Thank you for joining us for this 2nd session.

And thanks, James, for having me.

Last time, we were talking a lot about AI.

It was a great lesson for me, actually.

Um, an issue that affects the industry that I work in a lot.

So it was really incredibly useful.

And it was as much about sort of how rights holders can try and regain some control of their content.

Now, some of the methods that we talked about last time, which involved, some understanding of the device that was being used, also some understanding of the location that the person or that the crawler may be accessing that information from, today we're going to do a much deeper dive into those issues, aren't we?

We are.

We're going to be explaining a little bit more about how that works.

So you can ask questions. You just pop them in the chat window and they will pop up over here and we can come to any questions that you've got a little bit later on.

But 1st of all, let's dive in with the devices themselves.

So I'm holding in my hand a device.

How can you tell what this device is and why does that information matter?

So, let's do that in that order.

So whenever a device accessing the internet or the web performs that activity, so in the course of carrying out the access to a web page or an app doing something, it sends metadata effectively about the connection and the type of device.

So when it comes to device detection, the 1st thing we do is use that information.

And there's been a particular field in the internet standards, pretty much since year dot, called user agent, or it's a user hyphen agent.

This is talked about quite a lot in the industry, talk about user agents, and it's simply a sequence of characters that by convention contains information. About the operating system, the browser, whether it's a crawler.

So that was one of the things we talked about last time with robots.txt.

It might contain the model or something information about the model of device.

And then once you know the model, and you know the coding scheme that the vendor's using for that model, then you can start to work out, oh, okay, this is a TV from LG, and this is the model of the TV, or this is a Apple device, for example.

Then we have something else called user agent client hints.

So kind of everything was going on quite normally until the turn of this decade.

And then Google decided to make it a little bit more complicated.

So they added some other fields.

Now, these ones don't get sent with every single request.

These ones, some of them get sent with it every single request.

Some of them you have to ask for.

So there's a little bit of backwards and forwards that needs to occur between the client device and the web server effectively.

And then for other devices, we also use some JavaScript as well, which basically runs on the device and gets more information that we can use in order to narrow down what the device might be.

That's particularly useful for Apple, who don't really support those 1st 2 requirements.

So device detection, say 25 years ago, was relatively simple.

Virtually they were not actually that many devices in the world, that time relative to today.

And where they were, they all used this thing called user agent.

We now have user agent client hints, and we now also have the need to use JavaScript as well.

So technically it's become a lot more complicated.

And keeping on top of that, is really important because these devices and how they behave change all the time.

Now, why is that information important?

So, firstly, optimisation, you don't necessarily want to serve exactly the same content to a 100 square inch screen as a 6 square inch screen.

When it comes to analytics, you might want to know where you're getting good outcomes.

So suppose on a small screen device, you've got a call to action and on a particular permutation, it's just off the screen.

So users don't necessarily know to scroll down.

For example, they might miss that visual clue.

So the analytics can identify that and then the optimisation can then be used to deal with the problem.

Then you've got fraud, which is part of the same problem we discussed a fortnight ago around AI.

So we were talking about AIs. Pretending to be humans in order to get content, but they may not, the content owner might not want them to have access to.

But then you've got fraud, that might be being perpetuated, so that could be ad fraud, or it could be survey fraud, or even more serious forms of fraud where they're trying to sort of extract credentials effectively from, you know, from a computer service, for example.

So the device information can be helpful there.

A really, really powerful property is the age of the browser.

So if the browser's more than a couple of months old, these days, and it's not the current version, these days with the way the application stores work, and computers work, they tend to automatically upgrade themselves, when did you last do an upgrade, for example, it just happens, happens in the background, doesn't it?

So if a new versions come out and it's been out, let's say, 3 or 4 weeks and the old one is still being used, then that might be a little bit suspicious.

We also see when it comes to fraud, some of the fraudsters who are not particularly smart, who move to another version, which doesn't really exist.

So being a bit of a tangent for a minute.

But sometimes we get customers coming to us and say, oh, you're not reporting this version correctly and we go, well, that's right, because it doesn't exist.

Just because it's contained in a user agent, doesn't necessarily mean it's a real version, for example.

So that's a really important indicator for fraud.

You've then got what's happening on the on the internet.

So what devices are popular, what's not.

So that's kind of more for macro. Level analytics.

That information then also can be used in advertising.

So the way that advertising worked is that when the advertiser is presented with an opportunity to serve an advert, they are not having direct access to the device.

So this information can be provided to them, and the more information they know, they might change their behaviour.

So they might be prepared to pay more for an advert, say, for consumer electronics on a modern piece of consumer electronics, than an old piece of consumer electronics, for example.

So that can be quite useful.

Conversely, they might use that information to basically verify the reputation of a publisher.

So are they pretending something, is the publisher pretending to be connected television when in fact they're not?

You alluded briefly there to the kind of the huge explosion in proliferation in the number of different devices that have been over the last 25 years.

And my assumption is that every week there must be something new hitting the market somewhere in the world.

How do you keep on top of that?

Well, we're a little bit old-fashioned.

We have lovely, lovely people that work at 51Degrees, and part of their job is to research new devices that come out, so we do not scrape the web, using automatic crawlers and things like that.

I think that would be a bit ironic, given our position on scraping that I think is pretty well known.

So we research those devices.

And what customers can do is they can share usage information with us.

So that information I talked about earlier, user agents, user agent, client hints, JavaScript.

They can share that with us.

We can then aggregate that information and say, okay, well, what's new today?

Yeah?

So if it's an operating system, for example, or a browser, we'll see, oh, okay, there's a new one that's come out.

If it's a popular one, we would have probably got to it in the developer preview.

Sometimes there are security fixes that come out overnight or the long tail of apps.

We don't always research those proactively.

There's quite a lot of apps out there in the world.

So we take that information, it goes onto a queue, and then we research them broadly in the popularity order, which basically defines importance.

That might mean we have a new device we need to add.

So we'll go to the vendor and get the information about that device, and then qualify that with other authoritative sources, basically, but it's all done by real people who do an amazing job.

And our customers love the fact that they can contact us and if they want, we can explain exactly why we've entered some data the way that we have entered it.

Great.

So that's devices, sort of a mixture of active and passive means of working out what device is handling this at any particular time.

If we move on to location now. I suppose it's the same question again in a way, which is how do you know where I am? And why is that information important and how granular can you get with it?

Well, there's 2 aspects to location.

So there's the 1st part of your question, which is how do we know where you are?

And then how granular can we can we get?

So we work with 2 pieces of information.

The first, which I think is relatively easy for most people to understand, is the latitude and longitude as reported by the device.

So that could be from a satellite GPS.

Very accurate. Could be from network triangulation, might even come from Wi-Fi, depends on the device that's being used.

So we can take that latitude and longitude and we can turn that into a country and possibly even a postal address effectively depending on how accurate it is.

So that covers location.

But in order to get access to the precise and accurate latitude and longitude, that is not something that people are typically happy to do straight off the bat.

So if you visit a website and the 1st thing it says is tell me exactly where you are, especially if you don't know that website and that brand, you might be a little bit taken aback.

It was like a bit early to be asking that question.

So the IP address can be used in order to form an approximation of where someone is.

Now that's going to vary when it comes to location quite significantly.

So if you're on a cellular network, for example, it can be a very large geographic area.

If you're on residential broadband, it depending on how the operator works.

Could be using something called carrier grade gnat.

So this is where lots and lots of houses are kind of sitting behind the same network IP address.

So actually it might cover a lot relatively large geographic area.

If you happen to be on a specific hotspot in a specific Starbucks that's only used by that particular Starbucks, then it might be more accurate.

So the 2nd part is when you're working with IP address is the range of accuracy is huge, enormous.

So it can be useful for certain things to establish a context.

When it comes to the use cases, once you can work out the probability of a country, and you say, okay, well, it's probably going to be this particular country where someone's in, then you can do things like customise the pricing, so you might use the currency for that particular country when someone goes to, say, a pricing page, for example.

They won't even notice it's happening.

But if you always present US dollars, for example.

There's a lot of countries that aren't necessarily using US dollars, so you could use a currency that's more natural to the people of the most likely country.

Then we've got network.

So that can be quite important.

I touched on the different types of connections.

So residential, retail business, data centres, those kind of things.

That's important as well.

And then you've got human probability as well.

So just like I was saying to you earlier about the chances of a very old browser or an old browser, being a real person.

There's certain IP addresses, like, for example, you're getting what's supposedly an iPhone coming in from a data centre.

Well, it's a data centre running an iPhone, that sounds like it might be a proxy or something like that.

Again, you might treat that, treat that differently.

So, you know, those are the sorts of use cases where location is important and it varies depending on the business problem that's being solved.

So, you know, the convenience of providing currency is one thing that if you're, say, a streaming platform and uh, your licensing content from a particular organisation, uh, there might be restriction on that content to say, well, you can only make that available in these countries.

So you really want to be taking steps to make sure that you're not making that content available in breach of license, for example.

Great.

How does 51Degrees approach device, location differently to their competitors.?

Well, where to begin.

I think the key thing about us in the market is we came into these markets after the establish competition.

So what that means is we're able to learn from their mistakes.

And I think that that's been really helpful and that's allowed us to, firstly, we've paid attention to that.

And then we've been able to innovate around, you know, areas that we think are weaknesses in the market.

I explained a little bit more as I sort of progressed through the answer.

So let's take each of them, let's take both products together.

So I'm an engineer by background.

And I realised that what we need to be able to do is have a very, very efficient way of consuming this data, because it's used at scale.

Yeah?

A lot of this information can be accessed through a product like NGINX. NGINX is there's others HA proxy, Varnish, other networking equipment, sits on the edge of the network, and it's the sort of system that traffic flows through.

So you need real-time enrichment.

So it's got to be really, really fast.

So our engineering is built for speed.

We use something called C as a programming language, a bit old school, but everything runs C.

So we wrote things in C because it's just fast and it runs everywhere and it means we can take the same code base and use it in NGINX, which is, as I say, a very high performance piece of networking software or WordPress, which is, you know, used by many creators, many publishers.

So we get that benefit in our code and our code base.

And what we realised is that we can structure our data so that it can be loaded into memory or accessed from a disk very, very efficiently.

So what that means is you get very fast startup times and very efficient use of data.

So you can start up quickly, the data sitting there in the same memory space.

So the same computing space affectively is the application that's consuming it.

So there's no network traffic, no network calls or anything like that that are going on.

You can still deployed in different ways if you want to.

But that really underpins what we do from an engineering perspective.

I'm very proud of our engineering. We have a GitHub Repository, GitHub.com/51Degrees.

And on the front there we have the CICD dashboards.

So all our code is open source.

We don't hide behind, you know, private repositories and things like that.

We're very happy for anyone to have a look at our code.

We retain copyright on it, but we do license it very generously.

So that means engineers can get up to speed very quickly with our examples.

They can understand what's going on.

They can report bugs and we can do things in the open.

And then we take the same approach to the data.

So we are transparent about our quality metrics.

You go to resources on the 51Degrees.com website.

You'll see device quality metrics.

We update them weekly, so you can see what the distribution is and how things are changing over time. By different facets of the browser, the operating system.

The hardware model, so you can see that information, and we publish our numbers, whether that's the, you know, percentage of tax codes that we have covered, it's not 100%.

I think it's 98% at the moment, thereabouts.

So, you know, we publish those numbers.

So that's kind of part of our DNA.

That's the sort of philosophy of 51Degrees.

When we get into IP location, an IP network, IP human probability, there's a number of things we've bought to that space.

So we've got like location confidence, for example.

So the way we work when it comes to location is the sample of real devices that are reporting those latitudes and longitudes.

And sometimes we can have confidence in that.

We've seen enough data.

We're happy with the result that we're providing.

Sometimes we don't.

So what do we do in that situation, we return low confidence.

We then have medium confidence, high confidence, hopefully easy to understand.

High confidence.

Yep, make business decisions on the back of this. Is good.

Medium, maybe make low consequence decisions, low.

Pretend we haven't responded, probably.

Yeah, you know, something's not right here.

You probably can't rely on the information, and that, it's something that the legacy providers just, just, you know, don't do.

They built their data models before that was being considered.

What we certainly don't do, if there's, uh, when we say reporting country, is just say, oh, we don't know, we'll just use the registered location of the IP address range.

So, again, a popular sort of case that comes up, we deal with it fairly early on is, oh, these devices are coming back in this particular country.

But others are saying it's in another country and we say, well, the other country is the registered country.

We're pretty confident, if it's high confidence that we are seeing information from this IP address.

So maybe there's a VPN that's being involved there.

So we can bring that, you know, bring that to bear on it.

We have human probability as well as an attribute.

So 10 being good chance this is human.

One being definitely not treat them like our bad actors.

Yes, absolutely.

In the AI space.

We've got diversity now.

This is an example where we bring the 2 products together.

So looking at the diversity of hardware of operating system of browser or app by a particular IP address or IP address range.

So that can be quite useful to say, okay, well, this is a cellular IP address.

It's got a very high device versity.

That would kind of make sense.

It's a data centre with a high device diversity.

Yeah, well, that seems a bit more suspicious.

So we can provide that information.

And then just to build on kind of what I said earlier about both of these products.

We do use exactly the same technique.

So we have a highly customised data file that can be loaded straight into memory.

It's an engineer, the amount of data in IP location is far greater than device.

So device data files.

Small, IP foils big.

But we can use exactly the same technique.

So we can be super fast with that in process.

A lot of competitors will require separate servers and custom databases and things like that.

So we make it easy to consume.

Oh, sorry.

Well, I was I was going to let you talk.

I was going to pick up on the diversity point because that sounds to me, there's a bit of secret source there, starting to kind of weave together this device information in this location, this IP information into a kind of a bigger data set that you can then interrogate as a as a combined as combined bits of information strikes me as potentially quite important and offering some pretty unique benefits, I would imagine.

Yeah, well, there's a reason we're called 51Degrees, and that's because that's the northern latitude of Reading, where we are, and when I set up the business with Ruth, my wife.

We wanted a name.

We didn't necessarily think it was going to be ultimately a neon.

But it's kind of stuck.

And that was really about location because the problem I was having with another business was basically getting this data and being able to act on this data and getting it reliably.

And basically I was finding that there wasn't a solution that was accessible. At, you know, at that time.

So I thought, okay, I'll solve the problem myself, which is kind of the kind of inception story, the origin story. Of where we are.

But it's exactly the same use case.

You look at even the most basic analytics, it's country and device type.

Yeah?

A sophisticated analytics solution will go further.

And now with AI, they need more information so the AI can work out what's going on.

I kind of tear my hair out, looking at, say, some of the streaming companies, where, you know, they're investing in AI, but they're not feeding the AI, started calling this AI starvation. You know they're not actually giving the AI the data to work with.

So you take device for example, understanding the chip sets, because basically devices are just a small number of chips that get produced each year wrapped around with different plastic and different screens but basically it's the chip that matters. Actually analysing data by the chip set.

So you can start to see where you've got QOS problems, for example, quality of service problems.

You know, those kind of kind of use cases.

So really making that data accessible, conveniently, no one wants to do more engineering than they need to.

So we focus very, very much on the examples and making it easy for engineers to get up to speed and work with the product.

I'm very proud of what we've been able to do there.

Certainly our time to onboard customers is now, it just gets better and better every month, basically.

So like bringing all of these services together into one discrete package, it seems to be your approach, why?

What are the benefits?

What are the benefits, say, just to me as the end user in that?

Well, you've got different, you've got those use cases, okay?

So being able to say to the our customer, because we're a B to B business, them being able to consume this in a single product gives them a huge advantage.

They only have one API to consider one contract to consider.

We can, because about 90% of the code base is actually the same between these 2 products, that all the stuff about how do you load data, how do you get data, how do you traverse data, how do you present data back, et cetera.

They're all exactly the same.

It doesn't matter whether it's a country or a device type or a hardware model.

So they get this efficiency in being able to operate a part of their business with us as their supplier, that's just efficient and easy and they can just get started quickly.

Their engineers can be efficient.

We then make it easy for them to, you know, actually do something with that, with that data.

So we can package and present that data in a way that it's easy for them to understand and easy for them to integrate into their environment, whether that's analytics or optimisation or fraud or whatever it, whatever it might be.

So they get they get that advantage.

And then they get, you know, we're the only ISO 9,127,001 company.

We really do care about process and quality.

So we operate to that standard so they can be confident in the service, they can get a single support agreement where they can ask questions about, you know, both data sets effectively and how we work.

They get that open source, and they get the innovation that we've been able to bring where we've not just copied legacy, open source models and data structures, for example, which, again, might have worked well in the norties, but really now with networks being as sophisticated as they are today or device detection, having changed quite a lot.

It's now far more suitable with the innovation that we've bought, to be able to work on these things differently, if I can just give you one example.

I mentioned those user agents earlier.

Well, they used to kind of work left to right with the most important information being on the left.

Now, particularly with connected television, because there's no rules around it.

Sometimes the most important information can be in the middle or on the right hand side, if you look at it left to right.

So our innovation is being able to say, okay, well, we don't assume anything about the data.

We can jump straight to the locations that matter and traverse that data very quickly.

So I think really, the answer to your question is more consumers will notice a service provider that cares about these things subtly.

They might not notice it consciously.

Like, oh, it came back in context of my country.

It recognised my time zone.

I got the right currency.

I got a experience that was optimised for the type of screen that I've got.

Of course, other companies can support that.

What we do is just make it very easy and very efficient.

So that maybe companies that wouldn't have thought of doing that can now do that as a result of using our service.

Would this be the sparkle that you've talked about before?

Yeah, I think that's part of what we're doing.

I mean, fraud is not really sparkle, frauds kind of table stakes.

But yeah, the sparkle.

So you go to our website and the pricing page, creating a table of differences between different subscriptions, for example, there are only five, on a large screen.

You can display that so it fills the screen and it's very, you know, akin to a magazine or a printout or something like that, but on a small six, 7 square inch screen.

That's not going to work.

Yeah?

So pinching zoom isn't a thing anymore, thank God.

But actually how you then present that same data is very different.

So we use a different template on a small screen device on a tablet and on a desktop device because we're looking at how's the best way to present this information.

And we can do that really quickly.

That means that something like core web vitals, for example, which is a measure of performance of a website.

We can respond very well in that situation.

So that means for our high volume customers.

We got some of the largest travel sites in the world, for example, using as travel being quite a complex transaction.

They can then customise the template that they're using reliably.

Right.

How do how do people get started with this?

Well, firstly, you need to have a business problem to solve.

Once you do that, you're going to contact an engineer.

And that's where we would want the engineers to come and work through our examples.

So we've got the documentation, we've got the examples, they're all open source.

There are, there's a way of getting started on the website, so you can get a subscription, or you can contact us, or if you already have one product from us, then you can contact us and we can look at modifying your subscription to include the other.

So IP intelligence, if you've got device detection or device detection, if you've got IP intelligence, or location, of course, we do, we do as well, which is latitude and longitude to postal address.

So engineers can get started very quickly.

But normally they don't just do stuff on their own.

They do stuff because a business wants to solve a problem.

So we tried to put up a lot of information about the problems that we solve.

So advertising, for example, we can improve CPMs for publishers.

Conversely, we can actually spot opportunities to perhaps undervalued inventory for advertisers.

Fraud is a big growth space at the moment.

I think there's slowly a recognition that just saying an IP address might have been used for fraud before, so forms of fraud on its own.

It doesn't necessarily work, although just of sheer volume.

For example, actually the research that we provide in the fact that we can stand behind it is very, very important.

You've got e-commerce, travel, you know, those kind of things.

It's really about establishing that business problem and then recognising that we solve it in a very convenient way that really adds value to the business.

You talked before about, you know, the humans within your business actually actively staying on top of and researching things like new devices, new technology that you need to be able to respond to quite quickly to aid your customers.

My assumption is, and you've hinted at the fact that fraud is as big an issue as it's ever been and perhaps getting even bigger.

How does 51Degrees kind of stay at pace with, I would imagine all the myriad different ways that people are using to kind of to cheat these systems and to spoof these systems.

How do you how do you stay on top of that?

So we we provide the data that we provide.

So we provide refined ingredients.

Yeah?

You're not buying sugar cane?

Yeah, you're buying refined sugar, but you're not buying cakes.

Yeah.

Our customers are in the cake business.

We're in the finest ingredients business, okay?

So, ultimately it's for them to decide how to assemble those ingredients, what recipe that they want to use, and we could do a whole podcast on fraud.

But just, you know, quickly to look at 2 different use cases, you've got fraud to do, we say, advertising.

So trying to pretend to be a legitimate publisher when you're not.

Yeah, to try and defraud the advertiser, for example, that is a hot topic in the industry at the moment.

And then you've got fraud, perhaps a survey fraud, where people are offered a financial reward or something that can be approximated to a financial award in return to completing a survey, or you've got, you know, banking fraud, financial fraud, et cetera.

So each of these things has different, have different characteristics effectively.

What we're doing is making those raw ingredients available.

So we're saying human probability.

Yeah, we're saying location confidence.

We're saying connection type, we're saying the age of the browser or the operating system.

We're saying whether it's a, you know, a crawler or not.

We're bringing all that information to bear.

And then our customer in that space will need to also look at, well, how do they want to use those things?

Yeah, we were having a discussion earlier today in the office, you know, about the analogy with shoplifting.

Well, you know, shoplifting is actually really easy to stop.

You just put airport style security in.

But of course, that creates friction for people.

So when you're dealing with fraud, just like we were talking about AI last time, understanding the user experience that you want to create with dealing when dealing with fraud is also another factor.

So we're providing those ingredients.

And we stay on top of them by looking at the trends, we analyse our own data every day, as I described earlier in this conversation.

But we also look at it at a higher level as well, weekly to make sure we, you know, are spotting, spotting trends and things like that.

And then we work with our customers.

Great.

So, a couple of external questions.

You talked about, you know, the amount of data and the importance of streamlining the data.

How much of it is in the cloud, how much of it is on premise.

Oh, well, customers, so ultimately, we deal with a lot of data.

So we use the cloud, but customers can decide to consume our service from a cloud, implementation, so they effectively cool our service with the, we call it evidence.

So the information IP addresses, user agents, et cetera.

We then respond, and then they get the answer.

What I'm really pleased about and - I think we're going to do a plug for our next webinar later - but I'll save that.

Anyway.

Yeah, what I'm really pleased about is you can do that within the web browser or within the app, but you can also do it on the server and it's completely seamless.

It's basically the same back end, whether you're on premise or you're on cloud.

So that's great.

If you're happy to say, okay, my service can wait 22nd, 20 milliseconds to get an answer, then that actually works very well.

So if you take, you know, insurance fraud, travel, et cetera, that's kind of okay.

For advertising or those edge environments that I talked about earlier for networkingm it's not okay.

You don't want to wait 20 milliseconds.

So you've got on premise as well.

So you can go to the website, go to the pricing page.

You can get the cloud service through that route.

You want on premise, then there's a right hand option for bespoke, and then we can have a conversation with you about what you need.

What's really important to our process of working with the customers is that we're helping them establish the value they're going to get from our product and we're very happy to kind of walk away if there, you know, if there's no value. You know, we obviously want to solve a problem. For them, and that's what builds enduring relationships.

I mean, if I use 51Degrees, can I bring my own data into this?

Yes.

So, uh, uh, when was it now, 2017?

So that's like 9 years ago, isn't it?

We started the design process for version four, which is the product that we were on at the moment, the major version that we're on at the moment.

And we wanted this flexibility so that it was really easy for people to assemble these raw ingredients effectively.

So you can take our raw ingredients, but you can take others and you can assemble it into a service.

And this is called the 51Degrees pipeline.

We use that ourselves.

It's really only been over the last 3 years that we've started to see customers start using it, but the more the more that use it, the better it gets.

And what they can do is add their own components into the pipeline.

So if you think of it as a flow diagram where data's flowing through a pipeline, then you can add your own bits, right?

So you could say, okay, I've now got some registration information because I've got this particular request as an identifier that comes from registration information.

I want to add that into the data flow as well so that the resulting data I get from this process contains everything that I need in order to deliver my website experience.

You might have augmentation as well that could occur.

So we've got translations, for example, you might say, okay, I want to translate words where they can bring that in, they could bring their own translations.

And ultimately they assemble it all into this single pipeline, which is super easy to configure.

And then they can use that information to add that sparkle, for example, or make the fraud solution.

They take those different inputs and then they say, okay, this is how we want to treat forward in our particular context.

Just an important thing on that.

Again, we could talk about fraud as a whole separate subject.

They're going to see their own traffic patterns over the last, say, 10 minutes.

So actually fraud, it's pretty much essential.

If you're doing fraud, you also need to bring your own data, even if it's just the currency of data, because that's going to have a fact of something that we create that, say, 12, 24 hours in the past, won't necessarily reflect what's happening on your site today now.

So actually that's why our ingredients are part of these bigger solutions.

That leads me on to probably the final question then, which is if my company has questions about the results that are coming back, what's the approach?

How do we work with you under those circumstances?

So customers, well, we offer 2 types of support.

So we're open source, so that means GitHub.

So anyone, whether they're a customer or not, can come on to GitHub and say, hey, this is happening.

We usually respond by sort of saying, you know, great, please can you give us a little bit more information.

It's always helpful as engineers to actually understand the steps and the environment that it was used in order to recreate a particular problem.

That really helps us.

But we don't have an SLA around that, just like any open source project.

Those businesses that are customers can take a support contract.

So we can then help them, and we'll get people on a call with them, and we'll work through and potentially hold their hand through a particular issue.

Of course, we do sometimes have bugs.

We're not immune to bugs.

We do do what I think is seriously impressive CICD.

We test everything nightly and release everything nightly and we do it in the open.

You can inspect the build logs.

You can see all of that going on.

But yeah, there might be bugs and again, we can work through those.

And also another common request is new features as well.

So sometimes ultimately, if I can be convinced that the feature is one that we should invest in and we should make.

So it's one of those, why didn't we think of that, then of course, you know, that's where, in fact, that's where things like device diversity came from and some of our other properties that we're bringing out next month as well, come from that customer engagement.

But sometimes they might be specific to a particular customer, often their technical environment.

So we'll make those changes for them, but of course, we are a commercial business, so we do have to make that worthwhile.

Do you have any hints on things that are coming next month or is that still top secret?

Well, it's to do with our approach to location.

So.. Very relevant to us then.

Exactly.

So we understand these days with the modern web, it's not possible to say, yep, you were on this IP address, you must be in Reading or London.

Yeah, doesn't work like that.

You can't do, you know, spot, circle, right?

You're somewhere within that circle.

World doesn't work like that anymore.

So, we started by returning geographic areas, so these are regular polygons.

If you go to the IP tester or 51Degrees.me.

So IP testers, 51Degrees.com developers, IP tester, 51degrees.me, just comes up with a simple page that will show you this map, and you can see the possible areas that you might be in.

That's great, but not everyone wants to consume that information that way.

So being able to understand the countries that you might be in, turning the singular into the plural, with probability.

They are all hints at what we'll be bringing out next month.

Great.

You've also had it here first.

Indeed.

So great session.

We're going to be back in 2 weeks and things are going to take go off in a slightly different direction.

I will be vacating this chair and it's going to get much more technical.

The next recessions really for the developers.

That's right.

So what we've been doing in this session, if you're watching the recording, I hope you've got a flavour as to how we approach things, hearing it from me.

It's really about referring your developer, or developers to the next session, where we're going to be doing deep dives into how to build these recipes effectively, how to configure device detection and IP intelligence and consume it in a single service.

And they're going to be about this length.

So the technical part is going to be about half an hour.

And we're going to do a live coding session where we go through creating that pipeline.

So do refer to us to your developers effectively. Ask them to tune in to those sessions.

We're going to be covering.net, Java and go 3 of the most popular languages these days for our customers.

And we'll make that information available in usual places, and the sign-up process will be just the same.

That's right Brilliant.

Anything else before we wrap up?

No, thank you for asking me, you know, going through these questions.

Do contact us if we haven't answered your question in this session?

Very happy to arrange a conference call and answer your questions or go to GitHub and ask us your questions there where we're very open and you can see all the code.

We'll see you again in 2 weeks.

Thanks very much.