Data Products & Data Contracts in the AI age: Ownership, Standards, Semantics – with Simon Harrer, Entropy Data

00:00:00: The better you describe your data products, the better you have a contract more detailed.

00:00:05: The better your ontology is describing what your data means that better result of answer to agent and thats feedback loop we've been waiting for.

00:00:27: Hello welcome to the Data & AI Culture Podcast.

00:00:29: my name is Karsten Bange.

00:00:30: I'm founder CEO of BARC.

00:00:32: My guest today is Simon Harra co-founder of Entropy Data and deeply involved in data products, and data contracts.

00:00:42: We will talk about the benefits that companies can achieve with using data products and data contract.

00:00:48: And he really highlights the contract as the centerpiece of benefits but also off their big topic of assigning ownership into achieving decentralization.

00:00:59: so the core idea is That we unlock the potential of data, also now in an AI age where more and more agents need to interact with data.

00:01:11: And we as organizations need to organize that and find a way to let AI agents work efficiently and effectively.

00:01:22: Data products and data contracts are key ways to support this.

00:01:26: Listen to their episode enjoy it!

00:01:29: Hi Simon welcome to my podcast.

00:01:32: Yeah hello yeah thanks for having me

00:01:34: Simon, our topic today is data products and data contracts.

00:01:39: And there's a recent development why this discussion is super timely.

00:01:44: beginning of September the new standard has been released The open data contract standard in release three point two and you were one of the people that worked on That standards so I'd be super interested in hearing more about it.

00:01:59: What what's going on?

00:02:01: On the standardization phase?

00:02:02: but before we talk about that.

00:02:05: Let's have a bit of general overview, what is your take?

00:02:08: What are the statuses of data products and data contracts?

00:02:12: how do you see implementation in enterprises but especially also value they can actually generate?

00:02:19: I think great question.

00:02:21: The key thing for having data product and contract anchored into some semantics or ontologies.

00:02:28: it was an important part Both for let's call it human ready data and I read the data that both humans in AI need to know what they does out there.

00:02:43: Who is responsible owners owner of their data or what guarantees come around with data quality?

00:02:50: Does the data coming how can i use it?

00:02:54: under what terms and conditions?

00:02:56: um, uh... And what does a data actually mean anchored in the business ontologies and I think that's crucial.

00:03:05: That was crucial ten, twenty years ago.

00:03:08: to be honest it is what always... What i just said was always crucial for humans doing system integration building pipelines but now with AI agents has risen a new level because they need some way discover what's there, decide is this relevant.

00:03:32: Understand the data and then they

00:03:34: act.".

00:03:35: And an agent doesn't really pick up the phone if it's unsure and asks a human?

00:03:42: It will just continue on...it would make an assumption and continue on.. I think that's key partner.

00:03:47: we have to be more explicit in these metadata descriptions.

00:03:53: Alright interesting.

00:03:54: so um i think for the last maybe five eight years there has been a lot of activity around data catalogs and they were also meant to serve exactly the same purpose that you just described finding data that's available.

00:04:10: And getting information about it.

00:04:13: so why is data products in data contracts now such hot topic?

00:04:17: isn't the same?

00:04:18: what different?

00:04:20: how do see them?

00:04:22: I think key part ownership guarantees promises we can really verify in that combination.

00:04:31: Having a catalog scan all the data and the company, then assigning people and say let's hope they will improve the description of their data to really take ownership just by assigning from top for all the tables or thousands millions of tables in large enterprises I think failed spectacularly but it didn't work.

00:04:55: So I think the idea now is, and that's why data products in contracts are so populist.

00:05:01: Let's say let's build something that's valuable internally by someone who really cares to take ownership and puts it explicitly like saying a marketplace or some central place where others can discover it?

00:05:19: And then i think saying people with a whip behind their back.

00:05:29: now, you have to end some better description in there.

00:05:32: And this is no ownership.

00:05:34: Ownership means you have take that step of putting the whole entry and then you also say what do guarantee?

00:05:44: What quality debt do you

00:05:45: guarantee?".

00:05:47: That's the contract where they can verify terms or data quality on all other things what you promise around the data.

00:05:59: And really, I think that's... The key part is really your ownership and putting it explicitly somewhere.

00:06:06: That's a key change.

00:06:08: before we talk about how to do that exactly because i do see some challenges around the concept still.

00:06:16: Before we do that would like to get your experience on the effects.

00:06:23: so What are the when I do this?

00:06:26: When i'm change ownership, and have more data products my enterprise.

00:06:30: The question is still so what was improving?

00:06:34: where as my business benefit in doing so.

00:06:37: I think.

00:06:37: but when you look at these enterprises know when they apply that.

00:06:45: They reduce millions of tables there before they would use it to a much, much smaller amount that is shared and they focus their governance efforts on what it's shared.

00:06:57: And I think from there people then can really find things high quality data sets and build new innovation.

00:07:07: on top.

00:07:08: We had one of our customers implemented that.

00:07:14: It was interesting to say before It was about shipments data and before it was very hard to get the data.

00:07:24: And now they had a data product with shipments that are for the whole logistics company, and there is no available tool to be accessed as an API on the cloud platform or streaming platform And they could really choose what is their best way to access.

00:07:46: It was very simple, it was accessible and this enabled them to get new use cases because before I was so hard to get that data.

00:07:56: That's a great way to support new use-cases.

00:08:05: maybe didn't happen before.

00:08:07: people said oh its too complicated we just stop Right after we had some spark, I think maybe that helps.

00:08:19: You could say with unlocking data using data products you also unlock the potential in the data to generate business value out of it?

00:08:32: Yes!

00:08:33: Maybe its still an idea.

00:08:35: as a data mesh Now, when you look at the book and as a subtitle scaling getting value out of data.

00:08:45: And I think that's key part right?

00:08:48: Absolutely no.

00:08:49: i can absolutely confirm.

00:08:50: That's the key challenge.

00:08:51: we have so many client interactions where basically this is always The core of the whole discussion.

00:08:58: how Can We ensure that?

00:09:00: How can be more effective in generating value from Data So great, so data products and data contracts have a key role to play here.

00:09:10: Before we dive deeper into that please give me your definition.

00:09:14: what's you're understanding?

00:09:15: What is the data product in?

00:09:17: what does it data contract?

00:09:20: Data product always hard to define at start with this data contract.

00:09:23: first for me a data contractors really machine readable form machine verifiable form of all the guarantees and promises of a provider or data to my consumers, many consumers.

00:09:41: Think like an API specification around that specific dataset I'm sharing typically across teams.

00:09:52: it can also contain expectations towards the consumer but mainly the guarantee is provide.

00:09:59: Could you give us few examples?

00:10:01: Guarantees on what

00:10:03: What the content is, data quality, SLAs.

00:10:09: Also how to get support.

00:10:11: who's responsible?

00:10:12: Who's the owner of the data?

00:10:13: off course and expectations towards a consumer could be that terms on conditions around the dataset.

00:10:20: maybe there are some license agreements at limit.

00:10:22: what you can do with the data And I expect my consumers to follow that

00:10:27: Absolutely.

00:10:28: Or what I also see often is that you explicitly highlight, for example personal information or PII data.

00:10:38: And then your expectation is the consumer treats it with necessary governance and care?

00:10:44: That's a good example

00:10:46: too!

00:10:46: Yeah...and i think this key part of data contract is really essential where it becomes specific one specific dataset.

00:10:55: I could be multiple tables on a schema somewhere.

00:10:59: But that's kind of what the data contract protects.

00:11:03: and The data product is.

00:11:05: to me, it's like a building block.

00:11:07: now That has follow their definition of Shamak digani from data mesh.

00:11:11: It's combining code.

00:11:14: Let's this architectural unit architecture quantum Of get data from somewhere else.

00:11:21: you have an internal pipeline in that building block And then different output ports.

00:11:27: And the contract is at one output port, right?

00:11:30: So a data contract could offer version one with one contract and then version two with one contact or the same data on different technologies that might be in a different shape even because technology has an impact in the shape of data.

00:11:48: Then you have different data contracts.

00:11:49: it's kind my perspective here.

00:11:52: They don't product.

00:11:55: the key party is really.

00:11:57: I own this building block including the pipeline and internally

00:12:01: mm-hmm.

00:12:01: Mm-hmm.

00:12:03: So you you own it?

00:12:04: You have a pipeline, and then you provide data to others.

00:12:09: in what form?

00:12:10: What's?

00:12:11: what do you see?

00:12:11: typically house data provided?

00:12:15: Typically in one of the large platforms snowflake data bricks off course.

00:12:20: But you still see the occasional CVSV file on an SFTP server, of course there and but that's again.

00:12:26: You can still build a data product then concept in the contract is totally agnostic.

00:12:32: I've it's totally fitting to put uh... That CSV file under contract as well.

00:12:41: There's often also differentiation between different types of data products, like raw or original versus derived and others.

00:12:50: Do you think that is useful?

00:12:56: I liked this idea of source-aligned or more consumer aligned product thinking about who drives the data structure.

00:13:09: is it driven by an operational system?

00:13:11: and when this changes, well you have to follow.

00:13:14: You can't really do something about that because thats why the data's generated in here completely dependent.

00:13:21: so I like this line of thinking?

00:13:22: And i also liked this consumer line-thinking of wow!

00:13:26: This was used by my business uh... By my CFO or so on.

00:13:30: they had this particular thing and.. ..I have to present a date as though there are no standard in their domain now.

00:13:38: Apart from that, I'm not dogmatic.

00:13:42: We can call it raw or aggregated integrated and I've heard so many different types of data products but still more in the camp.

00:13:56: focusing on the contract let's say focus on interfaces.

00:14:01: how we called a box does.

00:14:03: this is really flexible.

00:14:07: I remember one of your presentations at one-of-our events where the title was Data Contract is not a contract and that obviously generated a lot of interest.

00:14:18: So people really, it wasn't an eye opener.

00:14:21: maybe you explain again shortly just so we get the idea.

00:14:26: what do want to say?

00:14:27: Great!

00:14:29: Yeah i think someone else actually posted they don't link.

00:14:32: then he got...I dont know..one hundred thousand views.

00:14:35: I would now frame it as a data contract is more than a contract.

00:14:42: To me, the perspective really is about... It's more in thinking of an API specification and not a bilateral contract.

00:14:54: that's more like i as owner describe what I guarantee and then we use by many consumers.

00:15:03: They can rely on my guarantees, but they also have to follow my expectations and I think that's the key message there.

00:15:13: Okay no very good.

00:15:15: since we talk about owners i mean thats often a big topic around data products or also data mesh where this concept became really popular.

00:15:25: it has been before with idea of data product.

00:15:28: data mesh really made it super popular, and often its combined with the idea of decentralizing ownership.

00:15:36: I see both actually!

00:15:39: I also see central data teams building data products.

00:15:43: so i would like to get your view on the market yeah?

00:15:45: So who are that?

00:15:46: typical owners of data products or other typical consumers ?

00:15:51: And how?

00:15:52: maybe you also some shifts here.

00:15:53: what's happening in companies?

00:15:56: I think the key part is that everybody goes a little bit towards decentralization.

00:16:02: They want to build data products, but they're still central data teams there especially in KMU.

00:16:11: so small and medium-sized businesses have more of their key central team than does all the processing.

00:16:21: But even these companies what try?

00:16:23: do this shift left?

00:16:25: They want to make sure that where the data is coming from, they take also part of ownership.

00:16:31: Please explain shift left.

00:16:33: Shift Left means doing it as early as possible during data quality assurance maybe even in operational systems.

00:16:42: or at a moment you extract the data form the operational system into your data platform and then best way would be people who own operation system also handle the shipping of their data into the data platform because they have most domain knowledge.

00:17:03: And we see that shift, it's slow but you can still push.

00:17:09: But when going to right side consumer side there were decentralized anyway already in different business domains.

00:17:19: so this was done.

00:17:21: what is now little bit broken up as are these central large data warehouses.

00:17:25: So it's more smaller datamarks, more specific.

00:17:28: so this is another trend I see.

00:17:30: and in between you have the state of contract layer with but this how i see its either market evolving But everybody moving towards smaller building blocks toward ownership.

00:17:43: This what my bubble.

00:17:45: that could be still a Central Data team handling.

00:17:51: I don't see any conflict in there.

00:17:55: It can be both.

00:17:57: Okay, very good!

00:17:58: I think it's important to not be too dogmatic here... ...to adapt the enterprise reality and typically also a very differentiated picture for many organizations.

00:18:12: They often have domains or departments that are super independent with lots of resources.

00:18:19: basically say, hey give me the power.

00:18:22: Give me the data and then I can do great things with it.

00:18:26: but in that same organization there are other parts domains departments.

00:18:31: let's say sorry no resources nor know how.

00:18:34: yeah would be great if you sent a team could build that for me so i can at least benefit from data.

00:18:40: And i think thats reality.

00:18:42: any dogmatic approach That will not really work because it's not realistic to set a true work.

00:18:51: So I like, but here comes you talk about responsibility.

00:19:00: would be for example ideal to have the source-aligned data products that people who have knowledge on the source systems or the source processes own?

00:19:10: these and others are more consumer aligned than possibly even consumers data products and own them on top of the source-aligned product, maybe.

00:19:21: But what we experience in organizations is that often it's not so easy to find people who take their ownership.

00:19:29: We hear that its actually a real problem for these people because they say hey thats just an additional task I've already enough to do.

00:19:39: or why me?

00:19:39: Or its not my job description!

00:19:43: What are your experiences

00:19:44: there?!

00:19:45: Well, how do I find the people that really own and the products with data products?

00:19:50: And not only to do that on paper.

00:19:53: Yeah That's a hard one.

00:19:58: Also get them motivated To take part in that journey.

00:20:05: There are two ways you can... One is from the source system side incentivize them to that they actually build data products.

00:20:17: When you think of maybe your engineers that built some operational system, I mean your core business processes and Maybe if you give them also access so the data platform where they should be at a product then You have their motivation because then they can see Their own data in the data data platform an experience that it's like a little bit of dog fooding incentivizing them, that worked.

00:20:46: That's one approach because then they really take ownership.

00:20:50: Also the other part is thinking of APIs or interfaces data contracts not about data products just contracts Just a. take ownership of interfaces And now and that's where it also works when you say when you share data You have to take ownership Of off that interface Make that clear.

00:21:12: whenever you share you are responsible.

00:21:16: And then that also works further down the stream, even more towards a consumer line or rather business because whenever you share to someone else outside your team maybe?

00:21:29: You have to say you need a contract because this is their responsibility because by sharing it already take ownership.

00:21:37: and I think It also moves the topic away from data products to let's share Data at the end of the edges.

00:21:51: Let's talk about where we share and that's this focus on that.

00:21:55: I think That's there.

00:21:56: make it easy to share Make it easier, too To see what responsibles do i have?

00:22:01: mm-hmm?

00:22:02: And that's the key is a little bit off Where we should shift in discussion because otherwise its very academic Is how do you classify these data products already build them?

00:22:12: Mm-hmm lets shift on.

00:22:15: I think that helps.

00:22:18: Absolutely, and I would say marketplaces can also play a role to make it easier to share.

00:22:25: what's your view on them?

00:22:26: Do you see an increasing amount of market places for data products

00:22:31: then absolutely crucial component because this is where everything comes together Because we need a central place.

00:22:37: when you put your data products way describe your contracts anchored in your ontology still too described their meaning And then someone can say, as a human I can go there and evaluate this is something from you or an agent.

00:22:51: Can't go the end saying, Hmm?

00:22:52: This is right for my use case but when they want You need to request access because otherwise you can't operationalize that.

00:23:01: you need to get access not just for scroll through them meta data graph Maybe one two.

00:23:08: actually i get the data and i think That's The key part In A marketplace Is to me it's this central place in our decentralized world.

00:23:21: It enables the decentralization, so-to say and without that you can't really decentralize because how do you know what others have?

00:23:29: How did they know about data of other people means?

00:23:32: or what do they guarantee me?

00:23:34: Can I rely upon them for my use

00:23:38: case?".

00:23:40: marketplace products or do they build it more themselves?

00:23:48: How did you.

00:23:49: Well, a little bit biased.

00:23:50: of course we are at entropy data.

00:23:53: where have our own data product marketplace?

00:23:58: but what we're seeing is there that three ways either you built via dedicated market place like us.

00:24:06: We also see if you have data breaks your snowflake.

00:24:12: use the capabilities that they offer, with their interpretation of what is a data product.

00:24:18: Or we also see some people are just firing up cloud code and building it themselves so... We also observe this.

00:24:30: You already touched on AI but I think its obviously crucial development everyone trying to cope right now.

00:24:37: I think one effect is quite obvious.

00:24:39: So if we want to use AI, We need data and good quality data And data products can provide that.

00:24:47: so i think That's pretty clear.

00:24:48: but on the other hand now with a next shift towards agentic AI you also mentioned that like To hear your opinion On what's happening?

00:24:57: Now especially With data products in data contracts once we have maybe thousands of agents In an enterprise possibly all, or most of them need data.

00:25:07: But what's happening now with the whole idea of products-data contracts?

00:25:12: Yeah I think they're becoming even more important because in some way when the agent wants to interact with enterprise data Because otherwise... What can't it

00:25:22: do?!

00:25:22: They are very boring!

00:25:24: So and my opinion that they need know what data is available.

00:25:34: They need to know in what quality they need.

00:25:36: To know, What does the data mean?

00:25:38: How can I combine that dataset with their data set and how to fulfill my actual goal from a agent perspective And then also how to get access actually accessing the data.

00:25:50: i think That's The whole the journey and In My Perspective this is like the marketplace layer in between the agents and the enterprise data on that layer in-between because we have the data products.

00:26:04: We can quickly, well maybe you're for a couple of tens or hundreds of data products in your company probably max thousands but I'm sure it's pretty sure not more than you are into contract to all their information.

00:26:20: guarantees may be even have that check results off your contracts event and then marketplace next to say do they promises actually live to the people, their owner actually live up to that promises.

00:26:34: You have a link to the semantics of the ontologies.

00:26:37: so you know this data and this status at that column here?

00:26:41: That actually means the same because it points for the same element in the ontology.

00:26:48: And um... The agent can follow that.

00:26:51: I think thats crucial.

00:26:53: then they can request access like access gets granted maybe automatically Maybe with human in loop and then they can also directly access the data.

00:27:04: And I think that's crucial because it is, the agent needs to do all of that actually fulfill their work.

00:27:10: The better you have...the better you describe your data products, see?

00:27:14: Better.

00:27:14: if a contract more detailed, the better your ontology is describing what your data means than better result in answer for an agent.

00:27:25: That's feedback loop we've been waiting on Because the agent now drives better metadata quality.

00:27:34: And it's no longer the whip of someone saying, please increase the metadata quality but no one had really a benefit.

00:27:41: and we see the benefit immediately with the results of the agents.

00:27:46: answer?

00:27:47: Absolutely!

00:27:48: Just to my mind came a quote from my podcast with Noemi Fabelo from Cambridge University Press And she said, there's no AI product without a data product.

00:28:01: Or that is not good to AI product with out the good data products and I think thats exactly what you just said.

00:28:07: we need this as a prerequisite.

00:28:09: Just quickly to add because it also did be able to see will the coding agents?

00:28:12: its very easy to build new data products but for them you need to know want to built in how.

00:28:19: at now not the how the house easy weekend instruct agent skills on how to build it quickly.

00:28:25: but then what to build, and what does already exist?

00:28:31: What to build upon.

00:28:32: And again this is we can describe what you want to build with a contract in the product that has not yet implemented.

00:28:38: just describing the metadata first... ...and let the agent implement it.

00:28:44: also looking at what's already existing data products are there.. We still need that layer for that!

00:28:53: Now the coding agent needs to interact In the same way as we said before.

00:28:59: And I think that's why it is so crucial in the current market situation.

00:29:04: Absolutely We mentioned a new standard at the beginning.

00:29:10: There are several standards to play a role here.

00:29:13: Please give us first an overview.

00:29:15: What standards there?

00:29:16: For me, the key part of this is the open data contract.

00:29:22: It's at the Linux foundation and the BTO project.

00:29:26: It's an umbrella project, so to say.

00:29:28: And under there we also have the open data product standard which fits very nicely together.

00:29:35: but then they describe the interface, the contract and then the building block.

00:29:47: What is important?

00:29:48: We need to describe what that actually means.

00:29:54: we are focusing on Apache Aussie, which is.

00:29:58: And they also have an ontology describing business data with entities of the properties and metrics in combination.

00:30:08: when you combine these three standards together basically get that core thing where he say who owns it?

00:30:16: What's the interface for the guarantees And what does the data mean in combination?

00:30:21: We think that's very powerful.

00:30:24: It sounds like it!

00:30:26: Who is working on these standards, who is defining?

00:30:28: how can I picture this?

00:30:31: Well for a second...I'm part of the Technical Serial Committee at the BITO project with the Linux Foundation for ODCS and ODPs.

00:30:38: so there are many large groups from different companies Consultancies vendors and users that drive that none under the umbrella under a process.

00:30:54: Let's say at The Linux Foundation, this same holds for Apache Aussie which is at the Apache foundation although it was originated from Snowflake or many companies including us taking part in that steering.

00:31:15: that would also become very big

00:31:20: as well.

00:31:23: The standard is only as powerful it's being used and we all know there are a lot of failed standards in the IT industry, how do you see development here?

00:31:35: Obviously you're a little bit biased to understand but still have an open mind with open eyes.

00:31:41: How does things develop here?

00:31:46: especially around the open data contract, we see a lot of adoption.

00:31:50: But whenever I talk to some companies or very big company is that A lot of people use it already internally but they don't really talk about it and don't publish on their blogs.

00:32:04: for some reason We have over one thousand GitHub stars at the repository.

00:32:11: so there might be an indicator And but maybe more interesting is.

00:32:16: there's also a popular open source tool around data contracts to automate upon that the Data Contract CLI and That's driven.

00:32:25: Also implemented maintained by us, and it has about one point four million downloads per month.

00:32:31: and that kind of shows that as some usage obviously they are now might also be a proxy indicator Of adoption.

00:32:41: and we also see that huge enterprise.

00:32:44: much larger companies than us are also adopting ODCS.

00:32:48: I think Databricks announced it, Action announced that there's something and smaller companies like Colibra in the data catalog space starting to adopt them.

00:33:02: so we see these market signals.

00:33:06: The other standards are not there yet.

00:33:09: We're ODPS, but when you start to adopt ODCS, ODPs makes a lot of sense because it works well together.

00:33:17: so I assume that this will come and we see a lot bus around Apache Aussie now Not a lot real signals yet by the bus.

00:33:28: So i can't really say about adoption.

00:33:33: Maybe in one sentence, why should I follow a standard?

00:33:38: What's my benefit as the

00:33:40: user?".

00:33:45: I think it reduces my cost.

00:33:48: I get a lot of benefits and can use open source tools that are available.

00:33:57: A follow-up path that has been threatened by others.

00:34:06: So maybe a few sentences for that?

00:34:09: Well, and for many lowering cost is very, very significant argument to make so that alone makes sense.

00:34:19: Simon Graves...

00:34:21: One thing we should add today might be also it's this vendor lock in problem when you think of how important that meta data is the data contracts, the data products and this semantics where we talked about it with the agents.

00:34:39: How critical that is?

00:34:42: And then you probably realize who maybe I want to control a little bit.

00:34:45: I don't want to head over in some vendor specific format.

00:34:51: That's why also standards help because you need to control that layer so crucial.

00:34:59: We're now building as foundation of AI usage in the future.

00:35:05: And when you start with a login now, it's very hard probably to get out off and then a few years.

00:35:12: Absolutely!

00:35:12: It is very simple... Some people also say I'll give that for free so we can get your logged-in after some time.

00:35:21: We see this movement as well.

00:35:22: So think about meta data serenity.

00:35:26: It's also cost if we come back to the costs savings, but on a long-term run.

00:35:36: No I completely agree and i just want to mention that We have a lot of discussions around data sovereignty But it is important not only talk about data residency.

00:35:46: So do I control over my data because its located maybe in my country?

00:35:51: But standards are equally important.

00:35:53: so can I access with how much login just from the way that data is stored in terms of metadata and accessibility.

00:36:04: So I completely agree.

00:36:05: Simon, unfortunately our time has uped.

00:36:07: we could talk much longer on this but it really enjoyed the talk as i think we understood where data products and data contracts are why they're so important also how to do it?

00:36:21: How to assign ownership or technically a comprehensive overview.

00:36:26: Last word is on you or you have it.

00:36:30: Maybe give us a quick input, so if someone wants to bring this forward in their organization and enterprise what are the most important things to address?

00:36:43: What's key success factors?

00:36:46: I think that the easiest way start actually just introduce data contracts maybe into new use project.

00:36:56: we have issues at a boundary between one team and another team.

00:37:00: Then just introduce the contract, automate with open source data contracts CLI in combination.

00:37:08: you will probably immediately see some benefits.

00:37:11: then go from there.

00:37:12: I think that's my best advice.

00:37:14: Excellent Simon thanks so much!

00:37:17: See ya soon bye-bye.