Latest / Inside Commerce: Ecommerce Strategy, CX and Technology Podcast / EP178: Driving More Value From Your Cart to Confirmation Page Journey, With Rokt Chief Product & Engineering Officer Bill Barton
Transcript
- 0:03Thanks for tuning in 3, part 4 part cast sponsored by
- 0:05attracting hypersonics and hosted by myself.
- 0:07James, good warm, welcome. If it's your first time we've
- 0:10put a lot of effort into recording this content of breeds
- 0:12you could free help us keep it that way refers to other people,
- 0:15let them know we exist, share it repost on LinkedIn, and social
- 0:19and we put our post and also, can you rate the show on Apple
- 0:23spots by YouTube or wherever you're listening?
- 0:25Really helps boost our visibility because of that warm
- 0:27glow inside. So thanks and we hope you enjoy
- 0:29the The episode. So let me set this up and then
- 0:32I'll shoot you I guess so I'll Focus today is on driving more
- 0:35value from your carts formation page Journey with rot.
- 0:38So having recently talked about profit optimization of
- 0:40e-commerce with hypersonics, we're drilling down into Tech
- 0:42solution that helps key Commerce teams.
- 0:44Had proved monetization of the on-site conversion funnel and
- 0:47our focus is with the team at Rock today.
- 0:49So we're going to cover things. I the problems of rocks is
- 0:51helping e-commerce business is solved.
- 0:53How the technology has been used by retailers and key product
- 0:56features and business Toulon. So all welcome to build Baltic
- 0:58is a cheap product and engineering.
- 1:00Near officer, I believe you I'm doing fine.
- 1:03This morning is good to be here with you.
- 1:05Yeah, thanks so much for jumping on.
- 1:07I'm really looking forward to this as well, to learn a bit
- 1:09more about the bright. I know you need up the R&D
- 1:12efforts, but would you like to give our listeners a bit of a
- 1:14favor? For what, who you are, what
- 1:16you're doing more product, is after, as I said, I need a
- 1:20product development. So that product management
- 1:22software, engineering machine learning science data,
- 1:25engineering, all of the, all the disciplines that go into
- 1:28allowing us to Build build and operate our service.
- 1:32I joined rocked about a year ago prior to that.
- 1:35I had spent 10 years at Amazon working on Alexa, but have been
- 1:40having a great time at rocked. Rocked is enjoying tremendous
- 1:45tremendous growth of. This is just a great place to
- 1:47be. I've interest do you still have
- 1:50Alexa on everywhere doing things for you or of you reach the end
- 1:54of 10 years where you no longer words are here at all?
- 1:56Got that I used to have about 20 devices just all over the place.
- 2:01It was it was all the wax all the time for me now that I'm out
- 2:06in the wild, I have a couple we have one in her kitchen.
- 2:11I use it mostly to ask for when the Boston Bruins will play
- 2:15their next. I saw Kira or what the Or the
- 2:19game was and then I have one that I use to control the fans
- 2:23in my home gym. I'm not.
- 2:25So basically, I live and breathe over 20 devices.
- 2:27That's like the Christian Bale method acting approach to
- 2:30certain that that we really embrace it.
- 2:34Yeah, well, you know, when you're working on a product like
- 2:38that, there's nothing like eating your own dog, food as the
- 2:42expression goes and and I ate an awful lot of Alexa dog food
- 2:46eggs. Well, let's let's let's get into
- 2:49The Rock because I think it's quite interesting technology.
- 2:51Platform another. I'm looking for to let a bit
- 2:53more and I think it'd be really useful for our core e-commerce
- 2:56based understand where it fits into the overall technology.
- 2:59You can system. The let's start with, you know,
- 3:01the focus once is you'd be his on the cops confirmation page.
- 3:04Purchase funnel in you walk us through like the most common
- 3:08user Journeys, that Roxas is part of on an e-commerce store
- 3:11and what roles you play each of those steps.
- 3:14Yeah, sure thing. So as he said, right focuses on
- 3:17what we call the transaction. Moment, which is really the user
- 3:20Journey during the checkout experience.
- 3:22On an e-commerce site. We don't try and solve problems
- 3:27around product selection and endemic advertising, and there's
- 3:30the more top of funnel experiences.
- 3:32There are people who are great. At that most notably, the
- 3:34e-commerce partners are often often.
- 3:37Very good at that. So we help our e-commerce
- 3:39Partners, unlock additional value during the checkout
- 3:43experience and I can get, I'll give you a typical user user
- 3:46Journey as you may think. Up.
- 3:49I'm a big fan of ice hockey, the Boston Bruins there in the NHL
- 3:53playoffs. And so let's say, I went to
- 3:56Ticketmaster and and I was looking to buy tickets for the
- 3:59next Boston. Bruins, playoff game, I go, I
- 4:02find my tickets. I put my tickets in the car and
- 4:06and now I'm ready to actually buy them.
- 4:08So now I get to the first part where rocked really helps a part
- 4:12of such as Ticketmaster unlock new value, which is the what we
- 4:15call the cart review page. So you can picture it as a
- 4:17consumer. You've already put Things in the
- 4:19card. You're ready to purchase in
- 4:21this, typically, a page where you're just looking one last
- 4:24time and what's in the car to make sure you actually want to
- 4:26buy it on that car review page. It's a great opportunity to help
- 4:31our partners. Unlock new value by presenting
- 4:34up cells, either up cells from their their own offering or what
- 4:38we call cross sells up cells from from third-party.
- 4:41So for example, on a Ticketmaster by have those
- 4:44Bruins tickets in there, Ticketmaster might have The
- 4:49opportunity for me to add parking near the venue.
- 4:52You can sometimes by VIP experiences at the venue, you
- 4:56can picture a third party offers to buy Boston Bruins Jersey, so
- 5:01that the family can show up with all their fan and Logo.
- 5:04We're on. So that card review page is is
- 5:07the first step of the user Journey where rocks can help a
- 5:10partner adding value to the card directly.
- 5:15What once you're happy with the cart possibly including some of
- 5:18these Rock to mediated upsells and you hit that.
- 5:21Yep. I want to buy it button.
- 5:23Then you get to the payment page and we're all familiar with
- 5:27that. You either have to enter a
- 5:28credit card or perhaps. If you're logged into the
- 5:31e-commerce partner, you might have one stored but it's where
- 5:34you're choosing. How you're going to pay for that
- 5:36cart? That payment page is
- 5:39traditionally a cost center for e-commerce Partners.
- 5:41You know, they're paying basis points to the credit card
- 5:44providers. And so usually it's a place
- 5:47where their margins are rewrote. A little bit with rock, served
- 5:50offers from repayments Marketplace, we can present
- 5:53offers for new forms of payment that the customer doesn't have a
- 5:57credit card to buy. Now pay later offer and we're
- 6:01able to do it in a way which does not affect conversion rate.
- 6:04We can dig into that later in the conversation and it's just a
- 6:07great opportunity to connect customers with new forms of
- 6:11payment at a time when that's very much on their mind and and
- 6:14doing. So, we unlock additional value
- 6:17for the, for the E-commerce partner Ticketmaster and I
- 6:20example then once you purchased you get to the confirmation page
- 6:25typically says, oh thanks for your purchase.
- 6:28That confirmation page is another great opportunity to put
- 6:31offers in front of customers so it might say something like oh
- 6:35thanks for your purchase of the of the Bruins tickets.
- 6:38Your purchase as unlocked, three, free months of Hulu, or
- 6:41those types of those types of offers.
- 6:44And we're able to deliver yet more value to the e-commerce
- 6:47partner and Every step of the journey, we're presenting
- 6:50things, which are relevant to the customer.
- 6:52So the customer engagement with the pages as high and and their
- 6:58engagement with the offers is extremely high attract, you'll
- 7:03Complete product Discovery, growth engine, create relevant
- 7:07shopping experiences that convert into sales and grow
- 7:09online Revenue with personalized, search
- 7:12merchandising, and recommendation Solutions powered
- 7:15by a I find out more at a tract.com I find that really
- 7:21interesting about the payments page because most e-commerce
- 7:23teams I've ever worked with would be terrified about
- 7:26changing anything and that anything other than yet because
- 7:29the traditional view point is get them as quickly through
- 7:32checkout as possible to maximize your conversion rate and then
- 7:35try and bringing props not selling off.
- 7:38And it's interesting because there are, you know, that though
- 7:40called the confirmation page. I think more people are starting
- 7:42to realize is an opportunity page.
- 7:45I'd love you to talk us through about that payments Patron.
- 7:47So, in terms of A to B is number one is, is how do you make?
- 7:53Sure the offers are relevant, what's the engineering and the
- 7:56data piece sits behind that? And be how you helping customers
- 8:00to accurately measure that so that they don't have, like, kind
- 8:04of? Because obviously things may
- 8:06work, but they sometimes don't work as well on particular
- 8:08websites. Yeah, those are all great
- 8:12questions and of course to write the the moment when a customer
- 8:15is about to hit that button and a charge is going to hit a
- 8:17credit card. And the e-commerce partner is
- 8:19going to get paid as a super sensitive part of the
- 8:22transaction journey, and they're extremely protective of that,
- 8:27that moment. So let me talk about that.
- 8:29First, we're able to measure our impact on the cart conversion
- 8:35rate, we're not customers, in fact, end up purchasing the cart
- 8:39and we're able to do it. Experiments and were able to
- 8:42show that we have no statistically significant impact
- 8:47on that car conversion rate while also creating these
- 8:51opportunities for customers to engage, with payment offers and
- 8:54the sunblock new value for the e-commerce partner.
- 8:57So, on the whole e-commerce for the e-commerce partner did.
- 8:59So it's incremental Revenue without a negative impact on
- 9:02their their conversion rate. Maybe let me pause there.
- 9:08Case you have follow-up questions about that conversion
- 9:10rate? Yes, you would you first know, I
- 9:13find it interesting because people are obviously there's a
- 9:16balance between not wanted to impede user Journeys, but also
- 9:19recognizing that people are under a lot of like cost
- 9:22pressures and trade in pressures, and wanting to drive
- 9:24profit. So a lot to know is the bit that
- 9:27I think a lot times people struggle with when testing
- 9:30things is hold out and test where you're able to validate a
- 9:33control cell a sees, nothing versus the test cell that
- 9:36actually sees a rather than just putting it on a measuring.
- 9:40You know, incrementally, how do you say, how do you work?
- 9:44I think this should be a question.
- 9:44A lot of things that have is how do you work?
- 9:46The exactly. Do you do test the testing
- 9:48natively in terms of showing not shown, or do you work?
- 9:52Need the Retailer's from Jungle to have an a/b testing back for
- 9:55like an optimizely or a be tasty to, to use and a, we Rockin beds
- 10:03A/B Testing deeply with it within a platform.
- 10:05So we Need our partners to do it for us, we can do it for the
- 10:08partners. We're always doing some amount
- 10:11of hold out so that we can measure increment ality and and
- 10:15effects on things like car conversion rate on the payment
- 10:17page. It's it's an integral part of
- 10:20the rock rock platform and so we're able to measure with rigor
- 10:27are our impact on things like car conversion rate.
- 10:30And by the way, everything I'm talking about that, we're doing
- 10:32on the payment page extends to the things that we do on the car
- 10:35review page. Age and the and the confirmation
- 10:37page as well. Yeah, I guess the other
- 10:41interesting thing is you can reduce conversion, right?
- 10:44Marginally and increase overall profitability because of a
- 10:48incrementality of the the court sighs.
- 10:52So do you spend a lot of time with clients and measuring
- 10:55things? I repeat, purchase rates of
- 10:56people who've redeemed offers versus those who haven't.
- 10:59Is that something that their clients typically look at?
- 11:01Or is it more measured on that heads to instant transaction?
- 11:05Oh, no. We we can look at things that
- 11:08affect lifetime value across, you know, repeat visitors.
- 11:11We certainly do that. And that the goal of the rock
- 11:13Marketplace is to create lasting incremental value for our
- 11:19e-commerce partners and to create engaging experiences for
- 11:22customers. So if there's anything any, any
- 11:26offer or content that we're putting in front of customers,
- 11:29it ends up having a negative impact, not just in the moment.
- 11:33So, even not, not just on whether they can.
- 11:34On that card or not. But whether you know their
- 11:36future repeat raids were extremely obsessed with not
- 11:42having any of those negative impacts and you mentioned you
- 11:44can offset a negative impact. If you're adding enough
- 11:47incremental value with the, for example payment offers on a
- 11:51payment page but we try and set a higher bar than that.
- 11:54So we're not just trying to be net positive.
- 11:57We're actually trying to make sure we don't affect the
- 12:00conversion rate and add incremental value on top of
- 12:03that. And let's do a couple of
- 12:07questions on the integration because it links quite naturally
- 12:10is what what is the integration of Carbonite?
- 12:12My some should be this is like a client-side JavaScript or do you
- 12:15do a pi server side Integrations for people depending on what the
- 12:20technical capabilities of the team or there.
- 12:23We have, we have a really broad array of Integrations.
- 12:27That are our e-commerce Partners on the supply side, and our
- 12:31advertisers on the demand side of the platform can use.
- 12:34So there We have a web SDK focused on JavaScript we have
- 12:39mobile sdks for Native iOS native Android blutter.
- 12:44React, native Cordoba. So we have a pretty wide range
- 12:48of those sdks and we also have direct apis.
- 12:52If at the clients and either side want to be calling to an
- 12:55API directly. We can do that as well.
- 12:58Often we have clients on one of the other side of the platform
- 13:02who use a mix. They might be using an s.
- 13:05IDK and tags and some of their web pages.
- 13:07They might be using an API to report certain types of events
- 13:12Okay, and you say again, another assumption would be that it's
- 13:18eCommerce platform, agnostic. So it's normal.
- 13:20It's not restricted to specific e-commerce platforms.
- 13:22Would that be a correct assumption or G?
- 13:24Work more naturally with specific vendors.
- 13:27One sales forces. For example I'd said we're
- 13:32platform agnostic. So we can we can work with the
- 13:36the partners with any other types of platforms.
- 13:38So it can be a completely proprietary.
- 13:42Act at the partner is built. If they're working with
- 13:43Salesforce, they can still integrate to us.
- 13:47We provide a nearly, no code integration path for sellers on
- 13:53Shopify, for example. So we, you know, our goal is to
- 13:58have as vibrant a Marketplace as possible whether with as many
- 14:02partners on the supply side, as many advertisers and brands on
- 14:05the demand side, it's that robustness and richness of the
- 14:09two-sided Market Place, which really is part of the value.
- 14:12Value that rocked brings. So we're always striving to make
- 14:16it easier for clients to onboard on both sides and just quickly
- 14:20come back to the the holdout test because that's I think
- 14:23that's a really important thing for people to understand is
- 14:28within that, are you able to exclude, or include specific
- 14:33user types of customer segments based on looking, you know, the
- 14:36data in the data layers over to be past you.
- 14:38So, I know that some premium Brands, I work with they have a
- 14:42Specific extra VIP segment who they would never offer any
- 14:45third-party promotions to but to their kind of aspiration
- 14:48audience and actually, yes, they would because I want to monetize
- 14:51thing. Yes, the answer is sort of.
- 14:55Yes, yes, and yes, depending how you split up.
- 14:57Be sure. You said a question is not part
- 14:59of. It isn't just about hold out
- 15:00some of what you're getting to. Now, it is targeting so Brands
- 15:05and advertisers can control audiences, that they'll reach
- 15:09with the, with their campaigns. They can do that via
- 15:14demographics. They can Target it to categories
- 15:18of e-commerce site. The, they have a lot of fun.
- 15:21Green control if they want it. So we're not just spraying and
- 15:24advertisers campaigns all over the supply side of the market
- 15:27place. The, The Advertiser has control
- 15:29and that control can include who they want to reach and who they
- 15:32don't want to reach in. A particular form of that is
- 15:36advertisers who have already acquired a customer.
- 15:42Can choose to Target everyone who's not already a customer.
- 15:45So they're not going to be spending money, trying to reach
- 15:48people who they already have his customers.
- 15:50If their campaign is a customer acquisition campaign.
- 15:53Hypersonics helps e-commerce companies make more profit every
- 15:57day. This AI driven platform delivers
- 16:00recommendations for pricing and inventory that lead to bigger
- 16:04profits. Visit hypersonics, dot AI
- 16:08forward, slash podcast to get a free trial.
- 16:12I'm from the merchants point of view, if it led, the merchants
- 16:15that there's putting these offers and promotions or up
- 16:18sells on their on their part Journey.
- 16:22What data points can you ingest? In order for them to control
- 16:27what and when and who? So let's say that example, by
- 16:31premium brand, that that's got a specific segment for people that
- 16:35it never wants to show this to where, you know, what data and
- 16:39gracious you typically have like yeah, you take in you.
- 16:42Take your stuff directly on the dates there.
- 16:44Are you integrating with crms I, how do you normally receive that
- 16:47information so you can control, you know, the display of
- 16:50promotions? Yeah.
- 16:52So, both for the e-commerce, partner is on the supply side
- 16:55and the advertisers on the demand side via, whether it's
- 16:58the sdks and or any p, is that the using to connect to the
- 17:01right Marketplace? They can provide us a lot of
- 17:05what Ferg each of them is first-party data.
- 17:07So the e-commerce partner can be provided as a lot of first-party
- 17:10data about that customer. The Advertiser can be providing
- 17:13us with their with their campaigns.
- 17:15On the demand side first-party data, they have about their
- 17:19customers, an essential part of the rock Marketplace is that
- 17:22rocked operates as The Trusted intermediary between the two.
- 17:25So we never cross pollinate, we're not sharing data that an
- 17:30e-commerce partner gives us about their customer with the
- 17:33advertiser. We're not sharing Advertiser
- 17:35data about their customer with the partner.
- 17:37Rocked in the middle is using all those data points to allow
- 17:40us to Respect targeting and and and suppression that the clients
- 17:47may be configuring. And then also to optimize
- 17:50click-through rates and engagement and outcomes, but an
- 17:53essential part of the Rock Bottom, again, is where it's
- 17:58essential to us to be that trusted intermediary.
- 18:01We're not sharing data across, we're not doing spooky
- 18:05retargeting, we're not building customer profiles that we used
- 18:09for any other purpose. Other than the Great.
- 18:13The most engaging opportunity. For customers of those eCommerce
- 18:16sites that I go across the red multi stole from because one of
- 18:20the challenges, especially in South Swell, where people have
- 18:22I've stalled, third parties and they got multiple storefronts is
- 18:26not having needed. Multiple accounts.
- 18:28One for each store from. I'd love to.
- 18:31I love to understand how you were her Rock.
- 18:33Would work with any Commerce as many e-commerce businesses,
- 18:36maybe while the UK site. A US site, any, you site and
- 18:40ones to be able to control the Target in.
- 18:42Down to like country, Regional language level.
- 18:47Yeah, well but we're currently operating 15 countries, five
- 18:50different languages. So it's very much a real issue
- 18:53for us. And it's there's no magic to it.
- 18:57We were able to Target campaigns and and match based on the
- 19:02country in which the e-commerce transaction is occurring.
- 19:04The language in which it's occurring.
- 19:08And so we it it's a native part of our platform to be able to
- 19:11ensure the bird. The right content is delivered
- 19:13to the right people and that includes country.
- 19:16Free language, in addition to other targeting criteria that we
- 19:20may have around demographics and act.
- 19:24Like I said, category of the e-commerce partner.
- 19:26So the multi-store, grunt the country in the language are
- 19:29really just a couple of Dimensions or what's a broad set
- 19:32of Dimensions across, which we can be doing targeting and
- 19:34matching, But is it will access by one like Central dashboard,
- 19:38then rather because on some some apps and I people have to have
- 19:41separate instances for each of the store fronts which whilst
- 19:44practical add whilst was it works ads like operational
- 19:47overhead. So is is are all the all the
- 19:49country talking. There's all just through a
- 19:51single account and single sign-on then.
- 19:54Oh yeah, yeah. We send it so pretty.
- 19:56Each Advertiser, for example, we can manage their campuses
- 19:59worldwide for each for each partner, we can be managing
- 20:03their inventory worldwide. They don't have to log into, you
- 20:08know, different instances of the rock management platform.
- 20:11It's we want our worldwide Marketplace.
- 20:15So I need you talk briefly about the 5:00 deadline, advertisers
- 20:18can control the where they removed tree is display because
- 20:22obviously, they don't want to display.
- 20:23Inappropriately or Target people that are relevant.
- 20:26If we flip that to the merch, what controls do you give
- 20:30Merchants to be able to, I decide what promotions and
- 20:34offers a relevant or what partners are willing to have,
- 20:37you know what types of business that they don't want to be
- 20:39associated with key talk people through what's automated versus
- 20:43will? They can configure and control
- 20:45themselves. Yeah it is.
- 20:49It's a very symmetrical yo balanced Market put so
- 20:52advertisers on the demand side. Can come you know if they don't
- 20:55want their brand advertise on a certain category of e-commerce
- 20:59experience. For example they can configure
- 21:01that and similarly on the e-commerce partner side.
- 21:04If they don't want certain brands advertising on their
- 21:07inventory, they can control that you know often naturally an
- 21:11e-commerce partner in say food, delivery is Is not going to want
- 21:17to have ads from other people competing in food delivery so
- 21:20they can say, okay no, no, no food delivery campaigns on my
- 21:24food, delivery inventory. So both sides are able to do
- 21:28that. Now naturally the less a client
- 21:33and other side of the marketplace is narrowing, the
- 21:37inventory of the other side that they could be masked.
- 21:40With the more opportunity is that this probably more value we
- 21:42can mark but we can work with our clients and you know some of
- 21:46A wide-open, others have have very narrow and specific
- 21:49requirements, and our platform enables both.
- 21:52Yeah, it's another point that I found quite interesting is that
- 21:56the products designed to assess a customer's likely to gauge?
- 21:59I think you alluded to earlier so it's not just like peppered.
- 22:02Everybody with offers. I love to know this because
- 22:05things are you know like due to engage I think sometimes people
- 22:10can find well that sounds quite nebula.
- 22:11So how is it? How is it calculated?
- 22:13What data points are used and what signals influence?
- 22:16I can you talk people through that.
- 22:17It's a sure thing that the starting point is the e-commerce
- 22:21partner themselves. So they as I said, can provide
- 22:25us a lot of information about their customer and that and at
- 22:28that moment, we were choosing what content to put in front of
- 22:31their customer in their checkout experience.
- 22:34We're really a solution provider to the e-commerce partner and we
- 22:37didn't react in their interest and again we're a trusted
- 22:40intermediary. So we can leverage a lot of the
- 22:42information about their customer to help predict over time their
- 22:49likelihood to engage with certain types of content.
- 22:53We can all and then we can also use demographic data.
- 22:56So like oh, this customer is is a male in their 40s, males in
- 23:01their 40s or more or less likely to engage with content from from
- 23:05these advertisers and then the advertisers themselves, giving
- 23:08us data which which helps us understand which customers are
- 23:12going to interact with their campaigns.
- 23:14And then, and in the middle, we're able to build this this,
- 23:17this model that predicts the likelihood of each individual
- 23:20customer to interact with a given campaign.
- 23:22Ain't but again, one thing we don't do is cross client
- 23:28boundaries. So for example, if we know from
- 23:31my interaction on Ticketmaster that I'm a hockey fan and
- 23:35there's an Advertiser out there, you know, the nhl.com logo where
- 23:40shop that they can Target they can reach me.
- 23:44If they want to Target hockey fans on Ticketmaster because
- 23:47rocks knows, I'm a hockey fan from a interactions on
- 23:50Ticketmaster but they can't Target me as a hockey fan on
- 23:53some different provider. If the only way that rock to
- 23:57knows I'm a hockey fan is from Ticketmaster so we're very
- 24:01careful to maintain firewalls and again we're not we're not
- 24:05doing some of the things which some people find spooky were on
- 24:09the right side of GDP are and apple tracking transparency.
- 24:14So we're able to leverage a lot of data but we but we do it in a
- 24:18pretty high hygiene privacy respectful way.
- 24:22Okay, cool interesting. I'm how important is historical
- 24:26data for your algorithms because you've obviously got the the
- 24:29learning I swear it, it's helping ottoman.
- 24:31Make smarter decisions based on all those those data points
- 24:35coming from browsing and purchasing and enzyme profiles,
- 24:39but some sites like a new side that's launched.
- 24:41It hasn't got any historical data.
- 24:43But how would it, how would it work in?
- 24:45Atkins is all do people typically need a period of time
- 24:47to the algorithms to be able to glean any data for.
- 24:50They can effectively Don't make these promotion decisions.
- 24:53Yeah, we can we do describing is what we call the called Stirrup
- 24:57problem and so harassed. Yes, it's Oracle data is
- 25:02extremely powerful and with modern machine, learning
- 25:05modeling, the ability or for deep learning models to improve
- 25:11their Performance Based on the more and more data, you thrown
- 25:13to the hopper, the top, including historical data, the
- 25:17better, they perform. So we are able to leverage that
- 25:20we're, we're doing billion. Transactions a year.
- 25:23So we have a lot of data that our models can learn from now
- 25:26with a new partner on the supply side, or a new Advertiser on the
- 25:31demand side and sometimes it's even an existing Advertiser but
- 25:34it's a new campaign there. Is that cold start problem.
- 25:37You don't initially have data about that specific campaign, or
- 25:41maybe it's a new customer. You're seeing through a new
- 25:44partner, and then we haven't seen the customer before, but
- 25:47we're able to use the synthetic twin techniques and look alike.
- 25:51Audiences. That sort of give that give the
- 25:55models, a good brush start in those situations.
- 26:00So in a way, you're leveraging, the historical data and the
- 26:03billions of transactions that we have to get a good starting
- 26:07point for a model of the customer and their propensity to
- 26:10engage with content. And the models learn pretty
- 26:12quickly from there. Once once things are actually
- 26:14life, I know you said you know building up your own date
- 26:18profile. The custody of basically
- 26:20aggravating all the data points and coming.
- 26:22I've taught other Merchants. Do you do within the
- 26:25permissions? It says, what you can do?
- 26:27Do you have like, cross stitching based at like it an ID
- 26:29level like a, an email, a user's email, or you just not touching
- 26:33pii. Then we do have internal the
- 26:37rock, the capacitor or data platform.
- 26:38So we are able to accumulate information about the customer
- 26:46via their interactions with rock serve content.
- 26:49It's, so it's our data. It's our view of the customer So
- 26:52the our e-commerce Partners obviously have their view of the
- 26:55customer, advertisers have to do their view of the customer, and
- 26:59we have a view of the customer so we can begin to learn what
- 27:02types of content a given customer engages with, but we
- 27:07don't do at me. Maybe the example, I gave.
- 27:11Like, we wouldn't allow a sportswear Advertiser to reach
- 27:16meet Bill. Bargain as a hockey fan. on an
- 27:21e-commerce site where the only reason we know on the cocky fan
- 27:25is from my interaction on Ticketmaster, So we have a lot
- 27:30of data but we do maintain firewalls.
- 27:33Yeah. Understood that make sense.
- 27:36You talked about the testing and I had hold out test which is
- 27:38interesting but can you talk? So the other A/B test options
- 27:42that you've got native, like what can people do is it about
- 27:45testing different advertisers strategies?
- 27:47Is it about creative testing on what what can people change?
- 27:52I guess in order to get you know to push kpi's like conversion of
- 27:55Revenue. Yeah, we we can run A/B tests on
- 28:00just about any Dimension or or a slice of the of the campaign and
- 28:05the targeting that we or our. Our clients one, certainly
- 28:11things are like creative text which, which phrasing of a
- 28:16credit card offer is going to be most appealing and likely to
- 28:19attract the most engagement. Should you talk about the low
- 28:22interest rate? Should you talk about other
- 28:24benefits of the card? Should you talk about?
- 28:27You know, balance transfer benefits and so on.
- 28:29So we're able to do a/b testing at that level but much for us.
- 28:34The magic isn't the ability to do a/b testing and say, oh, this
- 28:39is the better freezing for that credit card for for everything,
- 28:43what we really want to be doing is leveraging that defeat will
- 28:46be called the efficient Frontier, which is the ability
- 28:48of the models to learn that or a customer like me, you know, a
- 28:54certain phrasing is going to be better for a different customer.
- 28:56Summer. Some other phrasing is going to
- 28:58be better and making that decision on a per transaction
- 29:01per customer bases, which is what in the field, we call the
- 29:04efficient Frontier. So A/B Testing does help people
- 29:09get higher performance relative to not exploring between
- 29:13options, a b or c, but it's using those experiments to feed
- 29:18the model. So that now the models get it
- 29:20right per transaction. That's where we really get great
- 29:25lat. That's really interesting.
- 29:26She said you can basically I guess you can have a an offer
- 29:30that's modularized then and the how its presented varies from
- 29:35user to user based on the profile.
- 29:38That's a yes. So in maybe experiments have
- 29:42their place, we certainly do them.
- 29:44But to me, what they are is they're there are seed for how
- 29:48to structure the variance in the in the say The Advertiser
- 29:52content. Then our system is like oh we'll
- 29:54combine that font that color that Raising show an image
- 29:58showing the image small for the image medium.
- 30:01And so we can synthesize what amounts to a customized ad unit
- 30:05for a specific customer and a specific transaction from a
- 30:09relatively small number of variations that the advertiser
- 30:13provides and how they describe the possibilities for that ad
- 30:17unit. Yeah, that's nice for not having
- 30:20done some testing on things like remarketing Ansari marketing
- 30:25across territories and seeing how even just just subtle
- 30:27variance is an image can have an impact in response and
- 30:31engagement rates. Yeah, that ability to learn and
- 30:33also Pia is very useful interested.
- 30:36And then, and a key part of it, as a, we aren't just getting a
- 30:40static image. We're not, we're not just doing
- 30:43a display ad. That has some sort of static
- 30:45image where we have no insight into what the actual content is
- 30:49year with a click-through URL. We're getting structured data
- 30:54And then we're synthesizing what the overall experience is for
- 30:58the customer on the partner site.
- 30:59So that's part of what allows us to do that.
- 31:02Manifold synthesis and optimization on a per customer
- 31:05basis. Yeah.
- 31:07That's really interested. And Uncle one final question
- 31:11about the, the advertising Network where you got third
- 31:14party apis, you can be displayed in two.
- 31:16So you go, Ticketmaster is example.
- 31:19So how does it work for a merchant?
- 31:22You've got your own Marketplace. Obviously, people connect to in
- 31:24the advertisers and the merchants can select inside who
- 31:27does and doesn't appear. But how does it work for a
- 31:29merchant who's got? Potentially I to Partners they
- 31:32have commercial agreements with and they want their offers
- 31:35feature but they're not currently in your network will
- 31:37see what's a vetting process. Are you bring, can they bring
- 31:40anyone to the table or how does it work?
- 31:44Yeah, we can, we can work across the full spectrum of that.
- 31:46So some some Merchants come to us and they have zero
- 31:49relationship, the advertisers and that and that's part of the
- 31:52value that the right Marketplace Rings.
- 31:54There are, there are other partners who might have
- 31:56pre-existing relationships with some advertisers.
- 31:59Now, we have thousands of advertisers that we work with.
- 32:04So often a part of the chromosomes is oh well here's a
- 32:07preferred Advertiser like. Well that's great.
- 32:08They're already part of our Network and we can work with
- 32:11them on that. If it's if it's a new Ooh,
- 32:13Advertiser. Then we can pretty quickly on
- 32:15board that Advertiser on the demand side of our platform.
- 32:19Just like all the other advertiser's and and once that's
- 32:22done now, now the marketplace, just just does its thing and
- 32:27that absolutes, say that Advertiser for examples, like I
- 32:30only want to advertise on that partner of brought me to the
- 32:33Rocks table so to speak. Well, all of our existing
- 32:37Marketplace controls can allow that sort of narrow one-to-one
- 32:40matching of that's what the clients want.
- 32:43Thanks for thanks for. That's that's all the questions
- 32:45I wanted to run through. If anybody wants to probe
- 32:47further or is got a liar, use case, I want to run past you,
- 32:51how do they reach out? You know, how do they find more
- 32:54info but will anyone can go to rock.com, we have lots of
- 33:00information there. We have integration guides, we
- 33:04have case studies and and learn more, plenty of learn more
- 33:10opportunities where people can click can provide there,
- 33:12Information and our business development and customer
- 33:15success, teams would be happy to hear from people who are
- 33:18interested in learning more about the rock solution.
- 33:21So well, thanks hit solid, build, much.
- 33:23Appreciate it. I hope you enjoy the rest of
- 33:25your day. And thanks everyone for listing.
- 33:28Always appreciate it. Keep your ears open for next
- 33:31episode we drop, but every Tuesday and do rims.
- 33:33Give us a trait in wherever you're listening and watching.
- 33:35It really does help others ability.
- 33:37Thanks very much everybody. More information on this topic,
- 33:47head over to re-platform dot. F m, for our audio podcasts to
- 33:50discuss a project or if you'd like to chat about any of the
- 33:53topics covered in this episode in more detail.
- 33:55Please reach out to myself, James guard.
- 33:57And my co-host put Rogers, via LinkedIn and Twitter.
- 33:59Thanks again for listening and keep your ears peeled for the
- 34:01next episode.