In this episode of In-Ear Insights, the Trust Insights podcast, Katie and Chris discuss the important issue of who owns your data. They talk about why companies need to be aware of who has access to their data, especially when using tools like Google Analytics or generative AI models. Katie and Chris explain how you may be giving competitors access to your data indirectly through models trained on your data. They offer suggestions for ensuring you maintain control and ownership of sensitive company data when leveraging new technologies like AI. Tune in to hear their examples and recommendations for governance, collaboration, and setting up the right data architecture.
[podcastsponsor]
Watch the video here:
Can’t see anything? Watch it on YouTube here.
Listen to the audio here:
- Need help with your company’s data and analytics? Let us know!
- Join our free Slack group for marketers interested in analytics!
Machine-Generated Transcript
What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for listening to the episode.
Christopher Penn 0:00
In this week’s In-Ear Insights, let’s talk about data specifically, who owns your data and who’s using your data.
We’ve had a couple of interesting things happen recently one, we’ve had one client who wants to do a specific kind of tracking on their website and Google Analytics 4 simply cannot do it, you would have to make some significant modifications to the code of your website to accommodate GA for as opposed to making the tool work for you.
That’s one aspect.
Second aspect ChatGPT recently announced its ChatGPT Enterprise Edition, which we’re pretty sure it’s going to be reassuringly expensive.
And the promise there is that your data will not be used by OpenAI to train their models.
This goes in tandem with a huge number of lawsuits and questions and things going on.
about who’s who owns the data that these these models are being trained on.
That also goes with questions we’ve had in our Slack groups like our analytics market, a Slack group, which you can get to at trust insights.ai/analytics, for marketers, someone asking, Hey, I have some interest in a custom model for this.
And our first question is, do you have the data because if you don’t have the data, you can’t tune a model.
And you probably don’t want to be making part of the secret sauce of your business based on data that you don’t own that you didn’t put into the model yourself.
So all of this to say that right now, companies are at an inflection point, if they if you don’t own your data, or at least a good portion of it.
You may be at risk not only from just not beyond do basic things like analytics tracking, but may not be able to take advantage, say of generative AI to its fullest capacity.
So Katie, where should where should the decision makers be thinking about data ownership? Where should where should where should their heads be?
Katie Robbert 1:54
That is an excellent question.
As I’m sort of, I’m running through different scenarios in my head, that’s sort of where the policy is coming from.
Because I’m thinking about things like CDP’s, I’m thinking about things like, even as simple as Google Analytics, if you can’t export all of your data about your digital assets, then you don’t actually own that data.
And so it’s, it’s tricky, because not every company has the resources to build their own systems in order to collect that data.
It’s just it’s not scalable.
It doesn’t make sense.
That’s why there’s such you know, a deep bench in the in the mahr tech world and of software systems that do this for you.
But I would say one of the first things I would look at when evaluating a tool that collects my data is how do I get the data out of it? Or is it something that I can’t get the data out of, I think of LinkedIn as a really good example.
And so, you know, I can post six times a day, every single day.
And is that technically my data? Well, it’s my work, it’s my action, it’s the things that I have done, but I’m doing it on their platform.
And therefore, my work is their data.
It’s not my data, it’s data around my actions, but they still own it.
And I don’t have a right to try to extract that data from them if they don’t want me to.
And that’s sort of where we need to be thinking about, you know, if we want to be doing deeper, more focused analysis on customer behaviors on, you know, potential sales and revenue models, and all those things like how do we get that data? And that’s what I started thinking about.
So when you’re asking the question to go back, you know, what do decision makers need to be thinking about? It’s that it’s, what am I going to need to do the data? Why am I collecting the data in the first place? Why do I even care about this data point? I may not know what I’m going to do with it down the line, but I should be aware that it exists and how I can access it.
Christopher Penn 4:13
And I think that question of who owns your data really is, is at the crux of it.
If you are, you are the CEO of Trust Insights.
We have a lot of data on a lot of different systems.
You know, who owns that data? Do we own it? Does OpenAI? Own it? Does Google own it.
And for the things that we want to do, as a company strategically, what do we need to own that we currently do not own LinkedIn, like you said, is a great example.
Even our analytics, so we run the matomo operate analytics system, which is open source version of Web analytics.
It’s not as fully featured or as capable as Google Analytics.
But we own it, it’s on our servers.
We have access to the back end database to the raw data itself.
So if we wanted to, we could build models and things on which you you can’t do nearly as easily with a system like Google Analytics.
Katie Robbert 5:09
And, you know, I think that there’s, again, it’s sort of the you can’t build everything yourself, you know, we’re even Google relies on third party tools, probably like a CRM system or marketing automation system.
I’m just making assumptions.
But so, you know, even they don’t wholly own all of their data.
And so it really comes down to thinking through like, the basically the prioritization, what data do you absolutely have to have? So revenue data, for example, you probably want to own that data.
And so, yes, you can use a third party system like a QuickBooks or some other accounting tool, but you need to be able to work with that system to extract all of it in order to understand what’s going on with your business.
So that’s data that you need to be able to own do I need to own all of my social media data? Well, that’s not the kind of business that we operate.
So for us, it’s not as essential.
But if we ran a social media agency, that’s probably a higher priority.
For us, you know, thinking about things like Google Analytics we had, I don’t know that we had an easier time extracting data from Universal Analytics, but I think it wasn’t as riddled with issues as it is with Google Analytics 4 and their, you know, data API, where the data is different.
And that’s a question we get a lot is why is my data different in the system versus when it’s in liquor studio versus, you know, when I extract it through the API? That those are considerations? When you think about do I need to own this data? Do I need to have it all the time? What am I going to do with it?
Christopher Penn 7:02
You wait, you raised a really interesting point, particularly around social media data.
And this goes into the way generative AI functions in AI models in general.
One of the considerations that decision makers need to think about in terms of who owns your data is also then who has access to your data.
So if you are using a social media marketing tool, and they say, Hey, we’ve got generative AI that’s custom trained to help you write better social posts, right? That model was trained on something, if you are using, I’m going to make up a name, you’re using, you know, spam, social, at spam, social that has its own custom trained model, and spam socialist training on everyone who’s using the system, then by default, it’s trading on your data, which means that you are also by extension, giving your competitors access to a derivative of your data.
So if you write really great, you know, social media posts, and you spam social to post them on LinkedIn, and you get great numbers on it.
If spam social is using that data to train its language model, to customize it, and you have a direct competitor, also, using spam social, you are helping your competitor write better social content based on your data.
Katie Robbert 8:19
A little bit tangential, but it strikes me that every once in a while, like every few months or so I’ll start to see Facebook posts where, you know, users members will have these long posts of I do not give Facebook authority to use my pictures to use my thing, copy and paste, blah, blah, blah.
And it always hurts my heart a little bit.
One, because there’s no convincing them that that’s not how it works.
But to when they signed on to the platform, and did not read the Terms of Use.
They already gave up those rights to their own data.
And, you know, saying like, it’s like the old Michael Scott, I declare bankruptcy.
Well, that’s not how that works.
You can’t just declare it, and all of your money troubles go away.
There’s a lot that goes into it.
And when you’re on a social media platform, just by being on there, you’ve already given up a lot of your data.
You don’t own that data, that platform.
You know, even though they say we’re not going to sell your data, we’re not going to there’s a lot of loopholes and legal language with that.
And so it’s the, you know, it’s the same with any system with a CRM, you know, let’s say, you know, Salesforce for example, your data may be behind a password, but it lives on their servers, and they may be using your data to train large language models.
Christopher Penn 9:51
Exactly.
And so the question is that and you know, you’re loophole examples exactly right.
Facebook may not sell your data, but they train a model.
They based on your data, and then they sell access to that model, and they let competitors use that model.
And this is not even just for language models, this is for classical AI machine learning as well.
Selling access to that model takes essentially the important part of your data, the part that says this is what works and sell access to that.
And so I think for, for decision makers, for, you know, the CEOs of the world who are looking at the AI space, these are considerations that you got to take into account, who owns your data? Who has access to your data? And who is giving derivatives of your data to your competitors?
Katie Robbert 10:36
Well, and this is where, Chris, you’ve talked about the cautions with using tools like ChatGPT, and putting certain kinds of data in there.
So people are asking, you know, any and all kinds of questions of these tools, you know, you know, here’s my sales forecast for the last year, what can I do better? Well, guess what, they have that data now? You know, here’s my list of customers, who else, you know, what, what is the profile look like? Well, guess what, they have that data now.
And it’s going to continue to use that data to, you know, build recommendations for other people, including your competitors.
And so the solution there, and it’s not an easy solution is to build and train your own large language model.
But again, that goes back to, that’s not a reasonable thing.
For a lot of companies, they don’t have the resources, they don’t have the infrastructure to build their own CRM to build their own web tracking tool to build their own, you know, large language model trained only solely on their data.
You know, there has to be versions of that based on where what kind of industry you work in.
So healthcare, for example, it’s you kind of have to they’re forced into it.
So there needs to be a roadmap for the healthcare industry, there needs to be a roadmap for the financial or the legal industry.
But for B2B marketing, we shouldn’t be dealing with PII in the first place.
So we’re probably lower on the list of industries that are going to have access to our own tools.
Christopher Penn 12:20
I mean, we do though, if you think about it, if you go into Hubspot, and you go to your marketing automation system of choice, it’s swimming and bi, right.
You know, people’s names and job titles, their email addresses, their phone numbers, and stuff, you have stuff, you have enough data in there that if it leaks, you would have an identity theft problem.
That’s sort of the standard is, if it leaks, could someone compromise someone else’s identity with it? And as marketers, we don’t think about that, but it’s 100%.
True, we do.
So I actually got a question very recently about this about accounting, how you how do you set up systems to use language models in the accounting world, given that regulations are changing so often, and also the sensitivity of the data? And there’s, there’s sort of two approaches to language models in general, one is the perfect memory model, hey, this model is gonna get trained on everything.
It’s got to know everything, we asked questions, it gives us answers, that’s a very sort of older approach to these systems now that it’s still effective, but it contains a lot of problems.
And the newer model, a new approach is the hybrid model where you use the language model to do what it does best, which is language, and you connect it to your data sources.
So that it’s not, you’re not asking it for knowledge, you’re just asking it to write language that can interface with other systems.
So in the accounting example, you would have a database that you own, that contains all your accounting regulations contains your customer data, maybe.
And then you have some code glue, that glues it to a language model.
So you ask the language model questions, it goes and queries the database, the database returns answers, the language model converts those answers back into into regular language, the the model does not know information in the database, the model just knows how to speak.
And that sort of is the the architecture that a lot of companies out approach taking.
Because for one thing, it allows you to pull the model out, put another model in.
So as new models come about, you can swap in and to a creates those firewalls that allow you to say, okay, yeah, model your, we’re not going to give you this information, you just have to interpret the results.
And that’s about allows you to run a language model locally on your servers when the open source models.
They don’t have to know everything at that point.
They just have to be able to interpret language well.
And so that protects your company and protects your data.
Because now your model is running on your servers.
Your database is running on your servers.
And at no point are you handing this very sensitive information to a third party that may or may not be trustworthy.
Katie Robbert 14:58
So you’ve mentioned a couple of things.
so that I don’t believe are accessible to everyone.
So one is, you know, having the skill set to stand up a large language model that can have, you know, interchangeable, you know, training models, that to me strikes me as a skill set that not, you know, that account manager or, you know, a, an analyst might have, but you would need a data scientist.
So that’s problem number one.
Problem number two is you’re mentioning your servers, how many companies do you know not talking about like enterprise sized companies.
But like, you know, companies similar to ours, have their own servers or even know where to start to stand those up, because even those technically belong to someone else.
They’re our servers.
But we have to get the equipment from somewhere, we have to get the server space from somewhere.
And so I’m not trying to say that it’s not possible, I’m just trying to understand how it’s accessible?
Christopher Penn 16:04
And that’s a good question.
So it depends.
It depends on the sophistication of the company.
And because all most companies do have their own servers, even if they don’t know it, right, if you have a website, you have a server somewhere.
And the data protection on those typically is higher, have a higher standard.
Because that’s straight up it, that’s not even marketing anymore, that’s just straight up.
It is like this server is governed by these things.
One of our one of the Trust Insights services sitting on the shelf over there, it’s this little box that I have plugged in so and that’s why I do a lot of experimental stuff it because I got this thing on Amazon’s Black Friday sale last year, is it production ready, I would not put that into production, like as a for a client, because it’s, you know, the cat could knock it off.
But yeah, for companies that have these problems, like health care companies like finance companies, they already have either agreements with cloud hosting providers like Microsoft, or Google Cloud, whatever, that have very strong data protection, because they have to, or they have, you know, the legacy company, they probably have a room full of hardware somewhere that you know, with, with all the air conditioning stuff, so the companies do have these resources.
But this is where, again, this goes back to the CEO level stuff, not just marketing stuff, you’ve got an IT department, you’ve probably got a CIO, if you’re a larger company.
And this is where you need to have governance and integration across your company.
So that what marketing wants to do with data fast, it has to be integrated with what it says you can and can’t do, has to be integrated with what legal says you can and can’t do.
And all those departments have to work together.
And so your job as the CEO is to make everyone work together.
Katie Robbert 17:51
Easier said than done.
But I think that that’s a really good point, because I don’t think that there was that initial realization of how cross discipline, this kind of, you know, new technology really was gonna be so, you know, if we look at Google Analytics, for example, installing Google Analytics might involve your IT department, you know, to get permission to include it on the website, it might include legal to read over, you know, the Terms of Use and the data retention policies and say yes or no, or tweak them.
But then once all that is done, those two teams are essentially not needed to run Google Analytics anymore.
You need your marketing team, you need your data team.
But in this instance, when we’re talking about something like generative AI, legal and it can’t disengage from the process, they need to be true partners with marketing and data and sales, and everyone’s in.
So in some ways, it’s really an opportunity to introduce that collaboration.
When I worked.
The company that I often refer to a million years ago, that was a big problem.
Because I T we had the server room, my desk was near it, we had the server room, you open the door, you hear the whoosh of all the air conditioners and the server’s running, you know, it’s dark, you see the lights, it’s very like gothy and emo.
And so you see everything and our IT person was very easy, like all of the cables were perfectly aligned and like very, actually very aesthetically pleasing to look at.
But God forbid you asked this person to sit in the same room as development or sit in the same room of pretty much anybody.
And you knew that you had to like eat your Wheaties that morning and make sure you had six cups of coffee because the gloves were going to come off because the IT person definitely didn’t feel like anybody else understood how data worked, and just didn’t want anybody touching it, like, just don’t touch it.
You’re not I just don’t I don’t even want to talk about it.
No, don’t ask me anything, just don’t touch it.
And it was a very contentious conversation.
And so we had a lot of opportunities for collaboration.
But there was so many personality clashes.
So, I mean, you know, to be fair, it wasn’t just this person who was sticky, every other team felt the same way about the work that they did, it was a very siloed company, even for only being about 50 people.
And the collaboration was nearly impossible.
Because nobody was willing to understand why somebody else should know what they were doing.
And so a lot of companies I think, are going to be facing this cultural challenge right now, of their teams have been conditioned to be so siloed to think that their thing is the most important thing, that when you ask them to share their toys in the sandbox, they’re gonna start hand slapping, it’s not going to happen.
They’re gonna it’s gonna say, No, you can’t you can’t come into my area, you don’t need to know what I do.
That’s not your business.
And so when they’re trying to introduce technology, like generative AI, that requires that collaboration, I’m just seeing a whole lot of headache.
And that, Chris, to your question, that’s where the CEO has to start, the CEO has to say, how am I going to get all of these really strong personalities to sit in the same room together and not, you know, end up as an MMA match?
Christopher Penn 21:39
The answer to that, and this is kind of a management consulting trope is you introduce something like generative AI or digital transformation, or business process reengineering, or Six Sigma or something.
There’s some initiative that has a banner and a flag, maybe this signs up in the office, you know, classic office space stuff.
Katie Robbert 21:59
Oh, you’re not kidding, like a literal banner and a flag to wave like, I’ve exact on that.
Christopher Penn 22:05
Exactly.
But you do that, because that gives everyone a purpose and with generative AI, because so new people don’t have a grudge to fight on the gender of AI stuff itself, right? Everyone’s like, Okay, well, the CEO said, To do this, we got to do this.
And it can create, it can, if it’s done well, you can start to create those collaborative links between organizations that then have a halo effect, you know, hey, the folks in it are starting to actually talk to the folks in marketing in a non non adversarial way, on specific projects, that you can help create those relationships within the organization, then that again, that can spill over into other areas where it’s no longer as contentious you know, for IT and marketing say, Oh, by the way, you know, our CRM is still kind of a mess, we still need some help with this.
So it’s, it’s a good excuse.
If it’s managed? Well, it’s a good excuse to to, to knock down some of those silos at least a little bit, at least, you know, they build some planks, catwalks between the silos.
Because going back to the original question of who owns your data, yeah, in a large organization, especially your data ownership is shared across all these different silos.
You, even if even in the ideal situation, your company has different internal owners of data.
And to your point, they can be very, very protective of that.
Katie Robbert 23:32
I appreciate how naive and optimistic you are, when it comes to people management, like what you’re doing, well, you know, you’re not wrong, like introducing an initiative that’s new, and everybody can get behind is a good way to approach it.
But it’s not as simple as that, which is why a lot of times companies will bring in a neutral third party, not just to introduce the initiative, but almost to just referee the initiative.
Because it’s not that the people internally can’t manage the situation.
It’s that there’s, it tends to be very emotionally charged, even though it’s business.
You know, people take these things very personally.
And so bringing in a neutral third party, like Trust Insights, for example, is a really great way to let them be the bad guy, let them referee the conversations and introduce the initiatives that yes, you are capable of handling.
But it takes you out of the conversation as you know, oh, well, you know, you gave sales you know, 30 more seconds of airtime than you did marketing, you know when talking about what they own, so they must be more important than marketing.
Let the third party consultant take that heat.
Let you know.
For example, like if I were to go into a company, you know, my job would be basically to mediate the conversation and make sure that things are being He talks about new productive managers.
So when people get angry when people get frustrated, the heat’s not on the CEO, the heat’s on me as the third party who’s coming in to, you know, basically turn their world upside down, let me be the person who sort of takes that on their shoulder so that you, the CEO still remain the figurehead, the person pushing this forward the leader that everybody needs to look to.
So basically, my job as a third party is to make you look really fantastic.
And I have to take, you know, all of the, you know, bad feelings and curses and insults.
But that’s fine, because then you add a zero to the, you know, agreement, and everybody’s happy at the end of the day.
Christopher Penn 25:47
Everyone’s happy then today.
But also, those exercises allow you to understand to get better saying who owns your data, right, because when we do those kinds of explorations within companies, we often find out that there are significant data ownership problems, one of the ones that comes to mind is we worked with a biotech company that had three separate, independent, non manage instances of Salesforce, they were paying for salesforce.com, three times, their leads, and their opportunities were scattered across these three systems, they had different scoring systems at different opportunity stages.
And when we rationalize that, it basically they lost because of these discontinuities, they lost track of a billion and a half dollars in pipeline, because no one could find it was if it was in one system, and the person was looking for that.
But no, it’s actually the other system.
This is a case where the data ownership wasn’t necessarily in question because the company owns the data they put in a Salesforce, but they had no idea who owned the data internally.
And so when they rationalize to finally they’re like, Oh, oops.
Katie Robbert 26:55
And you know, and that is a big part of these exercises, is to your point, Chris, is using them as an excuse to get these people together to talk about who owns the data.
So for example, you know, if we go back to the Google Analytics instance, you may find that there’s a difference of opinion between IT development and marketing of who owns the Google Analytics data, it might say, Well, you’ve put it on my servers, my website, developers might say, Well, I had to actually do the work to implement it.
And marketing might say, but I’m the one who needs to use it.
And so everyone in that instance, is right.
However, it needs to be more collaborative than that.
So there are opportunities to find collaboration, collaborative path, but it absolutely needs to be if you’re introducing something new, like generative AI, that needs to be your first order of business is understanding what that governance looks like.
Because there’s so many ways that it could go wrong,
Christopher Penn 27:58
and very often does, often does.
So, to summarize, you need to know who owns your data internally and externally, where it lives, if you want to be able to use it for generative AI if you want to be able to use it for any new initiative where you have opportunity, but you also have risk.
So you want to de risk yourself, you want to de risk your projects, and you want to ensure that you can make use of the technologies.
Before you even start talking about which system and architecture you need to know Hey, who owns this stuff internally, because if you don’t, it’s gonna get really really hard to to make a project successful and comprehensive.
And the bigger your company is, the more of a pain is going to be.
If you have comments or questions or projects you’ve done just to try and figure out who owns your data and you want to share some thoughts or ask some questions go to our free slack group go to trust insights.ai/analytics for marketers, where you and over 3500 other marketers are asking answering each other’s questions every single day.
And wherever it is you watch or listen to the show.
If there’s a challenge you’d rather have it on instead go to trust insights.ai/ti podcast, you can find the show on most channels and while you’re there, the channel of your choice please leave us a rating and review it does help to share the show.
Thanks for tuning in.
I will talk to you next time.
|
Need help with your marketing AI and analytics? |
You might also enjoy: |
|
Get unique data, analysis, and perspectives on analytics, insights, machine learning, marketing, and AI in the weekly Trust Insights newsletter, INBOX INSIGHTS. Subscribe now for free; new issues every Wednesday! |
Want to learn more about data, analytics, and insights? Subscribe to In-Ear Insights, the Trust Insights podcast, with new episodes every Wednesday. |
Newsletter
Podcast
