3026: Boosting Conversion Rates with Predictive Delivery Promises

[00:00:03] [SPEAKER_01]: How is AI revolutionising the way we shop and receive products? What goes on behind

[00:00:10] [SPEAKER_01]: the scenes? Well today I'm going to be diving deep into the world of retail order management

[00:00:16] [SPEAKER_01]: with my guest Chap, who is the Vice President of Product Strategy and Operations at Next

[00:00:23] [SPEAKER_01]: Duple. And with more than 25 years experience in the retail omni-fulfilment domain, Chap

[00:00:29] [SPEAKER_01]: has been at the forefront of transforming how retailers manage orders from giants such

[00:00:34] [SPEAKER_01]: as Best Buy to premium brands like Red Wing Shoe Company and so many more. But at Next

[00:00:41] [SPEAKER_01]: Duple, Chap is leading the charge in redefining omni-channel fulfilment by leveraging advanced

[00:00:47] [SPEAKER_01]: AI and machine learning and microservices architecture. So in this episode today,

[00:00:54] [SPEAKER_01]: we're going to explore how something called predictive promising powered by AI is reshaping

[00:01:01] [SPEAKER_01]: customer experience by accurately forecasting delivery times and reducing manual inefficiencies.

[00:01:07] [SPEAKER_01]: And we'll also discuss how gen AI and dynamic inventory services are enabling retailers

[00:01:13] [SPEAKER_01]: to maintain optimal stock levels and streamlining operations. So whether you are curious

[00:01:20] [SPEAKER_01]: about the impact of AI on retail or just want to understand the future of order management

[00:01:25] [SPEAKER_01]: systems, my guest Insights will offer a much clearer view on some of the innovations

[00:01:31] [SPEAKER_01]: that are already transforming the industry. So what role will AI play in the future

[00:01:36] [SPEAKER_01]: of retail and how will it influence the products we see on store shelves? I want

[00:01:42] [SPEAKER_01]: to take a time out to express my gratitude to everyone who supports my mission of

[00:01:46] [SPEAKER_01]: delivering daily content to you in 165 countries. I couldn't do it without you and I couldn't

[00:01:52] [SPEAKER_01]: do it without my sponsors. So quick shout out to any defence contractors listening out

[00:01:57] [SPEAKER_01]: there. CMMC 2.0 compliance doesn't have to be a headache. Consider Ki-Works your

[00:02:03] [SPEAKER_01]: fast track to authorisation and as a FedRAMP moderate authorised solution, they cover

[00:02:08] [SPEAKER_01]: nearly 90% of CMMC 2.0 level 3 requirements. For you that means less time, less effort,

[00:02:16] [SPEAKER_01]: less cost and while others struggle with DIY approval processes and clunky apps,

[00:02:21] [SPEAKER_01]: you'll breeze through with their zero trust framework. So don't let compliance slow

[00:02:26] [SPEAKER_01]: you down is what I'm trying to say. Simply visit Ki-Works.com to get started.

[00:02:30] [SPEAKER_01]: That's Ki-Works.com to learn more about Ki-Works' secure content platform for

[00:02:36] [SPEAKER_01]: CMMC compliance. But with my thank yous out of the way it's now time to jump right

[00:02:41] [SPEAKER_01]: into today's interview with a fantastic guest.

[00:02:45] [SPEAKER_01]: So a massive warm welcome to the show. Can you tell everyone listening a little

[00:02:50] [SPEAKER_01]: about who you are and what you do?

[00:02:53] [SPEAKER_00]: Absolutely, good morning. My name is Chap Aiken and I've been at Next

[00:02:56] [SPEAKER_00]: Duple for a few years now. Most of my career has been in retail on the

[00:03:02] [SPEAKER_00]: order management side so kind of where digital commerce meets supply

[00:03:05] [SPEAKER_00]: chain and at Next Duple I help our clients get the most out of our

[00:03:10] [SPEAKER_00]: technology and then help our product team take those client pain points

[00:03:14] [SPEAKER_00]: and make sure we're working on the right things in our product.

[00:03:18] [SPEAKER_01]: Well I'm looking forward to learning more about Next Duple a little bit

[00:03:21] [SPEAKER_01]: later in the podcast but to begin with, I always ask my guests to try

[00:03:25] [SPEAKER_01]: and demystify a buzzword, some technology or anything that maybe we

[00:03:30] [SPEAKER_01]: hear on a regular basis but a little bit too afraid to ask exactly what it

[00:03:34] [SPEAKER_01]: means. So today I want to ask you to introduce everyone listening to the

[00:03:38] [SPEAKER_01]: concept of predictive promising. So can you tell me a bit more about that

[00:03:43] [SPEAKER_01]: and how you see it transforming the landscape of retail order management,

[00:03:48] [SPEAKER_01]: particularly in things like improving fulfillment efficiency for example?

[00:03:53] [SPEAKER_00]: Yeah absolutely and luckily predictive promising is actually one of the

[00:03:58] [SPEAKER_00]: easier AI ML models to understand as essentially taking a retailer or

[00:04:06] [SPEAKER_00]: anybody's shipment past, shipment performance, historical data using

[00:04:12] [SPEAKER_00]: that then to turn around and predict what is the estimated delivery date of

[00:04:17] [SPEAKER_00]: a product going to be using that history. So instead of using rules to

[00:04:23] [SPEAKER_00]: decide based on when someone's ordering, what carrier might be, what

[00:04:27] [SPEAKER_00]: distribution center it might ship out of, they use actual historical data

[00:04:33] [SPEAKER_00]: to then come up with a prediction of when something might arrive at

[00:04:37] [SPEAKER_00]: someone's house. Now from an fulfillment efficiency perspective,

[00:04:43] [SPEAKER_00]: this doesn't really make the fulfillment process more efficient,

[00:04:47] [SPEAKER_00]: however it helps retailers make accurate promises which can drive

[00:04:52] [SPEAKER_00]: conversion and competence in the consumer experience and then of course

[00:04:57] [SPEAKER_00]: the retailer side, it simplifies their ability for them to make that promise.

[00:05:05] [SPEAKER_01]: Perfectly explained there and next to Paul of course you're on this mission to

[00:05:09] [SPEAKER_01]: level the playing field for omnichannel retailers. I've got to ask,

[00:05:13] [SPEAKER_01]: there's a lot of hype around AI at the moment, a lot of businesses trying

[00:05:16] [SPEAKER_01]: to understand what does it mean for them and what problems can it solve

[00:05:19] [SPEAKER_01]: and AI and machine learning are integrated into your order management studio.

[00:05:25] [SPEAKER_01]: Can you tell me a little bit more about that and how it ultimately enhances the

[00:05:29] [SPEAKER_01]: accuracy of delivery estimates and customer satisfaction and I suspect

[00:05:33] [SPEAKER_01]: a few other things because it's great to hear the actual value that this

[00:05:36] [SPEAKER_00]: technology can offer. So predictive promising is really exciting

[00:05:41] [SPEAKER_00]: because not only can you predict an outbound promise to a consumer but you

[00:05:47] [SPEAKER_00]: could predict or predict an inbound promise from a supplier into a

[00:05:52] [SPEAKER_00]: distribution center as an example and the way we use AI ML in terms of our

[00:05:57] [SPEAKER_00]: next to board or management studio is certainly someone can toggle that

[00:06:02] [SPEAKER_00]: feature off and on if they decide that they don't have enough confidence

[00:06:07] [SPEAKER_00]: in the data but essentially down at the skew slash vendor level we can

[00:06:16] [SPEAKER_00]: decide do we have enough training data so that we have confidence that

[00:06:21] [SPEAKER_00]: this prediction is going to be accurate and so when a promise is being made

[00:06:26] [SPEAKER_00]: if the retailer has decided you know what at this particular supplier skew

[00:06:33] [SPEAKER_00]: I don't have enough data to make that predictive promise down at that level

[00:06:38] [SPEAKER_00]: you could turn on something or turn off something to allow that promise

[00:06:44] [SPEAKER_00]: to be made to the consumer.

[00:06:47] [SPEAKER_01]: And to bring to life what we're talking about here are there any real world examples you can share

[00:06:53] [SPEAKER_01]: where predictive promising has directly impacted a retailer's bottom line in

[00:06:58] [SPEAKER_01]: operational efficiency and the reason I ask that is I think every tech project is under

[00:07:02] [SPEAKER_01]: close scrutiny now for what ROI is it going to deliver, what business value is it going

[00:07:08] [SPEAKER_01]: to generate for us etc. So it'd be great to hear some real world examples of how

[00:07:12] [SPEAKER_01]: you've done that with this.

[00:07:14] [SPEAKER_00]: Yeah actually this is one of the more straightforward AI use cases I think in

[00:07:19] [SPEAKER_00]: the industry so when you make a predictive promise typically it's actually faster than

[00:07:25] [SPEAKER_00]: what you're promising today because retailers like to put in conservative promises you know

[00:07:31] [SPEAKER_00]: make sure that they deliver against that promise but that typically means a slower promise

[00:07:36] [SPEAKER_00]: so what we've seen an example that we're working with a retailer is using predictive

[00:07:42] [SPEAKER_00]: promising for all of their drop ship suppliers and they're seeing a five and a half day

[00:07:50] [SPEAKER_00]: improvement in the delivery promise that they will be able to make to consumers.

[00:07:55] [SPEAKER_00]: Another retailer that we are doing a POC with that's using stores for fulfillment

[00:08:02] [SPEAKER_00]: is going to recognize a two and a half day improvement in the delivery speed of the

[00:08:07] [SPEAKER_00]: promise that they're making to consumers and that just translates into bottom line

[00:08:12] [SPEAKER_00]: conversion improvement right, the faster you can make a promise the more specific you can

[00:08:16] [SPEAKER_00]: make a promise there's tons of data that supports that drives increased conversion

[00:08:22] [SPEAKER_00]: and then consumer satisfaction.

[00:08:24] [SPEAKER_01]: And I'm curious how would you say your approach to AI driven order promising how would

[00:08:29] [SPEAKER_01]: you say that differs from some of the more traditional methods that are out there and

[00:08:33] [SPEAKER_01]: what are the particular advantages that it might offer to retailers facing today's

[00:08:37] [SPEAKER_01]: market challenges because the pace of change is moving so quickly at the moment we

[00:08:42] [SPEAKER_01]: great understand the old world to the new world.

[00:08:44] [SPEAKER_00]: Yes, the irony with this one is is that what actually prevents a lot of retailers

[00:08:49] [SPEAKER_00]: from making a specific delivery promise today so you know many of them have shipping

[00:08:54] [SPEAKER_00]: windows on their site still today that says you know you can order something and get it in

[00:08:58] [SPEAKER_00]: three to seven days as an example.

[00:09:01] [SPEAKER_00]: The amount of data analysis the configuration and then the ongoing monitoring of all of

[00:09:07] [SPEAKER_00]: the inputs that go into a traditional promising engine which would be you know manual

[00:09:13] [SPEAKER_00]: configuration of data loading of transit tables that takes a lot of effort and

[00:09:20] [SPEAKER_00]: and it of course needs to be maintained and there's tons of variability in that

[00:09:25] [SPEAKER_00]: what is the shipping performance of a node on a Friday afternoon versus a Tuesday morning

[00:09:31] [SPEAKER_00]: for example for a given item and so predictive promising actually eliminates all the setup

[00:09:39] [SPEAKER_00]: required because you're using that historical data.

[00:09:44] [SPEAKER_00]: So its outcome is better as we discussed it's faster and it's more accurate but then

[00:09:51] [SPEAKER_00]: the actually ability to do it is somewhat enabled by the AI model itself because it

[00:09:58] [SPEAKER_00]: allows retailers to get away from all this manual work that traditional promising engines

[00:10:02] [SPEAKER_00]: have involved.

[00:10:04] [SPEAKER_01]: Just to drill a little bit deeper in this are you also able to explain just

[00:10:07] [SPEAKER_01]: how your dynamic inventory service how that uses AI to improve fill rates reduce

[00:10:12] [SPEAKER_01]: inventory related issues for retailers too because this feels like somewhat of a game changer

[00:10:17] [SPEAKER_01]: too.

[00:10:18] [SPEAKER_00]: Yeah absolutely you know the classic issue for omni-channel retailers is finding that

[00:10:23] [SPEAKER_00]: balance between how much store inventory you would show online how much to keep as

[00:10:29] [SPEAKER_00]: safety stock to protect the fill rate the customer experience and then how much to

[00:10:34] [SPEAKER_00]: reserve for walking consumers and our model attempts to balance those keeping the safety

[00:10:40] [SPEAKER_00]: stock as low as possible so to maximize online sales while keeping that fill rate high

[00:10:47] [SPEAKER_00]: and keeping in mind that walking customer and most retailers are just kind of focusing

[00:10:53] [SPEAKER_00]: on getting the safety stock right but in grocery use cases for example with high

[00:10:58] [SPEAKER_00]: skew velocity you want to protect offline sales for that day as well.

[00:11:03] [SPEAKER_00]: So we used forecast data sales history bill rate performance to come up with a

[00:11:09] [SPEAKER_00]: recommendation on that available to promise I would show an online customer how much

[00:11:15] [SPEAKER_00]: to keep for that offline customer and the right safety stock level to protect the fill

[00:11:20] [SPEAKER_00]: rate and in addition we allow the business customer to decide which one of those is

[00:11:27] [SPEAKER_00]: the most important.

[00:11:28] [SPEAKER_01]: And with generative AI also rising in popularity how do you see that foreseeing

[00:11:35] [SPEAKER_01]: its role evolving in retail order management particularly in terms of inventory order

[00:11:39] [SPEAKER_01]: inquiries dynamic safety stock calculations those kind of areas feel like a good fit for it but

[00:11:45] [SPEAKER_01]: what are you seeing here?

[00:11:46] [SPEAKER_00]: We're actually using gen AI right now in our inventory service to help IT and business

[00:11:52] [SPEAKER_00]: teams improve the speed of issue discovery so instead of writing queries to find out

[00:11:58] [SPEAKER_00]: where an inaccuracy in inventory data came from I could interrogate the data using a chat

[00:12:05] [SPEAKER_00]: and that will go you know in natural language tell me where is the issue being caused with

[00:12:11] [SPEAKER_00]: this inventory data.

[00:12:13] [SPEAKER_00]: In the future I imagine an AI model that can proactively find issues with the inventory

[00:12:19] [SPEAKER_00]: data for example if I saw four online inquiries for availability fail because there was a

[00:12:26] [SPEAKER_00]: disconnect between the sales channel of view of inventory and the backend system

[00:12:31] [SPEAKER_00]: I could proactively move that skew to out of stock or proactively send a message to

[00:12:36] [SPEAKER_00]: an operator with possible root causes that they should look into.

[00:12:41] [SPEAKER_00]: The tons of upside I think is as we think about gen AI and order management.

[00:12:46] [SPEAKER_01]: And obviously we're talking about incredibly complex technologies here so what kind

[00:12:50] [SPEAKER_01]: of challenges do retailers face especially when transitioning to an AI driven order

[00:12:55] [SPEAKER_01]: management because I'm sure there'll be a lot of people very apprehensive when

[00:12:59] [SPEAKER_01]: hearing about this and any kind of technological change especially after those tech projects that

[00:13:04] [SPEAKER_01]: always seem to take an age to get implemented and then you've got the culture

[00:13:07] [SPEAKER_01]: change that goes with that as well so can you tell me about those challenges and how

[00:13:11] [SPEAKER_01]: you're helping ease them through that transition as well.

[00:13:16] [SPEAKER_00]: Yeah a lot of people will mention data and of course that is important but I'll focus on

[00:13:21] [SPEAKER_00]: another element which is explainability so when people implement these models they want

[00:13:28] [SPEAKER_00]: the ability to understand how the decision was made by the model and a lot of the past

[00:13:34] [SPEAKER_00]: experiments that we've seen specifically in the order management domain with ML models

[00:13:40] [SPEAKER_00]: is that lack of explainability has caused the retailer to not get through that change curve

[00:13:46] [SPEAKER_00]: because they were unsure of how the system was making a decision.

[00:13:51] [SPEAKER_00]: So at Next Duple we actually use gen AI to provide that explainability that allows

[00:13:58] [SPEAKER_00]: a user to see what attributes caused the decision to be made what was the

[00:14:03] [SPEAKER_00]: the attribute that had the most weight into the decision but in addition to that

[00:14:09] [SPEAKER_00]: before you even get to these models we allow retailers a weighted a flexible way to define

[00:14:17] [SPEAKER_00]: all the costs and the penalties that might go into an AI decision so they make sure that they

[00:14:23] [SPEAKER_00]: have that data correct in the system in the first place and then when they get to that model

[00:14:28] [SPEAKER_00]: they'll have confidence in the data but then that added explainability will allow them

[00:14:33] [SPEAKER_00]: to gain that further confidence to say yes I can see why this decision was made

[00:14:38] [SPEAKER_00]: and then I can go if I don't like that I can go tweak the model to to improve the outcome.

[00:14:45] [SPEAKER_01]: And of course we're talking here in autumn, autumn is here we're already weeks away from

[00:14:50] [SPEAKER_01]: pumpkin spice, black friday emails and all that other crazy stuff so looking ahead what

[00:14:56] [SPEAKER_01]: innovations in AI and ML do you anticipate will have the maybe the most significant impact on

[00:15:02] [SPEAKER_01]: retail order management and indeed customer experience? Yes one the model that we haven't

[00:15:08] [SPEAKER_00]: talked about yet that I'm really excited for is node selection and instead of you know a

[00:15:16] [SPEAKER_00]: of an order management system is to take all these business rules and then select the best

[00:15:21] [SPEAKER_00]: node that meets you know service goals cost goals utilization goals and today that's all done

[00:15:28] [SPEAKER_00]: with rules by business users and tomorrow where I think we're headed is businesses telling the

[00:15:35] [SPEAKER_00]: system these are my actual KPIs I'm trying to get right I want to move from five days to three

[00:15:42] [SPEAKER_00]: and a half days of transit time as an example supplying the inputs to the system and then having

[00:15:48] [SPEAKER_00]: the system make the decision based on trying to optimize against those actual end outcomes

[00:15:56] [SPEAKER_00]: and I think that's going to unlock potential loss savings and network speed improvements

[00:16:01] [SPEAKER_00]: that retailers can't see today in the rules that they have set up so I think that's

[00:16:06] [SPEAKER_00]: tremendously exciting for retailers. And I am conscious we've been fully focused on

[00:16:12] [SPEAKER_01]: the future and where we're going but as we come towards the end of the podcast now I want to find

[00:16:17] [SPEAKER_01]: out a little bit more about you and ask you to look back throughout your career I'm sure you've

[00:16:21] [SPEAKER_01]: picked up a few war stories along the way so is that a funny or interesting story that has

[00:16:27] [SPEAKER_01]: happened to you throughout your career that that we are able to share today? Yes back

[00:16:32] [SPEAKER_00]: when I was at Best Buy I had the fortune of being in a future of work remote work pilot

[00:16:39] [SPEAKER_00]: as a results only work environment that Best Buy was doing and it was a very innovative time this

[00:16:46] [SPEAKER_00]: was back in 2006 this is a very innovative concept at the time of course fast forward we're all

[00:16:51] [SPEAKER_00]: dealing with you know hybrid work remote work but at the time it was quite innovative

[00:16:56] [SPEAKER_00]: and I was actually selected to be interviewed by Leslie Stahl on 60 Minutes about the future

[00:17:04] [SPEAKER_00]: of work in 2006 so it was my 15 minutes of fame you know being interviewed on how that

[00:17:15] [SPEAKER_00]: process was unfolding and everything that we were learning there at Best Buy

[00:17:20] [SPEAKER_00]: so and fast forward 20 years and here we are today and remote work is a part of the

[00:17:26] [SPEAKER_01]: fabric of our life now. Wow what a fantastic story I love that it does it's that interview

[00:17:32] [SPEAKER_00]: still on YouTube somewhere you must have it stored somewhere right? I think I have a DVD copy

[00:17:37] [SPEAKER_00]: of it somewhere I'm not sure it's I'm not sure it was one of the most popular episodes

[00:17:40] [SPEAKER_01]: as 60 Minutes ever did it's probably not on YouTube. I love it and for anyone listening just

[00:17:46] [SPEAKER_01]: wanting to find out more information about any of the things we talked about today and find

[00:17:51] [SPEAKER_01]: out more about AI powered predictive order promising where should they look at any way

[00:17:56] [SPEAKER_00]: Yeah I would just say you know head to nextupel.com all of our AI models are out there available for

[00:18:04] [SPEAKER_00]: people to look at of course we're on LinkedIn and we do have some content out on YouTube as

[00:18:09] [SPEAKER_01]: well if you search for Nextupel. Well thank you so much for coming on today and demystifying

[00:18:15] [SPEAKER_01]: the world of predictive promising and how applying AI and generative AI the impact that

[00:18:21] [SPEAKER_01]: it has the real world use cases that are already transforming retail order management

[00:18:25] [SPEAKER_01]: and the value that it's bringing and ultimately how it's impacting the products everyday consumers

[00:18:31] [SPEAKER_01]: are looking for on the shelves whether or not they will be there of course but that is a

[00:18:36] [SPEAKER_01]: story for another day so just a big thank you for joining me on the podcast. Oh thank

[00:18:41] [SPEAKER_00]: you so much it was terrific to get to talk about this and as I said super excited about

[00:18:46] [SPEAKER_00]: where AI and ML are going to go in that field of retailer management. So as we've heard today

[00:18:51] [SPEAKER_01]: AI is not just a buzzword in retail it's actually the driving force behind tangible

[00:18:57] [SPEAKER_01]: improvements in order management inventory accuracy and even customer satisfaction and I

[00:19:03] [SPEAKER_01]: cannot thank CHAP enough for coming on and illuminating how Nextupel's innovative

[00:19:08] [SPEAKER_01]: approaches are enabling retailers to essentially predict promise and deliver like never before

[00:19:15] [SPEAKER_01]: ultimately transforming how products move from those warehouses right into our hands as

[00:19:21] [SPEAKER_01]: consumers but what stood out for you in today's discussion what was was it the potential of

[00:19:26] [SPEAKER_01]: predictive promising or enhancing customer confidence or even exciting applications of

[00:19:32] [SPEAKER_01]: generative AI and inventory management whatever it was I think it's clear that AI is set to

[00:19:38] [SPEAKER_01]: revolutionize retail and offer businesses tools tools that they need to thrive in this

[00:19:45] [SPEAKER_01]: increasingly competitive area but as we wrap up think about these advancements how they

[00:19:50] [SPEAKER_01]: could impact your own experience as a consumer and equally as a professional in the retail industry

[00:19:56] [SPEAKER_01]: and ask yourself how will AI change the way you shop or the way you manage your business

[00:20:01] [SPEAKER_01]: these are the questions that are helping shape the future the future of retail and I hope

[00:20:07] [SPEAKER_01]: today's conversation has sparked just a few ideas and insights for you but as always thanks

[00:20:12] [SPEAKER_01]: for tuning in and until next time keep exploring this intersection of technology and business

[00:20:18] [SPEAKER_01]: and hopefully you'll join me again bright and early tomorrow here on tech talks daily

[00:20:22] [SPEAKER_01]: because we're going to do it all again bye for now