Latest / The Retail Podcast / AI Mastery: UK's Top Grocer Leads with Innovation

Transcript

AI Mastery: UK's Top Grocer Leads with Innovation · The Retail Podcast

• 0:00
  Hello and welcome to our post NRF Interviews.
• 0:03
  So let me introduce Paul Windsor who is Head of Retail and
• 0:09
  Industry at Snowflake. So Paul, welcome to the podcast.
• 0:14
  For those who don't know, why don't you tell us a little bit
• 0:18
  about yourself and then a little bit about the company that you
• 0:22
  work for, Snowflake and and what they.
• 0:24
  Do, yeah. Thank you, Alex, and thanks for
• 0:26
  having me today. Yes.
• 0:28
  So a bit of background about myself.
• 0:31
  My career in retail can be split split perfectly down the middle.
• 0:35
  I spent the 1st 19 years of my career at Sainsbury's, where I
• 0:41
  think in those nineteen years I've got an incredible
• 0:45
  understanding of the retail business and it was a great 19
• 0:50
  years in the company. But my last job I had at
• 0:53
  Sainsbury's, My last role was all about the data component and
• 0:57
  how data is shared and collaborated on with large CPG
• 1:03
  companies. Now that may into my next 19
• 1:07
  years of my career, which is always it's been in the data and
• 1:10
  analytics space. So I spent 19 years pretty much
• 1:14
  working for companies that are helping retailers solve their
• 1:18
  data challenges, whether that was with data warehousing in the
• 1:22
  kind of first generation of when data warehouses became something
• 1:26
  that retailers invested in. I've spent a number of years in
• 1:30
  the whole BI space and then recently in the AI space around
• 1:35
  automated machine learning. And two years ago I got the
• 1:38
  opportunity to join Snowflake, which is the leading data
• 1:42
  platform that was born in the cloud and snowflakes.
• 1:47
  Mission today is to mobilise the world's data.
• 1:51
  So this opportunity to give access to retailers for the data
• 1:57
  they need with the data consumers that need to consume
• 2:01
  that data in a way that makes it easy and and approachable in the
• 2:06
  way that they they tackle their their data challenges.
• 2:10
  So, so, yeah, so been with Snowflake for just over 2 years.
• 2:14
  I lead our industry go to market for Snowflake in EMEA.
• 2:18
  We have a number of retail customers today that are
• 2:22
  publicly, publicly I can talk about it's using Snowflake.
• 2:26
  One of them happens to be, if I can point to it correctly Alex,
• 2:30
  Under Armour which are obviously I've got my Under Armour and
• 2:33
  Snake like talk on today, that's all about how they use Snowflake
• 2:37
  from their supply chain perspective.
• 2:39
  And then my old, my old company that I work for, Sainsbury's is
• 2:43
  also a customer of Snowflake as well as John Lewis and River
• 2:48
  Island if we kind of localise this down to kind of the UK
• 2:51
  market. So that's about me and a little
• 2:54
  bit about about Snowflake, but I'd love to get on to the
• 2:56
  conversation of NRS as well. Yeah so I just want to this is
• 3:02
  in the spirit of you know what my audience is thinking cause
• 3:06
  every talk at NRF that had anything with AI in in the title
• 3:11
  irrespective of what technology, vendor or partner it was from.
• 3:17
  You can imagine all or all of the the the major players had AI
• 3:20
  in their title. Everyone you know everyone was
• 3:24
  super interested. So I just want to be, you know,
• 3:26
  cut to the chase and in in my mind and I'd love your opinion
• 3:30
  on this, how much of AI that we talk about is true AI as opposed
• 3:36
  to machine learning plus, right. So for those who are data
• 3:41
  specialists, they'll be like well that's just that's machine
• 3:45
  learning. We've been doing that for years.
• 3:46
  That has nothing to do with AI or generative AI.
• 3:52
  And and I'm just thinking from a from a strategy perspective
• 3:56
  before we sort of sort of go deeper down and down.
• 3:59
  If we just keep it at that high level, what are your thoughts in
• 4:03
  actual AI outcomes? And if you want, why don't you
• 4:06
  define what AI means to you and Snowflake And then we can sort
• 4:12
  of distinguish because that was the biggest frustrating thing
• 4:15
  for me that people would just like talking about machine
• 4:17
  learning as it was a I and hoping that no one would ask
• 4:19
  that question and like hang on a minute, that's just machine
• 4:23
  learning. That's like something that you
• 4:24
  could put in an Excel spreadsheet and extract better
• 4:27
  outcomes from. So go, go ahead, tell me.
• 4:30
  Well, I mean it's it's it's a great point to kick off with,
• 4:33
  right? So without a doubt, those three
• 4:36
  days in NRS, the number one hype and the topic and the theme was
• 4:42
  Jenna in large language models without a doubt.
• 4:44
  But I agree with you, not many of those people that were in
• 4:47
  those speaking sessions went much deeper than just sort of
• 4:52
  reflecting on the fact that this is something that they'll be
• 4:54
  looking to try and experiment with.
• 4:57
  Yeah, from snowflakes perspective and perhaps it's
• 5:02
  sometimes does get missed off here is that you can't have an
• 5:06
  AI strategy without a data strategy.
• 5:10
  Absolutely. Now, sometimes this gets lost.
• 5:13
  Now this is something that we're trying to kind of.
• 5:16
  Interrupt you there, Yeah. Is that is the reason because
• 5:20
  there's a separation between the two?
• 5:22
  Because the people talking about the subject are not from a day
• 5:26
  like yourself? From a data background is what?
• 5:30
  What's the reason for the disconnect between the two?
• 5:32
  Well, I think, I think we've now come into a new era of how AI
• 5:37
  can be used because generative AI is something that can now be
• 5:42
  used. Large language models is
• 5:44
  something that can be used and deployed.
• 5:47
  But if you think about it, it can't be used without the data.
• 5:51
  So if you don't have a data strategy, and by a data strategy
• 5:56
  we mean you to be able to truly get the best value out of
• 6:01
  building a Gen AI model or a large language model, you are
• 6:05
  going to need to have access to the data that you need to build
• 6:10
  that model. Now this has been a challenge
• 6:13
  for a number of companies for years.
• 6:15
  This is how Snowflake has grown so fast as a business is that
• 6:18
  lots of data sits in multiple silos.
• 6:22
  You can't just connect a large language model when data sits in
• 6:27
  multiple silos. You need to have that data
• 6:29
  unified. You need to have that data in
• 6:31
  one place. You need to have it governed.
• 6:34
  And also Alex, what's really important here is you need to
• 6:36
  have the kind of framework behind it so that when you start
• 6:40
  to build these large language models, it's contained inside
• 6:44
  your own platform for security purposes.
• 6:48
  But the first thing that any company needs to do when we
• 6:51
  start to hear from our customers, they're teens hear
• 6:54
  about Jenna and large language models.
• 6:56
  We always ask the question, you know where are you in terms of
• 7:00
  your data strategy? Do you have the data today ready
• 7:03
  to start to take advantage of those large language models in
• 7:06
  general, I and I don't believe I heard that in NRF.
• 7:09
  I heard exciting. I heard market signals increase
• 7:14
  the efficiency of the supply chain, but it didn't hear
• 7:17
  anything about the fact that first and foremost you need to
• 7:19
  have a really good unified data strategy.
• 7:23
  So I'm I'm not going crazy thinking that, right?
• 7:26
  Why not at all? That that is like there would
• 7:29
  cause, I was thinking but just maybe there's an an NLM that
• 7:33
  you've not thought about Ali that's actually magically
• 7:36
  presenting the data into these platforms.
• 7:41
  No, because all the platform providers that's their, you know
• 7:44
  come come with us on the journey and we'll take care of.
• 7:46
  I'm thinking if your data is not ready and as we know most of the
• 7:50
  industry, yeah is not ready because everything sits in
• 7:55
  silos. So let let's just talk about the
• 7:58
  the, if you like the silos and breaking that down as well.
• 8:02
  So does does my finance data need to sit?
• 8:06
  Like how do I manage that? My finance take the market
• 8:09
  marketing data? My ERP data?
• 8:12
  Where? Where do I start with, with my
• 8:14
  silos? Within my business or my systems
• 8:17
  to be fair? Yeah.
• 8:18
  Well, this is one of the, this is one of the key strategies
• 8:21
  that Snowflake is helping to solve for.
• 8:22
  Our retail customers of course have the source systems where
• 8:26
  the data actually is generated and curated, but that's not
• 8:30
  where you would take advantage of building large language
• 8:35
  models, Gen AI, machine learning models, analysing that data to
• 8:39
  understand business performance, looking at past historical sales
• 8:43
  etcetera. We all know that that's just
• 8:45
  where the data is sourced and generated from.
• 8:48
  You need to bring that into a platform which is where the data
• 8:52
  can be consumed. So where Snowflakes been helping
• 8:55
  retailers for a number of years is to bring exactly those
• 8:58
  functional data sets that you just mentioned here.
• 9:01
  You need to have your finance data alongside your marketing
• 9:04
  data alongside your supply chain data.
• 9:07
  If you're thinking about running a campaign that is a promotional
• 9:11
  campaign to your consumers, you need to understand their
• 9:16
  behaviour. So that's all of the customer
• 9:18
  data residing in the same platform as your inventory data.
• 9:22
  Because if you're going to be promoting stock, you need to
• 9:24
  understand if you actually have the stock available to actually
• 9:29
  promote those brands. And you need to have your
• 9:31
  finance data available to understand the costs of running
• 9:35
  these promotions. So bringing all of that data
• 9:38
  together is an absolute necessity and this is what we've
• 9:41
  been helping our retail customers in the first instance.
• 9:45
  Now Gen AI and large language models is just taking that onto
• 9:48
  another level. But again, this is just another
• 9:51
  capability to start driving those outcomes and those
• 9:55
  predictions. But the underlying part, which
• 9:58
  unfortunately is not the interesting exciting parts of
• 10:02
  many people, Alex, is you need to have that governed unified
• 10:06
  data platform way you can consume that data from.
• 10:10
  And then I'll go one step further as well, which is
• 10:13
  something that many retailers are not necessarily truly taking
• 10:17
  advantage of to a wide degree is in order to kind of build real
• 10:23
  good predictions on future behaviour or detections of
• 10:27
  market signals, you need to access third party data as well.
• 10:31
  So think about all those third party datasets, Alex, that some
• 10:34
  of the biggest data providers in the world are monetizing today.
• 10:39
  Now everybody draws their attention straight away to
• 10:42
  weather data because it's been something that's been talked
• 10:44
  about and the retail industry for 20 years.
• 10:46
  But we're also talking about demographic data, economic data,
• 10:50
  ESG data. We've got all of these types of
• 10:54
  datasets that could also be enriched alongside your own data
• 11:00
  to then run those kind of AI and those predictions as well.
• 11:05
  So you need access to that third party data.
• 11:08
  Now another little tiny plug for Snowflake here Alex, it's it's
• 11:12
  Snowflake has the largest marketplace on its platform, the
• 11:17
  third party datasets. It has over 2200 third party
• 11:21
  datasets sitting on the Snowflake platform.
• 11:24
  So if you are a Snowflake customer and you'll you'll find
• 11:27
  all of your own first party and all your own data unified
• 11:31
  together to start to really understand your business.
• 11:34
  But you feel that there might be a competitive advantage to
• 11:37
  access third party data. That data is available to you as
• 11:41
  well as part of the platform for you to enrich it.
• 11:44
  Bring that together and now you're really scaling the way
• 11:48
  that you can do predictions on things like AI.
• 11:52
  So if we take sort of the data economy in action and from your
• 11:56
  experience from whichever you know whether it's Sainsbury's
• 12:01
  Underarm or one of the the retailers that you, you you you
• 12:05
  mentioned and understanding that they must have had some data
• 12:09
  strategies that you know they they already set and and now
• 12:13
  leveraging your platform. Can you share any sort of I I
• 12:19
  don't know if any of them were in customer data or the loyalty
• 12:23
  programme area but but I'm what I'm curious is what the use case
• 12:28
  is and then what the intelligence is that's being
• 12:31
  drawn out to make actionable business outcome.
• 12:36
  So can you have you got any examples?
• 12:38
  For us, yeah, I do. So I I think if you look at
• 12:42
  snowflakes, opportunity to drive success for our retail customers
• 12:49
  today, the typical conversations that we're having, So we've had
• 12:52
  those data conversations. You need to unify your data, you
• 12:55
  need to bring it together, you need to be able to give it
• 12:57
  seamless access to that data of all your data consumers.
• 13:01
  That's the first part of a road map of a data strategy journey,
• 13:06
  bring that data together. We are now having conversations
• 13:11
  with our retail customers today which is around surprise Alex,
• 13:15
  this is, this is no kind of surprise for you.
• 13:17
  And also what is now a hot theme as well, Customer 360 does not
• 13:23
  go away and it continues and it will never be sold.
• 13:27
  Customer 360 will never be sold. We know that.
• 13:30
  But it's about now this opportunity to bring all your
• 13:33
  data together from a customer perspective.
• 13:35
  I don't know if you picked up on this at NRF, but one of the
• 13:38
  sessions I think I went to and I think it was for an Abercrombie
• 13:45
  and Fitch I think was speaking at that intersection and they
• 13:48
  were talking about well moving now from personalization to
• 13:53
  individualization. So this idea now the
• 13:56
  personalization is we want to personalise as much as we can
• 13:59
  our relationship with the consumers.
• 14:02
  Now we're getting down to that level of granularity with the
• 14:05
  data. The data is all unified
• 14:07
  together. You've got the capabilities with
• 14:10
  modelling in AI and this potential around large language
• 14:13
  models. Can we get to the point where we
• 14:15
  can start to individualise those relationships with our customers
• 14:19
  as well? And you can see that happening
• 14:21
  just here with Sainsbury's. Sainsbury's are now offering ten
• 14:26
  products every seven days to their Nectar card members that
• 14:31
  is absolutely personalised and individualised to those shopping
• 14:35
  behaviour. We'll be able to bring that data
• 14:38
  together. If you're a network card holder
• 14:40
  today, every Monday you'll get those and new offers that are
• 14:46
  absolutely individualised and personalised to you to be able
• 14:50
  to do that. That's really understanding
• 14:52
  previous customer shopping behaviour and doing that really,
• 14:55
  really well. And you can see from their
• 14:56
  results and they talk very positively about how Nectar
• 15:00
  pricing is driving revenue from that sector.
• 15:03
  It's a customer. 360 is a big, big factor.
• 15:06
  Yeah. I've got, I've got another two
• 15:09
  I'd love to just touch on. I'll be the second one is no
• 15:13
  surprise again is that supply chain.
• 15:15
  So you know Under Armour talk publicly about the fact that
• 15:19
  they use Snowflake for the importance around sharing data,
• 15:24
  Alex, to make the supply chain more efficient.
• 15:27
  We know that there are so many situations in the supply chain
• 15:31
  where that flow of goods can break down.
• 15:34
  The idea that you can support that with data sharing is
• 15:37
  really, really important. And where Snowflake does this
• 15:40
  exceptionally well is we can help to share data between two
• 15:46
  companies without moving or copying the data.
• 15:49
  So this is a really good example where in supply chain you need
• 15:53
  to share data with sales of inventory of demand and what you
• 15:59
  don't want to do is turn that into a supply chain of data
• 16:03
  having to move Where Snowflake does this exceptionally well is
• 16:07
  around that sharing of data between two companies.
• 16:10
  So that's supply chain is another key strategy for a lot
• 16:13
  of our retail customers. And then the final one that I
• 16:16
  want to sort of touch on is I I don't know if you heard it as
• 16:19
  well, but retail media, I mean retail media is where the net I.
• 16:22
  Was just about to that was gonna be my final question for you
• 16:26
  because later media is absolutely how do I combat the
• 16:30
  battering I'm getting from inflation by generating new
• 16:34
  revenue from other means. So yeah, absolutely far away on.
• 16:37
  That well, well look you know I, I again I captured a tonne of
• 16:41
  notes here from last week. This was US numbers, Alex, but
• 16:45
  for those listening in today, you've got Retail Media
• 16:49
  generating $60 billion of revenue in 2023.
• 16:53
  That's expected to increase to $100 billion in 2028.
• 16:58
  So four years away from that increasing to $100 billion,
• 17:01
  They're talking again very, very positively NRF about the next
• 17:06
  era of we're in the golden generation of retail media.
• 17:10
  And this is because now we've got this opportunity physically
• 17:14
  in store to start generating those advertising and those
• 17:19
  promotions through those digital screens that you can use and
• 17:22
  place around the store. And then you've got that
• 17:25
  absolute fantastic instant response to whether or not that
• 17:29
  advert, that advertising and that product has resulted in the
• 17:33
  product being purchased and then the closing that loop to how
• 17:37
  much was that return on investment for that advertising.
• 17:40
  So we're seeing a lot of our retail customers now wanting to
• 17:43
  get really, really solid on their first party customer data
• 17:48
  because they can use that to really decide which is the best
• 17:51
  times to promote which items through their stores.
• 17:54
  And then the other part of retail media Alex, which is
• 17:57
  again where Snowflake really comes in is we're now seeing
• 18:00
  this growth of off site media advertising.
• 18:04
  And again the numbers, the numbers were staggering.
• 18:07
  So again at NRS last week they talked about currently offsite
• 18:12
  media revenue, it's about $6.7 billion in 2023.
• 18:17
  That's taking what you can advertise of goods into
• 18:23
  advertising and media channels which is away from your on site
• 18:27
  channels. That growth is set to go from
• 18:30
  6.7 to 24 billion in the next three years and that's because
• 18:35
  you can take that data now and you can start to monetize it
• 18:39
  through off-site channels. But it's Paul, just interrupt
• 18:43
  you on that. But I see that as so I get the
• 18:47
  outcome that if you're delivering adverts in store then
• 18:53
  delivering those adverts off site somewhere else.
• 18:57
  Yeah. But again because I've sort of
• 19:00
  grown up in this industry you know the martech that that you
• 19:04
  need to to to develop something like that, I don't know, I don't
• 19:08
  I don't see it as easy as so. So the I guess the possibility
• 19:12
  and the numbers are there to say look if you can do it, it is but
• 19:18
  I don't know if I took the UK grocer market bar two of them, I
• 19:22
  don't think the others could do that, right.
• 19:24
  I don't know if you've got experience in the UK market and
• 19:28
  obviously one of the grocers you you're working with.
• 19:30
  Yeah. But I don't see how they could
• 19:34
  because obviously then you're you're you you're just looking
• 19:37
  at the integration with sort of expanding that problem of data
• 19:42
  sets out across other other assets that might be owned by
• 19:47
  other people. Well, it's interesting the
• 19:49
  integration passed the really important words that you just
• 19:52
  reflected on here. So of course the grocers are
• 19:55
  leading the charge in terms of the way that they can do this.
• 19:58
  They can take that marketing budget and they can invest that
• 20:03
  in offside channels. This is where the risk is at
• 20:07
  right now is obviously protecting personal identifiable
• 20:11
  information is EI customer data, your data and my data.
• 20:16
  And what you don't want to do is once you've brought that
• 20:18
  together inside a data platform like Snowflake, and you've
• 20:22
  actually governed it and you've got it secure and now you're
• 20:25
  protecting that data and you're using it for your own insights
• 20:29
  and your understanding of your customers and your business.
• 20:31
  And now you're thinking about a media strategy where data needs
• 20:34
  to actually be shared outside of your platform and with
• 20:39
  advertisers and media agencies, et cetera.
• 20:42
  Now we're into the real era of data cleanrooms.
• 20:45
  Now this is where the conversation may get a little
• 20:48
  bit more too technical, but for those listening in today, data
• 20:52
  clean rooms is a fantastic way of being able to share data but
• 20:58
  not expose the data. It's a beautiful way of sharing
• 21:01
  data and insights without moving the data.
• 21:04
  And so we're starting to see capability baked into platforms
• 21:09
  like Snowflake to allow you to be able to share data without
• 21:14
  moving or copying or exposing it.
• 21:15
  And I think that's where we're going to see the platform
• 21:18
  capability enabling that growth of that retail media.
• 21:22
  So that advertising can happen off site, but the data stays
• 21:26
  governed and secure. So that's where we're, we're
• 21:29
  talking to most of our customers today.
• 21:33
  Brilliant, Paul. This was going to be a quick
• 21:36
  fireside chat, which I think has given the audience a lot of food
• 21:40
  for thought. Any sort of where do you, where
• 21:44
  do you think will be at next year's NRF?
• 21:47
  Well, because it's like almost. Do you have any views of where
• 21:51
  this is all going? Will it be AI .2 dot O or
• 21:54
  something? I thought you might ask this
• 21:58
  question so, so when I when I went 12, I went 12 months ago.
• 22:01
  I think you were saying at the start before we started this
• 22:03
  podcast. You've been for the last 13
• 22:05
  years. Last year I took away three key
• 22:08
  themes. The first one was how do we, how
• 22:11
  do we help serve the consumers that are dealing with the cost
• 22:16
  of living crisis and inflationary pressures.
• 22:18
  So it was, it was very much about dealing with with those
• 22:22
  kind of consumers. The second one was how do we
• 22:26
  move from promising to proving that we are delivering an ESG
• 22:31
  sustainability strategy. And the third one was how do we
• 22:34
  optimise our assortment and our prices again connected to the
• 22:38
  cost of living crisis that was very much in focus 12 months
• 22:42
  ago. That felt quite a challenging
• 22:46
  time and very much in media. If you look today, the three key
• 22:50
  themes was Gen AI, it was retail media.
• 22:56
  And the third one was where did I get the third one, please
• 22:59
  don't let me go. Yeah.
• 23:01
  The third one is connected consumer, Alex, with the three
• 23:05
  that were absolutely focus this time that felt more optimistic.
• 23:09
  If you asked me in 12 months time, I think we may be at the
• 23:12
  point where the NRF is going to be asking retailers to prove out
• 23:18
  how they've adopted Jenai large language models, how they've
• 23:22
  been able to get to the point of individualising those
• 23:25
  relationships with customers. I think we're going to start to
• 23:27
  see more proof now in just concept talking.
• 23:32
  No, that's wonderful. Paul, thank you so much for
• 23:34
  giving up your time in the day and I look forward to seeing
• 23:39
  what you guys carry on doing to support retailers go through
• 23:43
  this journey of transformation. Well, listen, I'm really
• 23:47
  appreciate you inviting me on here, Alex.
• 23:49
  It's been a pleasure and great. Thanks so much.
• 23:52
  Thank you.