Latest / The Payments Shed / Ep. 63 - LIVE from FTT Payments with Zilch: Reimagining Loyalty, AI, and the Value of Payments Data

Transcript

Ep. 63 - LIVE from FTT Payments with Zilch: Reimagining Loyalty, AI, and the Value of Payments Data · The Payments Shed

• 0:05
  We're live here at FTT Payments and today we're joined by Sean
• 0:10
  Heatherman, CTO at SILTS. Thank you very much for for
• 0:13
  joining us today. I want to start about data as an
• 0:18
  infrastructure first and foremost.
• 0:20
  I think payment data for me especially has what when I look
• 0:24
  at the industry has always been the by product.
• 0:26
  What feels now that it's really changing?
• 0:30
  What do you believe is really different now versus a a while
• 0:34
  ago? Yeah, I think, I think what's
• 0:37
  tended to happen, there's a lot of and by the way, thank you
• 0:39
  very much for having me, having me.
• 0:40
  Let me start over. That's really good to see you
• 0:42
  guys. But what's tended to happen in
• 0:45
  the past is people have just collected the transactional data
• 0:47
  and not really done anything with it.
• 0:49
  They haven't thought about how to attribute it.
• 0:51
  They haven't thought about how to to really analyze it.
• 0:54
  And from the beginning at Zilch, we had a mantra of no data left
• 0:58
  behind. We wanted to be able to collect
• 1:00
  all the data that we could and make sure we could see a
• 1:02
  sequence of information. And when you think about it,
• 1:05
  payments data is some of the highest value data you can get.
• 1:09
  It is literally a person has made an important decision, in
• 1:13
  some cases a life changing decision, OK, when they're
• 1:16
  making that payment. And to not have that connected
• 1:20
  to understanding what the customers about is, is is
• 1:23
  frankly criminal. We we, we should all we should
• 1:25
  all be thinking about data as one of the key first class
• 1:29
  pieces of our infrastructure. And I think just on see that
• 1:32
  realization is through a lot of organizations.
• 1:34
  Now, the data on its own isn't terribly useful, but it's in
• 1:38
  combination with a whole bunch of this.
• 1:39
  This is what the person was looking for.
• 1:41
  This is what the person did shortly afterwards.
• 1:43
  This is the kind of places that they've shopped in the past.
• 1:46
  That kind of data is really, really useful.
• 1:48
  So really trying to get that maximum value to understand our
• 1:51
  customs and to be able to provide that information and
• 1:54
  personalisation to them. It's.
• 1:56
  Good to hear that narrative, particularly in the payments
• 1:58
  industry, like we have so much data, but we haven't done the
• 2:02
  right things with it necessarily over the years.
• 2:03
  Like par data is constantly something I'm getting asked
• 2:06
  about at the moment. Like I think like again, like
• 2:07
  the accessibility to that is not, is not always there and
• 2:10
  businesses could really benefit from that.
• 2:12
  You talk about first party transaction data with a real
• 2:15
  asset. What does that actually unlock?
• 2:19
  Well, for me, first party data is the data that we have.
• 2:22
  It's the days that we collected. So as an example, when we
• 2:25
  started Zilch, and I mean it's almost closing on eight years
• 2:29
  ago now that I started, one of the decisions we made was we
• 2:34
  were going to insource our Ledger.
• 2:35
  What we wanted to do is make sure that exactly this kind of
• 2:38
  information was fundamentally under our control.
• 2:40
  While we aren't sourced acquiring and we aren't sourced
• 2:43
  payment issuer processing, we wanted to be the the decider of
• 2:47
  record. We wanted to be the ones
• 2:48
  actually making the decision on the transaction based on the
• 2:51
  data we got. As you say, you've got the card
• 2:53
  transaction data, which isn't always the greatest, but it's
• 2:56
  getting better these days. But you've also got the
• 2:59
  combination of what the person's been doing in the app.
• 3:01
  So, so really what we want to do is make sure that we had that
• 3:05
  data in our systems, in the formats that we wanted it, that
• 3:09
  we could apply our semantics to it rather than using some third
• 3:13
  party Ledger system, which maybe didn't really work the way you
• 3:15
  wanted to. And that's just at one example
• 3:18
  of of what we tried to do. Is that what makes embedded
• 3:20
  finance intelligent and not just functional?
• 3:23
  I, I think, I think embedded finance, if you look at it, a
• 3:25
  lot of it is these individual pieces of platforms and, and
• 3:28
  really you've got to think about well, what's orchestrating that
• 3:31
  whole system. I'm taking this in Zilch's case,
• 3:35
  I'm taking this situation where the customers registering with
• 3:38
  the app, we're doing a credit check, we're doing KYC, we're
• 3:41
  doing some stuff like that for regulatory reasons, fraud
• 3:44
  checks, all those kind of things.
• 3:45
  Then they're going along and there's linking a debit card.
• 3:49
  Well, what kind of debit card? Then they're going along and
• 3:52
  they're making a purchase, well, where they're making that
• 3:54
  purchase. And so really tying all that
• 3:57
  together is the glue that we do. So even though we don't process
• 4:01
  the actual transaction against the bank, the acquirer does
• 4:03
  that. Even though we don't actually
• 4:05
  integrate with MasterCard and Visa, the issue of processing
• 4:08
  does that. All of these funnel into our
• 4:10
  orchestration layer. We make the fundamental decision
• 4:13
  based on our data and our copy of that data.
• 4:16
  I want to talk about loyalty built on top of payments as
• 4:20
  well. Like loyalty has always been
• 4:21
  about points and rewards. Traditionally that model feels a
• 4:26
  little bit outdated. What do you think is broken with
• 4:28
  that model at the moment? Speaking personally, have just
• 4:31
  become suspicious of points a little bit as though we've
• 4:34
  always tried to tie the points very closely to pounds, but
• 4:38
  everybody kind of knows the points degrade in value.
• 4:40
  You can get less of the points. It's kind of an easy way for the
• 4:43
  retailers or the the, the merchant to, to cut costs a
• 4:47
  little bit. I think what you're starting to
• 4:49
  see if you look at for example, Tesco and Sainsbury's is they're
• 4:52
  actually doing loyalty as pricing.
• 4:53
  OK. So it's not that you get points,
• 4:56
  it's that you get a better price of the checkout tool and that
• 4:59
  gets very, very interesting. So I think loyalty programs are
• 5:03
  definitely an area where there's been a lot of innovation, a lot
• 5:05
  of focus these days. One of the things that I think
• 5:08
  that retailers should think about is your loyalty program is
• 5:13
  only capturing a thin slice of the customer behaviour.
• 5:17
  So if I'm swiping my Tesco Club card, OK, cool, I'm going to get
• 5:22
  that at e-commerce and I'm going to get that on every single tap
• 5:25
  and pay transaction. Well, maybe not.
• 5:27
  Am I going to see where else this customer is making
• 5:30
  purchases? Well, no, maybe not.
• 5:31
  OK. And that's where I think, you
• 5:33
  know, companies like Zilch can actually provide a lot of value
• 5:36
  to retailers is we can help augment those loyalty programs.
• 5:39
  It isn't an either or kind of situation where you're doing in
• 5:43
  house loyalty or using the partner like Zilch.
• 5:46
  It's a question of both. If you're focused just on your
• 5:49
  own in house loyalty and you're assuming you're going to capture
• 5:51
  that very first transaction all the way through, you're going to
• 5:53
  have the whole life cycle. Well, it's basically only Amazon
• 5:56
  that has that, right, maybe Timu and a few others these days.
• 6:00
  So, so I really think having a loyalty program that focus on
• 6:03
  both sides of those probably really, really good.
• 6:05
  I'm glad you mentioned supermarkets as an example,
• 6:07
  because obviously we're filming this in a week when Lidl are
• 6:11
  very much in the news for a small change to their loyalty
• 6:14
  program, which has caused absolute uproar in their
• 6:17
  customer base, right? And they do focus on it was more
• 6:20
  that monetization element, right of loyalty.
• 6:22
  And you turn that financial, I spend with points, which equates
• 6:26
  to X amount of money to I can now get a bread roll.
• 6:30
  People don't take it as well, right?
• 6:32
  Like it's, it's the backlash has been real for those guys.
• 6:35
  Yeah, and you see it all the time.
• 6:37
  I mean, British Airways had went through a whole bunch of these
• 6:39
  kind of examples a while ago. People get very passionately
• 6:42
  dedicated to what their points can get them and the various
• 6:46
  rewards they get. So you got to be very, very
• 6:48
  careful change that. The advantage of the pricing
• 6:50
  model is it's very, very simple. It's well, here's the list price
• 6:53
  and here's the list price if you have the car.
• 6:56
  So that's very direct and very immediate for the customer.
• 6:59
  Yeah. So does this create much more of
• 7:01
  a direct relationship between those merchants and the
• 7:05
  consumers? And how does it change the value
• 7:08
  that is delivered at that point of spending?
• 7:11
  Well, I think it's exactly that, right?
• 7:13
  If, if, if you're now swiping your card and if you think about
• 7:16
  the way, the way that you sort of swipe a card these days at,
• 7:18
  at checkout, you're swiping out with your, you've got to scan
• 7:21
  your Tesco card, then you've got to use a payment card.
• 7:23
  It's a bit of a clunky sort of experience, right?
• 7:26
  But you do that because you're going to get these benefits,
• 7:29
  right? So that's the sort of premise of
• 7:30
  loyalty programs. Since I'm going to get these
• 7:32
  benefits in the future. As I say, what I quite like, and
• 7:35
  this is me personally, it's not necessarily everyone, what I
• 7:37
  quite like about the discount model is, again, as I said, it's
• 7:40
  very much connected right there on the till.
• 7:41
  I can see it, OK, I would have spent £40, but because I used my
• 7:45
  Tesco Clubcard, I'm spending £30.
• 7:47
  Fantastic. I've just saved £10.
• 7:49
  That's real. It's immediate, it is in my
• 7:50
  pocket and payments gives you that ability to do that at the
• 7:54
  point of checkout, whereas a points based system just
• 7:57
  doesn't. And then just putting your, your
• 7:59
  CTO hat on because I think it's always fascinating when we have
• 8:03
  a CTO on, on the show because the perspective is so different
• 8:06
  to some of the other guests that we have.
• 8:07
  You guys are obviously at a certain size now where this
• 8:12
  narrative that you have, I want to dive into because you talk
• 8:15
  about designing for where you're going, not where you are.
• 8:18
  I don't want to get your your feedback on what that means in
• 8:20
  principle, but also as an addition like how sustainable
• 8:23
  that is as your growth continues to go as it is a business like
• 8:27
  Zilk. Yeah, I mean, I think there's,
• 8:29
  there's always this, especially when when you're found in seats
• 8:32
  here like myself and you, you write there at the very
• 8:34
  beginning. There's there's there's quite
• 8:36
  difficult trade off. Do you get something out the
• 8:39
  door as fast as possible and not worry about the future?
• 8:42
  And the good news about that is you get something out the door
• 8:44
  very, very quickly. The bad news is it just doesn't
• 8:47
  skip. And if you got to our scale,
• 8:49
  you'd be, you wouldn't get to our scale.
• 8:50
  I mean, that's just the reality. So you wind up having to rewrite
• 8:53
  everything re architect everything.
• 8:55
  The other extreme you want to avoid is essentially a gold
• 8:57
  placing everything. You're making everything perfect
• 8:59
  for this imaginary scale that may never materialize.
• 9:03
  OK, and that's just as bad because now you're not even in
• 9:07
  the market at all. There's a fact that's worse.
• 9:09
  OK, So what we tried to do is we knew from the beginning what our
• 9:13
  vision was. We knew it was very much around
• 9:14
  customer protection and customer benefit.
• 9:17
  We knew what our ambitions were. And so then we had to draw a
• 9:22
  middle path. There's how can you make
• 9:24
  decisions that are pragmatic today, but don't box you in in
• 9:28
  the future. And, and Jeff Bezos always has
• 9:30
  a, a very interesting way of thinking about decisions, which
• 9:33
  is he talks about type 1, type 2 decisions.
• 9:35
  And the one type is when I make this decision, I can undo it
• 9:39
  fairly quickly. And the other kind of decision
• 9:41
  is, well, if I make this decision, I'm locked in, I'm
• 9:44
  going to go sign a three-year contract with an issue, a
• 9:46
  processor, for example. So you got to think about that
• 9:49
  from an architectural perspective as well.
• 9:51
  Can I make, can we go and do a sort of quick and dirty where I
• 9:55
  know we can optimize it relatively easy down the line
• 9:58
  rather than doing the quick and dirty, which I know is going to
• 10:00
  be very hard to optimize down the line.
• 10:03
  And that's generally where we try to do it.
• 10:04
  We're not necessarily doing the optimization on day one, but we
• 10:07
  know what the path is to optimize should we need a tap.
• 10:11
  As a CTO and having skilled payments company, especially
• 10:14
  from an infrastructure, we already mentioned all of the
• 10:17
  modular components that you've integrated working with that
• 10:20
  data. But what are some of the biggest
• 10:22
  mistakes that payments companies are making whenever they're
• 10:25
  scaling their payments infrastructure?
• 10:27
  One of the biggest mistakes I see generally in in most
• 10:30
  organizations is kind of conflating, well, thinking that
• 10:35
  quality and speed are two different things that that you
• 10:39
  have to trade one or for the other, OK, that you can have
• 10:43
  quality or you can have speed, you can't have both.
• 10:45
  And, and I fundamentally disagree with that from the
• 10:47
  ground up from from my perspective, if you're doing
• 10:51
  things, and I'm not saying goal placing again, but if you're
• 10:53
  doing things with the requisite degree of quality, it causes
• 10:57
  less distraction, it causes less noise, it causes less customer
• 11:00
  problems, it causes less pain in general.
• 11:02
  And that allows for teams to focus more and allows them to
• 11:05
  deliver more quicker. So fundamentally across the
• 11:08
  entire industry, not just payments, I'd say that's one of
• 11:10
  the biggest things that I look at and I go, people are missing
• 11:12
  a trip, They're trying to cut corners left, right and center
• 11:15
  and they're not realizing sure, that works in the very short
• 11:18
  term. And we still, we do do it from
• 11:20
  time to time, but you've got to pay down that debt.
• 11:23
  You've got to go and fix that, because otherwise what you're
• 11:25
  going to do is spending more and more time, more and more effort,
• 11:28
  just keeping the ship running and not actually advancing it.
• 11:32
  And we talked about AIA little bit today.
• 11:35
  I think it's inevitable that AI comes up as a topic of
• 11:38
  conversation even all all these years later.
• 11:41
  The first sort of events where it was, it was subsequent.
• 11:44
  It changes and evolves day in day out, right.
• 11:46
  So you've spoken about consumer first AI.
• 11:50
  Why should that be the design brief and not a constraint?
• 11:55
  And again, I think this goes back again to to Zilch's
• 11:58
  founding principles, right. We were always very customer
• 12:01
  first from the beginning. We went off and got regulated
• 12:05
  for BNPL before there really was any BNPL regulation.
• 12:08
  We worked with the FCA to go do that regulation.
• 12:12
  We don't have to. We did it because we wanted to
• 12:14
  make sure that our customers would have access the same kind
• 12:18
  of protections they get from other finance regulated
• 12:21
  financial services companies, things like the ombudsman and,
• 12:24
  and and those kind of protections and similar.
• 12:27
  When we decided to launch via credit card via card program,
• 12:30
  one of the advantages was Section 75 protection, which
• 12:33
  again gives customers that ability to repudiate the
• 12:35
  transaction if it wasn't them. So we've always focused very
• 12:39
  much on the customer and with AI, it is the same.
• 12:43
  The question we're asking ourselves is first, first start
• 12:45
  with customer. How does this help the customer?
• 12:47
  How do we make sure we're not going to protect customer?
• 12:50
  If I'm putting an AI chat bot in front of a customer that's going
• 12:53
  to answer questions, I want to know that we've done as much as
• 12:56
  can reasonably be expected to make sure that it's going to
• 12:58
  provide accurate and useful information.
• 13:01
  And if you start with that as your premise, OK, I think most
• 13:05
  of the other things flow out of it.
• 13:07
  And I don't think it slows you down again to the point of the
• 13:09
  speed versus quality. It maybe slows you down a little
• 13:13
  bit getting out the door, but it also stops a lot of noise and a
• 13:16
  lot of chaos and a lot of problems.
• 13:18
  So you've got to think about these things from that
• 13:20
  perspective a little bit. And again, you can take
• 13:22
  pragmatic decisions, right? You can take a pragmatic
• 13:25
  decision today where you know, OK, well this is good enough for
• 13:28
  now, but I know the plan to make it more secure, more reliable,
• 13:33
  more predictable. Dan, look, of course, nowadays
• 13:37
  we see pretty much every other week there's a new model coming
• 13:39
  out. It doesn't matter if it's in
• 13:40
  Tropic, the Open AI or even Gemini with their models.
• 13:45
  But in financial services especially, like you already
• 13:48
  mentioned, we're working with so many of the different regulators
• 13:51
  as well. How do you balance between being
• 13:54
  innovative and utilizing some of those newer technologies, but
• 13:58
  also managing trust like you already mentioned, like making
• 14:01
  sure that things are safe, secure, and compliant as well?
• 14:03
  How do you balance that? Well, I think you've got to look
• 14:06
  essentially is, is what are you going to use this model for,
• 14:09
  right? So, so let's go along and say,
• 14:11
  you know, the, the product team demoed an internal agent that
• 14:14
  they use, they, they want to use recently to move.
• 14:17
  Well, that's a very, a very low risk model, right?
• 14:20
  It's got a human in the loop. The human's meaningfully in the
• 14:23
  loop. It's going along and doing some
• 14:24
  market research and stuff for the product team.
• 14:27
  So let them get on with it, right?
• 14:29
  Give them the tools and let them get on with it.
• 14:30
  OK, Give them some advice and say that prom's not great.
• 14:34
  You maybe need something else that's going to verify it.
• 14:36
  You know, you can give them a few sort of tips along those
• 14:38
  kind of lines. Now at the other end of the
• 14:41
  spectrum, we've worked very closely with one of our partners
• 14:44
  to provide AI agents that will respond to customers and help
• 14:47
  customers with their with their issues.
• 14:49
  And that's working very, very well.
• 14:51
  But from that perspective, we wanted to be super, super
• 14:54
  careful. We've got to think very
• 14:55
  carefully about the training corpus of the information the
• 14:58
  agents used. We ran a sort of side by side
• 15:00
  and shadow modes. Quite a lot of time we sat with
• 15:03
  our partner and they, they found it quite painful.
• 15:05
  We sat with them and we wanted to understand how their
• 15:08
  architecture worked, how their testing worked so that we were
• 15:11
  comfortable that we could put this in front of our customers
• 15:13
  and then we lay it on our own test.
• 15:15
  So again, it's, it's what's the risk model you're applying And,
• 15:18
  and can you sandbox that risk? Can you limit the blast radius,
• 15:21
  right. If I have that product agent
• 15:24
  that's running in Atlassian in our case, well, that's, it's
• 15:27
  blast radius. The most you can do is affect
• 15:29
  stuff in the blast and it can't reach out and, you know, attack
• 15:31
  our database or whatever the case will be.
• 15:33
  So you can scope these things pretty tightly.
• 15:35
  I think the danger comes when you try and have these agents
• 15:38
  trying to cross boundaries or get too many, too many powers,
• 15:41
  as it were. The problem is that those are
• 15:44
  the most interesting agents. But yeah.
• 15:47
  We'll learn as we go, I think in that space.
• 15:49
  Yeah. Look, really interesting
• 15:51
  conversation. Thank you very much for joining
• 15:52
  us. We have a final question that
• 15:53
  everyone on the Payment Shed podcast get asked, which is if
• 15:56
  you were to banish one thing from this industry forever, put
• 15:59
  it on the shelf of shame, what would you?
• 16:01
  What would you choose? It would be the speed versus
• 16:04
  quality thing. It would 100% be that.
• 16:06
  There's no doubt a matter in my mind.
• 16:07
  It's it's, it's these are this is a mindset that needs to die.
• 16:11
  Well, that's an easy one to nominate, I think, and and get
• 16:14
  rid of. No, Sean, thank you very much
• 16:15
  for joining us. Welcome people.
• 16:17
  Get in touch with you. Find out more about Zilch.
• 16:20
  Excellent. LinkedIn is is a great place to
• 16:21
  reach out. I'm I'm not great at paying
• 16:23
  attention to LinkedIn, so you may have to be a bit patient,
• 16:26
  but it's a great place. Where there are people to be
• 16:28
  saying that too, I'm sure I'm. Sure.
• 16:31
  I go through phases. Like every month or so I do a
• 16:33
  flurry of LinkedIn and then and then I don't.
• 16:36
  Well, enjoy the rest of the event and thanks for coming on
• 16:38
  the. Show absolute pleasure.
• 16:39
  Thank you very much. Cheers guys.
• 16:40
  Thanks a lot.