looks at the user behavior and starts targeting people based on that. We see massive returns, basically. So we basically doubled our ROI when we started looking at how people are behaving once we give them a discount. And that's when we realized this is a super powerful tool. And it's unfortunate that it's only, it's super complex to actually build something like that. And, you know, we need, we can build a team around that, that can basically offer that as a service to e commerce companies.

Noam Szpiro: And so that's what we set up to do about a year and a half ago. So at the high level, the idea is just trying to understand what is the cause and effect of giving people promotion. And then from that, trying to understand the incrementality of each individual offer based on what the user has done on the site and then giving them the best possible offer.

So sometimes that's no discount, you know, if the ROI is actually negative on that discount. And sometimes that, you know, some users need a little bit more of nudge to get them to complete the purchase. So we do that too. And then we also do that all based on the cost of the items. So we take a look into gross profit as well and optimize things based on that.

Noam Szpiro: So that's kind of at a high level. We started the company, as I said, a year and a half ago. We raised a large sea ground then and since a year ago we started working with clients and now working with Dor Dryas, one of the clients there too. Anything to add, Mark?

Mark Lotman: Yeah. So to your question, like, you know, if you look at the world today outside of, within consumer, kind of outside of the big tech platforms, there's a lot of disjointed tools that help you run discounts today, but do it in a way that's very, very complicated from operational perspective because you have to tie between excel spreadsheets, different tools you have, it may be on the site, different email implementations that you do with your agency, kind of your marketing agency or in house.

So you kind of have to string all of those different tools and then on the back end of it, from a customer perspective, consumer perspective, a lot of the times folks see very different discounts throughout their consumer journey because they're coming from different teams and because they're coming from different systems.

And so you have this situation where even before we lay over the AI, we have this sub optimal experience both for the operational marketing teams of brands and also for the consumers, which creates a lot of distrust kind of in a lot of work. And so outside of the benefits that not mentioned in just creating more value for brands on their promotional spend and the ROI on that, we find that brands kind of love using our platform also for, you know, just streamlining operations and helping them kind of unify a customer experience and a customer journey into something that's a lot more consistent and kind of trust building instead of trust eroding.

Wiehan  Britz: The problems people are facing. And while you were chatting, I was actually looking at one of your landing pages here. I'll quickly share my screen just to show you exactly what. But, and I think they know their website quite well because whatever they've said right, right now is reflected on the website. I like this notion of scaling incentive decision making across millions of customer interactions. So we've worked exactly like you've said. So we work with small, medium, large brands. And the moment you get all these different touch points interactions, it becomes al trying to a, like you say, plan out the incentives.

Wiehan  Britz: What should it be? Doing calculations to a task, you know, different stages of the life cycle, it becomes a nightmare. And I love the questions you're also posing here is, can I reduce my discount spend without losing customers? We get this constantly and we do a b tests where we reduce the discount amount, but then the place order rate reduces because people now aren't interested because you've kind of stripped out your, your discount code completely.

So now you've got this, the weighting factor where on the one end you're saving money margins, on the other end you're losing volumes. Who should receive a incentive when, how, what? This is brilliant. But how does this work in the background? Do you guys feed it with like millions of touch points? Do you need historical data? How do you get started with this? Like, how can I get this machine to think for me?

Noam Szpiro: Yeah, so, yeah, I'll answer that just before that. I do agree with a lot of what you said, and one point I'll add too, is that we do see a lot of ads, a lot of brands actually do a b test, which is great to actually understand. Okay, this is the best performing discount and I'm just going to give that to everyone. The problem with that is it's kind of like doing an A B test for an ad that you're doing on Facebook.

Sure, like you find out if the ad is a good thing or a bad thing, but that's just the beginning. You need to constantly keep optimizing it and you can't just say, okay, this works, and I'm just going to forget about it. You need to make sure it's always optimized and it's always changing. Like the best possible coupon isn't the same throughout the entire year, it's obviously changing.

Noam Szpiro: And then on top of that, it's hiding the heterogeneous effect of people. So some people would affect respond positively to coupon, some people will just maybe not actually change your conversion and therefore hurt on aov, essentially. So that's why using a tool like ours is actually beneficial, because it uncovers all those things. But to your question on how this actually works.

So the machine learning we use behind the scenes, it's called causal inference. And that is actually very different than what exists today in the market around machine learning or data science in general, in the e commerce enablement space. So a lot of the tools out there you see today are around personalization. And what they typically do is they look at historical data and build correlations between what they've done, what users have done in the past.

Noam Szpiro: That works really well for certain areas of the business. A typical example of that is, you know, recommending products that go well together. For example, if a lot of people purchase a battery and torch together, it's likely a good item to offer them to buy together. That's great. With, with promotions, it's a little bit different because you actually don't want to look at correlation. You want to actually look at causation. So the problem here is to make sure that if you're giving someone a coupon and they used it, that doesn't mean it was a good coupon. You want to give someone a coupon that made them actually purchase, whereas without the coupon, they would not have purchased.

Noam Szpiro: And so to do that, it's a little bit more complicated. So what we do is once we integrate into a company, we have a short learning period where we call it an explore phase, where we're observing how people are reacting to different incentives. Essentially, that gives us an initial model where we can build on top of that and kind of build an initial model that tells us how are people reacting in general to these discounts.

And, you know, we might find, okay, like users that spend over nine minutes on the website, they typically need a $5 off discount, and that is the best discount for those people. And then we, we have millions of signals that we're processing, basically. And then for each individual user, we end up coming up with a different prediction on what is the incremental value we will get from giving them a discount. And then we do it based on that. And then after that week, we can move to an optimized model.

Noam Szpiro: And then throughout that period, we're constantly learning at the same time. So we're always, there's always a little bit of explore, we call it. So even if we think anyone who's called mark should get 10% off, sometimes mark will get 20% off or will not get a discount just to see, hey, are we still accurate? Or maybe there's a seasonal shift, something like that, where it might actually mean that we should not give Mark a discount anymore.

Wiehan  Britz: The one thing that I also love is. So you talking about the science behind it? Well, science, I think everybody calls it a science. I have looked at some of your screenshots as well, and feel free to share your screen to show us around a bit is you guys do have a dashboard where you reveal some of these findings or observations in the data set. It looks like the discount percentage should be 16%. I find with a lot of predictive analytics AI tools, it's a bit of a Pandora's box, black box. You just don't get exposure to any of these things. Which I think is critical in today's world is can you allow the email and SMS marker to actually communicate to whoever what these numbers are, what the machine is telling us? So I don't need examples that you can show just to show us how it gets spits out on the other side.

Noam Szpiro: Yeah. Let me try and log in from my laptop and see if that will.

Wiehan  Britz: Work perfect, because I do see you guys also integrate with attentive and Klaviyo and sail through and Shopify. It's quite extensive, so I can see the complexity around it.

Mark Lotman: Yeah, so a couple of things, just both on the setup before and just on the black box point on the setup before. You know, everything that Novgorod described we can make happen within a couple of days. So the integrations are really simple and our team is taking care of onboarding. Onboarding, on average takes between two or three days at this point.

And then secondly, to your point of a black box, the way we're seeing it, we kind of want to be the to make marketeers kind of superheroes. And what we want to try to do is have them have good reasoning behind discounts, which is something that's really hard to do today. And I think that a lot of folks feel like they're guessing a little bit when they're coming with a new promotion. And we just want to take that guesswork away.

Mark Lotman: We don't want to replace that with this mysterious model that we're just going to run in the background and nobody knows what's happening. And instead, what we want to have people is both have the confidence that they're making the right decision, but then also be able to explain that decision and potentially also use the same reasoning. The same kind of features that we're using in our models to run ads have other extensions that they're targeting folks with that would just extend their capability beyond incentives.

Noam Szpiro: Cool. Let me quickly show you some screenshots. Part of what we do is we want to just provide some analytics on promotions. So how are promotions being spent in general? So that's part of the app. The other part is around the optimization. So in this case, this is maybe a flow that we're optimizing. You can see here, this is the distribution that we found to be the most optimal one. So some people are getting no discount, some people are getting 25% off and the rest are something in between.

Noam Szpiro: And then you can see here, how's AOB, how many shoppers revenue all of that? We always like to do experiments to make sure we're actually providing value to the brand. So in this case we're seeing an 83% uplift, which amounts to about 240k annually. And then you can see where the uplift is coming from. Is it coming from conversion or from aov? In this case it's both. And then you can see how many orders per user, things like that. I'll just touch on that quickly. This is where you can see what you mentioned, asked about before, where you can actually see what are the top reasons we're giving people discounts. So in this case it's number of items in the cart has a huge implication for how sensitive you are to discount.

Noam Szpiro: So you have between one and 20 and one, you're not very sensitive above. Two is when you become pretty sensitive. Number of email interactions is a big one as well. Part subtotal where you actually signed up, things like that, essentially. Just one more thing I'll show you here. This is what we call the promotional strategy. So this is where we let marketers decide, a what is the discounts that, what are the discounts that they want to give? And then b how aggressive or conservative do they want to be with discounts? In this case, maybe they pick sustainable growth. So the expected uplift is about 46%.

Noam Szpiro: And this is the distribution of discounts that you will get with this strategy. And then maybe the next month you want to be a little bit more aggressive, you want to hit your revenue goals. So you can pick promotional growth. That means 70% percent of users will get an aggressive coupon. So the 25% off and that will generate 57% or approximately 57% in revenue uplift. And then on the other extreme, you know, you can always pick something that means, you know, most people aren't going to get a discount. So in this case 50% of users aren't going to get a discount. And then the ones that will get a pretty low discount and then that will still generate about 33% of revenue uplift.

Noam Szpiro: So the idea is that the marketers can come in and pick the strategy that works for them and then with a click of a button they can change that. And then they can also change what is the actual discount that they want to give. So if they want to remove one of the discounts, it's just a click of a button. Nothing to change in the emails because the emails are already dynamic, essentially.

Wiehan  Britz: Interesting question on that one. So if I get this right, so you still need some input from the user. Exactly. Like you said, we're willing to go up to 25% in discount value. So I need to set that in the system. And then the strategy, do you flip flop? Can you choose that every when? How does that work?

Noam Szpiro: Yeah, so, you know, when you, after you set it up, you can, if you want to just set it and forget it. Just pick sustainable growth, select these promos and leave it be if you want. But, you know, a lot of our brands want to occasionally test a new discount, for example. So they want to see, yeah, they want to see maybe free shipping over $30, maybe that's the right thing to test instead of giving 20% off.

So they might try adding the new discount and seeing in a few weeks how that's performing. Is the model giving it more often than it used to then, you know, sometimes some brands typically do, for example, for Labor Day they did, they maybe do a, sorry, Memorial Day, they do maybe a discount on the site and then in that case they want to maybe match it in email or maybe want to remove the discount in the email. So they can easily do that basically by going in here and changing things.

Mark Lotman: And one thing, so the point of the promotional strategies, options is to keep control with the brand. The way we're trying to think of ourselves as something that enables brands to scale incentive decision and do it in a way that's more based on kind of science than guesswork. And so, you know, we don't want to take control over the strategy of the brand. We don't want to take control over, you know, the goals we're trying to optimize.

Instead, we want to enable them as we pursue them with incentives that are aligned to those. And so we see that brands, you know, shift strategy between different seasons or two, Dom's point around different events. And our point is not to be prescriptive and say, oh, this is what you should do. Instead, you know, we're giving them the full menu of what they could do with incentives and it's within their control to decide what do they want to pursue.

Mark Lotman: And so that is something that is always going to be left with kind of the brand as the ultimate decision maker, kind of the overall strategy and kind of inputs that they're able to convey to the model and to us. And make decisions based on that.

Wiehan  Britz: Got it? Got it. Love this. Tell me also. So let me go for the sustainable model, because we want to test it out. How do you then push that back into the email output? So if there's a flow, email, campaign, email, a paid ad, how do you then push that out? So klaviyo communicates with your model. How, how do we go about doing that?

Noam Szpiro: Yeah, sure. So I can show you how that works on our side. Let me share my screen again.

Noam Szpiro: Typically, the way we do this is by doing a web hook in the flow that basically informs us a user has entered this flow. So let me go through an example here. So if we click on this. So this is our webhook that tells us someone has entered this flow. And in the background, what we do is we look up this user and see what is the discount, if any, that this user should receive. So this webhook just has information a little bit about type of user, what's the name of the flow, things like that, once they go through it. So we can preview it here, I can click send webhook, and then we can go to the user's profile. And then in the background, we basically generated this coupon code.

Noam Szpiro: And then now what we can do is we can use that tag we assigned on the user in the email itself, basically. So if before, maybe you said, welcome ten is the coupon code. Now what we can do is something like this. So we can write use code person monoclecoupon code for the discount description, and then that would show up. So we can preview it. And then this is the guy we sent the webhook for. And then you can see this is the coupon that we just generated for them. And then likewise, we can just hide this section.

Noam Szpiro: If the user did not receive a discount, I think this one received the discount. Let's see if we can find one that did not. There we go. So that person didn't get a discount. So it just disappears.

Wiehan  Britz: Very interesting. I think what I like about this most is I know we've briefly mentioned abstaining up all these complicated split paths, so you can test the 5% versus 10%, aggressive versus not aggressive, pushing this over to monocle. Let Monocle do this for you. And you look at the monocle data to find those statistical answers. The uplift, I think that's just brilliant. You also had something.

Before we move on or wrap up, you had in a dashboard gross profit label as well, or option, do you first need to enter your, your unit economics your cogs and so forth into the system? Or do you guys not dabble in the space of, help me figure out my margins here. What does that look like? Does it still need to come from the business head? Correct.

Noam Szpiro: Yeah. So the way we can do this, we can build a machine learning model that is optimized on revenue, or we can optimize it based on gross profit. So it's completely up to the brand to decide what is the metric they want to optimize on. If they want to optimize on gross profit, we just need cogs information or any information that they, any formula that they use to calculate the gross profit. So if it's already inputted into shopify, we can grab that automatically, basically. And then we can start basically producing predictions for how it will affect gross margins, essentially. So, yeah, we can, we can do that.

Wiehan  Britz: Is it based off of percentages? A gross profit margin of 60%. And you guys work out the ideal, ideal discount distribution to hit those margins. So you do the smart calculation around if we want to hit 65%. This is what it looks like.

Noam Szpiro: That's pretty good. Yeah.

Wiehan  Britz: Damn, that's smart. It'll never go over. It'll obviously go under or no discounts to margins. I think that's smart because I think a lot of people are just thumb sucking the numbers. I think one of you guys mentioned that is, I think it's 25%. Cool. Let's just go for that. Where if you can set it to be dynamic in that way and you guys do take care of those calculations, I think that's the brain that we need to use AI for because we are not all economists and data scientists. So I do love that capability. Freaking smart.

Noam Szpiro: Yeah, I agree.

Wiehan  Britz: Very cool. And tell me the you said that. So when you connect a webhook and it has a lookup of a particular person, is that just looking at your Shopify data plugged into your platform?

Noam Szpiro: Yeah. So we have a Shopify app that basically understands what the user has done on the site. So we grab data from there. If there's third party tools that the company uses to collect data, we can plug into those and then to just supplement the data. But yes, typically it's all coming from Shopify directly.

Wiehan  Britz: Super cool. Okay, before I let you guys go, what are you guys seeing for some of the people that have used the platform in terms of revenue, uplift, profits and so on?

Noam Szpiro: Yeah.

Mark Lotman: So as Noam said, we're working with brands already at scale brands, mostly in North America. We started from email and sms and expanding onto on site promotions and offers. And that essentially the way to look at us is we're kind of a unified engine across both in terms of results, we're saying, and there's a number of case studies on our site, we're seeing a pretty consistent lift of between 30% to 35% on gross margin.

That is comparing us to a previous discount policy. So let's take an example. We have brands that had this fast and hard rules around, if you are in checkout abandonment, then you will get 15%. Well, if you're in cart abandonment, you will get 10%. So we do usually is after we develop the model, we compare ourselves to those discount policies.

We do a split A/B test, and we find that a lot of the times, there's a lot of people that are just getting into the wrong flow, if you will, in terms of the discount they're getting. And that dictates a lot of the outcomes. And so we really are able to kind of improve results on that. Brands that ask us to open longer time periods, we're also able to kind of drive results with a longer LTV.

So if, you know, if you're a subscription brand, we're also able to kind of drive increase in LTV over a longer time period. It obviously takes us a little bit longer to also measure those effects. I'd say that kind of the biggest brand so far that we've done case studies with, you know, honey, love, death wish, coffee, TPJ are all kind of national us brands and are all within that 30, 35% bucket and gross, gross profit from, from the flows that we optimize.

Wiehan  Britz: Got it. Nice. I love that last question from me. Then we can wrap up. You do say you play well with subscription services platforms. Recharge Kio, all those platforms and loyalty programs. Do you guys pay well with any loyalty programs giving incentives or how does, how does that work?

Mark Lotman: Yeah, so, I mean, we haven't integrated into loyalty program yet, but we are talking to folks in that area and that is something that is a natural extension.

Wiehan  Britz: So that's still on the roadmap. Okay, smart. Especially if you guys can also take over that loyalty side of things, because I feel like loyalty is also such a static thing in itself. If someone hits a certain tier, give them 30% off. Like why? Exactly. Okay, love this. We can wrap up. How can someone get in touch with you guys? A demo free trial.

Wiehan  Britz: That looked like as a next step.

Mark Lotman: Yeah. So our site has a book demo section. I think that's the easiest way to get in touch. Otherwise, send us an email and yeah, excited to chat to people.

Wiehan  Britz: Love it. Good session, guys. I love this power of AI. It's not going to take over the human base and our jobs. We need to embrace these people. Thanks, guys. Thanks once again for joining me and it was a great session. I think there's a lot bigger we can and yeah, I'll check you guys soon.

Noam Szpiro: Thanks, man. Thanks.


Enjoyed this conversation? Follow Wiehan Britz on LinkedIn and don't forget to join 5,000+ hungry D2C enthusiasts who lap up our weekly insider insights on eCommerce email marketing.

Related Reads Curated for You

Ready to unlock Profitable Growth?

Get a free 30-minute consultation with a senior eCommerce expert.
No obligation to sign up - come prepared with questions.

Monster head