#120 Marketing analytics updates and a BigQuery health check

[00:00:00] Dara: Hello and welcome back to The Measure Pod, Matthew, second episode. How are you feeling? 

[00:00:20] Matt: Oh, yeah, I enjoyed it, I enjoyed it. I’m enjoying the prep that goes into these talks as well, like forcing me to get out and look around at, at the industry and, and the news, et cetera. ’cause we were, I suppose it’s, it’s worth saying that we, we, we originally had a guest that had to unfortunately pull out at the last minute. So we, we go in it alone, Dara and I. Today, 

[00:00:44] Dara: we, we, we quickly scrambled and scoured the internet for news. And I, I shouldn’t say we, I’ll just be honest. I’ll put my hands up. I’ve done absolutely no prep and you’ve done. You’ve done all the prep, but I expect at least half the credit. No, that’s fair enough.

[00:01:00] Dara: You it’s co-host. It’s not a co-host. Exactly, exactly. But yeah, so the reason why, we’ll, we’ll do a little, we won, we might not always do this, but let’s do a little, a little kind of, bit, bit around, you know, what we’ve been up to. So I’ll start and explain why I’m gonna get my disclaimer out of the way and explain why I’ve done nothing and been really lazy and not put the work in to prepare for this one.

[00:01:21] Dara: Laziness is definitely not true actually. ’cause I’m, I’m, I’m packing and clearing ’cause we’re moving house and as with pretty much everything in life, I tend to underestimate it and, and I underestimate the task and I overestimate my own abilities and my own organizational skills. I’ve been, I’ve been offline and doing, I mean I call it real work.

[00:01:43] Dara: real work manual. Yeah, real manual work. Yeah. Yeah. Although really it’s kind of clearing and packing and moving, and doing lots of, lots of trips in a van. Moving stuff around. Clearing your own mess. Clearing our own mess. Exactly. Yeah. So that’s left me a little bit on the back foot, which is why I neither you, but I trust you.

[00:02:01] Dara: I knew you’d go away, you’d do those homework. So I have complete confidence in you that you’ll help to pull this episode together. 

[00:02:08] Matt: We’ll see, won’t we? I think I did for a moment. Consider just making it completely about snooker, because I think having any sort of podcast during the world championships is hard Not to be sneaker related, but I did go and find some data related things.

[00:02:23] Dara: Well, we are both snooker fans, so I think maybe we need to at least, you know, and we’re also data people, so I think we need to, to, we need to put this out there. If people would rather this became a snooker podcast, then make sure you let us know and we’ll count up the votes. And if, you know, if the people speak, then we’ll, we’ll just switch to snooker 

[00:02:43] Matt: Next episode will probably be shortly after the world Championships wrap up. So why don’t we commit to pulling some snooker data and snooker do some snooker analysis that we can talk about on that. On that podcast, I commit By which commit, I mean, we probably forget about it as soon as this is over and never do it, but, 

[00:03:02] Dara: Well, I was thinking you like, as it’s becoming a theme, you could do it and then I could just take the credit verse. We’ll see. Yeah. Okay. So, so what, what have, what have you been up to then? Have you been up to any, have you been moving and packing boxes? 

[00:03:15] Matt: No, I’ve been very much stationary, sitting during the sun a little bit and, and yeah, genuinely watching a fair amount of books in my evenings, that’s pretty much as exciting as it’s got for me.

[00:03:28] Matt: So has this recording been interrupted? Have you, you basically got sneakers around in the background? I haven’t actually today I haven’t watched any, even though there’s some good matches on today, I’ve not watched any, ’cause I’ve been in meetings, people, oh, some people, people don’t like it when you’ve, you can hear sneaker in the background and you can continuously looking off screen. People are too fussy, aren’t they? I know. 

[00:03:47] Dara: Okay, so we are, we’re gonna cover some news and then we have a bit of a topic, I dunno why I say a bit of a topic. We have a topic that we’re gonna, we’re gonna cover, but we’ll get to that, you know, no spoilers. But yeah, we could do a quick, quick run through some, some news and I, I mentioned, I should say as well, I mentioned last time in our, our first episode together, that we were gonna kind of tweak the format a little bit as we go.

[00:04:14] Dara: And I think this is probably proving that we’re gonna kind of mess around a little bit with the, with the upfront sections of the podcast. We’re gonna have a few different sections around news or if there’s anything happening in the industry, we’re gonna cover that. And then usually we’ll follow it up with a topic.

[00:04:29] Dara: But you might, you might expect to hear a couple of changes as we, as we fine tune that over the next couple of episodes. Okay. So news, what have you, what have you found, what have we found, Matthew, collectively, you and I, what have we found? 

[00:04:43] Matt: I’ve found a few bits. I. I tried to kind of relate each of them to marketing analytics in some way to kind of keep it on shoehorn, shoehorn it in.

[00:04:55] Matt: because there’s a real problem at the minute with news in any industry, it seems to be that the AI dial is turned up to 11. So there is a continuous stream of AI news and oftentimes quite big things. So it’s very difficult not to just talk about AI all the time. so I’ve tried to spread the gamut a little bit, but at the same time, for example, news about GA four is pretty light.

[00:05:24] Matt: There’s not much happening there in terms of new features and, and things like that. So, we’ll see. It’s, I, I’ve, I’ve tried my best. So it’s ai. I always thought it was steak sauce. Yeah. So the first one I pulled out is. Chat gt. So straightaway, AI chat, GPT shopping, I dunno if you saw that they’ve released a load of new features that hooks it up.

[00:05:51] Matt: I think I, I might be wrong, but I think they’ve, they’ve sort of collaborated a little bit with Shopify and you have much more detailed searches for shopping and it returns products back into the chatty BBT interface and you can kind of have a much more streamlined journey, go and purchase things and everything from there.

[00:06:10] Matt: And the reason I thought that was interesting other than it clearly is, was that it seems extremely relevant to e-commerce. It feels like a pretty significant shift potentially in the way people are going to begin shopping. and what the hell does that mean in terms of tracking all of that and understanding those optimizations as well?

[00:06:37] Matt: Optimization, yeah. So like. Is there a new term for search engine optimization? Has it, has it, is there a chat? LLM optimization? 

[00:06:46] Dara: I feel like I did see something, but it’s obviously not very memorable because I can’t remember. I’m sure there is some kind of SEO for, you know, like whatever the equivalent is, but I can’t remember.

[00:06:57] Matt: Oh, yeah, yeah. So I just thought that was interesting. It’s what, it’s another example of, I dunno, just people, you, you can imagine people interacting with the internet in a fundamentally different way and, and potentially not leaving these, these chat interfaces. We kind of talked about it a little bit last week with like conversational analytics, 

[00:07:18] Matt: but maybe data, a different, a different paradigm to traditional dashboard, general used dashboarding. But, I just thought that was a pretty significant change. 

[00:07:29] Dara: It is, it is significant and, and it is all these ramifications that we don’t, we don’t fully know about or understand yet. Like even just in general, as. People’s searching behavior or fact finding or whatever, whatever it is that people used to go to a website for, that’s gonna happen less and less.

[00:07:50] Dara: So, you know, one thing is opt opt, the optimizing part of it. Maybe I’m completely wrong here, but, but in my head, optimizing is one thing, and there might be ways to make your information more optimized, but there’s nothing you can do to track people searching for you. If that’s all happening within a, you know, if that’s not happening on your property, then there’s, there’s not gonna be any way to actually track, you know, unless the, unless the AI providers start to give out tools, you know, things like search trends, you know, there, there might be the equivalent of that for chat GVT or Gemini or whatever.

[00:08:27] Dara: But otherwise, you know, if people start to move more and more into using a. Interface of an AI platform for everything like shopping and all of their browsing, all of their research, then that’s scary for, you know, companies in terms of the reduction in the amount of data that might be available to know what people are looking for and what people are doing.

[00:08:49] Matt: I suppose it, I suppose it’s kind of the opposite to search bots in a way that you’d, you’ve wanted a no, a no index to say like, I don’t want to include all these search bots in my traffic and understand like Google’s core on my site and, and messing with my metrics. Whereas, if the example that comes to my mind is recently we, we, we’ve got quite a lot of content that we’ve created at Measure Lab around data form, and as such, Google’s documentation data form is pretty thin on the ground.

[00:09:17] Matt: And we, you know, we ran quite highly for that kind of content. Recently I was messing around with chat GBT deep research and asking it some data form questions and loads of them. Loads of the sources that it pulled out were from Measure Lab. And, and in a way I’m wondering, is it going and pulling that information on the fly?

[00:09:36] Matt: Like it’s doing its own Google search, it’s going to the website, it’s pulling the data out, in which case that you want that bot in your traffic to understand how that all feeds through into, into the end user experience. Yeah, you do. Yeah. You want to, you wanna 

[00:09:51] Dara: I know that it’s happened. But I guess you, you, you, I mean you, you take that ’cause it’s better than nothing, but you’re not gonna know much about the actual user behind that.

[00:09:59] Dara: But then I guess maybe that’s not that different to using a search engine. No, well, smarter Brains than us or figure it out. I’m sure it’ll be fine. Yeah. Yeah. It will. Did you say, just on the details of it, that you can’t actually complete a Truly, you can’t complete approaches yet? 

[00:10:18] Matt: Or can you, I dunno if you can pull, complete it within the interface, but I know it’s surfacing and I might be wrong on that if I should have read further into this before I started spouting about it.

[00:10:28] Matt: But I, but I know that it, that, I don’t know if all of this is generally available yet. I think some of it was like, you tend to do it in pro users and then release it down to general users, slowly. but I know it was, it was much more so that you could, so like a tabular interface of products being surfaced within the chat, the chat box and, and being able to click through and, and look at those products.

[00:10:52] Matt: I’m not completely sure if you can then complete the purchase within there, I would assume it would send you off to, to their payment gateway or, or to the product page. 

[00:11:01] Dara: But, yeah, ’cause you would have to go through, but eventually you, you know, it’s unlikely that they’re gonna have a payment gateway set up to work, which at GBT, but that’ll happen. It’s only a matter of time. 

[00:11:11] Matt: It may, maybe it already has, and we’re just talking like it hasn’t, I don’t know. 

[00:11:14] Dara: Yeah. Maybe we should probably put a disclaimer on our news. Say some of this news may not be true, but 

[00:11:20] Matt: Yeah. This is, yeah, we, we, we, we skimmed it. Yeah. It’s kind, it’s kind of, we’ll talk about it like we know what we’re talk 

[00:11:26] Dara: Yeah. It’s kind of a pinch of pinch of salt news. It’s not, you know, it’s not the news, it’s just stuff that could be news. 

[00:11:33] Matt: Yeah. I think one thing I would like is to, is to really make this more of a shock jock type podcast with sound bites and, and jingles. So maybe Pinch of salt news and some sort of jingle. Will knock up.

[00:11:45] Matt: Yeah. will be the producer of the podcast. Yeah. Anyway, let’s move on because we clearly dunno what we’re talking about in that bit. There’s a couple of things that have come out, for the, for the people who listen to the last podcast where Darren and I talk through. All of the updates were announced at Next 25.

[00:12:02] Matt: A lot of it was coming soon with, with a.dot. but a few things have started to trickle through into BigQuery now. We have been playing with them at Measure Lab. In fact, we had a call this morning where we were sort of all sharing and, and walking through some of our new things. So Canvas Assistant, which I think is the one you were quite taken by, if I remember correctly, Dara, we’re playing with that this morning.

[00:12:23] Matt: So that’s, that’s within BigQuery inside of Canvas, you can kind of just use natural language and it’ll start creating different nodes of, analysis and visualizations and, and things like that. which I think is really, really interesting to anyone who’s got data in BigQuery ’cause it’s making that data just a little bit more accessible.

[00:12:45] Matt: Again, with that comes risks, obviously, which I think is kind of our topic for later a little bit. What, what you do to prepare for those risks. And, but yeah, it looked really, really interesting and, like it’s really. making all that data even more accessible within BigQuery. 

[00:13:01] Dara: Yeah, I’m glad that one’s said that was the one I was, well, one of the ones that I was keen on. And, and I guess the little test I wanna run is like, how effectively can I use that? ’cause I think I’d be a good benchmark for, you know, somebody who has some, you know, some level of understanding. 

[00:13:16] Dara: I, I wouldn’t be going in there completely not knowing what I’m doing, but at the same time, I think I would be a good target user for that because I, I could lean on the, you know, the assistant bits of it to do some of the heavy lifting, but I’d probably know, you know, I’d know, I’d know what to ask it and, and, and, and hope, hopefully some idea I’ve had to sense check what comes back. so I’m kind of keen to play around with that and see what I think of it. 

[00:13:41] Matt: What, what I would recommend to anyone who does wanna play with it is definitely do have a play with it. But what, what you could potentially do is create a data set and put, just put some smaller, smaller couple of days worth of data in there.

[00:13:57] Matt: that isn’t as large and just interrogate that and play with that and get used to the features of it. Obviously don’t, don’t point it at anything raw because if you’re letting it just go off and start spinning off and doing things and it’s just pointing at a, I dunno, 3-year-old data set with billions of rows in it, you, you could end up with a bit of a hefty bill.

[00:14:18] Matt: So, that’s a good, good approach is to just ring fence some data for exploration. I think that’s what we’re gonna do internally. It’s gonna get a couple of days of data and just get everyone to be able to go and have a play and, and see what it can do. 

[00:14:31] Dara: Yeah, no good advice. 

[00:14:34] Matt: And really cool in terms of telling a data story. I think as well we’ve played in the past with notebooks like Python notebooks, being able to tell story analysis stories for clients. ’cause you can have that plain text element and visualizations and kind of disperse that amongst the bits of Python code that are doing the transformations. I think this is an even more accessible way of doing that.

[00:14:56] Matt: It’s just this very visual interface of seeing a journey of transformation and visualization and explanation, and being able to just share that canvas with somebody so they can understand what you’re talking about and how you got to that conclusion is pretty cool. 

[00:15:10] Dara: So what, what’s the difference? Is there a difference, maybe there isn’t, between using, to using CoLab and using Canvas in, in BigQuery, are they, they’re similar at least, right?

[00:15:23] Dara: Are they, are they the same or? 

[00:15:26] Matt: I, I, canvas is built on top of Python notebooks. So, obviously CoLab CoLab is a Python notebook. I think it sits primarily on top of BigQuery notebooks, which is the hosted version of Python notebooks that exists within BigQuery. So you can actually build out your data canvas and then at the top you can convert it to a notebook and it will take that and drop it into a notebook.

[00:15:48] Matt: You carry on from there. So even if you’re a bit more of an experience. Analyst, with Python skills. You can maybe start there just to sort of work your way through and pull out a few threads and then once you’ve found some of interest, convert it over to a Python notebook and pick up where you left off and, and carry on, do some more deeper analysis.

[00:16:07] Matt: Okay. Nice. So discovery and catalog, cloud storage data. I don’t think I need to say anymore about that. 

[00:16:14] Dara: That’s obvious. Does what it says on the tin. 

[00:16:18] Matt: What, what this is so that there’s, there’s a few. One, one thing that’s happening with BigQuery, which we kind of alluded to last week, is it just seems to be gobbling up various other data services around Google Cloud.

[00:16:30] Matt: In into itself. It’s kind of eating everything and bringing it under the big query banner. and There’s various services around governance and data cataloging and things like that that existed in, say, data plex or data catalog and these other services that are now starting to appear. One of them is this.

[00:16:50] Matt: Discovery catalog for cloud storage, which is essentially all that means is it is looking in like cloud storage buckets and being able to pull out metadata and understand the schemas and things of that structured and unstructured data that’s sitting in your cloud storage bucket and allow you to interrogate it in BigQuery pretty much straight away.

[00:17:09] Matt: So for, for say a, for say a, somebody who may not have loads of data engineering, resource in the company and things like that, it can really increase their ability just to interrogate that raw data and just start to start to play with it and speed up the speed up the time to be able to explore the data.

[00:17:31] Matt: and it could reduce that technical friction down quite a lot. Not played with a huge amount, but it looks interesting. Maybe a precursor to that multimodal table thing we talked about last week as well. Yeah. Having all of that in one place. 

[00:17:46] Dara: I’m gonna ask you questions about this, even though you just said you have another chance to play around. I’m gonna, oh know. Yeah. I’m gonna assume you know this inside out. I’m, I’m trying to get my head around like, I, I get it, like I kind of get it on one level. 

[00:18:00] Dara: So if you’ve got a bunch of data sitting in cloud storage, rather than in nice niche, big CRE tables, you can use this to go off and firstly, it’ll just go off and scan what’s there and it’ll tell you what’s structured and what’s unstructured and all the rest of it.

[00:18:15] Dara: But what can, you, can, can you actually, can you actually do something with that then? Or is it like, I’m trying to think how to phrase my question like, is it kind of like an audit tool or is it actually something that you can use? Can you actually run queries through it? 

[00:18:30] Matt: Yeah. Yeah, I believe you can. It is creating big tables, big lake tables, and yeah, you can query them once it’s kind of cataloged the data that’s sitting within cloud storage, I believe theory is you can query it from within BigQuery.

[00:18:46] Matt: So that’s what I’m, that’s kind of what I’m getting at when I say that, it removes that friction of being able to just interrogate that data. So you can imagine you’ve got a lot of, I don’t know, CSV files or parquet files or whatever else sitting in cloud store, and you just want to interact with it from BigQuery and not have to worry about how do I get that, ingest that data with data transfer service and match up the schema and this, that, and the other. You can just start poking at it. 

[00:19:11] Dara: There’ll be a, maybe this is a stupid knowledge. No, there’s no stupid questions. But maybe, but maybe this is a, maybe I’m not understanding it, by asking this question, but what, why would you bother then putting the data into BigQuery? If you can just access, and I don’t mean for any, like obviously there would be reasons to have some data in there, but would this, would this actually create an argument for leaving some of your data just in cloud storage and not actually having it in BigQuery in the first place?

[00:19:49] Matt: I dunno. I, I guess I can think of a couple of examples of potentially where you, you’d want to get it into, I, I if it’s gonna be something that’s more piped in and regular and sort of updated on a daily basis, you may want to be bringing that in into BigQuery because you then want to take that data and model it and combine it with other data and, and get other outputs potentially.

[00:20:16] Matt: Sitting in it, hitting the cloud store in the first instance would allow you to start interrogating and joining and playing with that data before you begin building out pipelines and pulling the data in on a regular basis. As one sort of use case. There’s, there’s probably a ton more that I, that I’m, again, I mean this is pra salt news, so, the, the depth.

[00:20:41] Matt: I love that. But yeah, I, I think that’s one example where it might just be easier just to, just to begin to, to play with it without having to go into the worries of, of building out infrastructure or importing it into BigQuery before you’re ready to maybe the type of content. ‘

[00:20:56] Dara: cause again, thinking of the, you mentioned about like the multimodals point that we talked about last week, and like if you, if you start, I asked a question last week around like, it, you know, is there gonna be, more of an incentive for companies to start putting more of their data?

[00:21:12] Dara: So not just the data we typically think of in terms of what sits on a query, but even things like assets, such as images and video files and PDFs and whatever. and you mentioned CSBs as well, so like, I guess it could be like a bit of a. Dumping ground’s gonna make it sound really bad. But, you know, there could be a bit of a case of like, let’s get all of our company data in there.

[00:21:32] Dara: It’s not all gonna be structured, it’s not all gonna be kind of used in regular reporting, but we might just stick everything in there. And then the stuff that doesn’t need to be organized and put into big query tables, it would give you a way, it would give you an interface to go and actually create a, so let’s say you’ve got a whole bunch of, you know, let’s say you’re a company, you’ve got a whole lot of older data in CS V files that you now have, you know, an updated database or whatever, but you could just drop all those CSVs in and then run queries on them as opposed to trying to, you know, gradually migrate them.

[00:22:09] Matt: Right? Now this isn’t, this is, this is structured and semi-structured data, but what the, what they specifically call out is Parquet, A-R-O-R-C-J-S-O, and CSV. that’s why I say maybe it’s a precursor to those multimodal tables and being able to pull, pull in the unstructured stuff, which has, I don’t think that’s been released yet.

[00:22:29] Matt: I haven’t seen anything about it. But yeah, you’re right. I think Big Lake, then pulling that into BigQuery and, and making it more accessible to that historical data could be really useful. Cool. moving on. GA GA four, try to grab some GA four information, but primarily I, we did, we did talk a little bit about from time to time, you know, when, when we’re coming to the end of a month, maybe wrapping up changes and things that have happened within a particular platform.

[00:22:58] Matt: So thought maybe we could, I’d have a look at what new features have come online in GA four over April. So this is April. This isn’t in the last week. but the generated insights stuff came into GA four this month. There’s a lot of those similar features appearing throughout BigQuery and Looker Studio and conversational analytics, but it’s.

[00:23:21] Matt: On particular reports and data sets, having an LLM look through the data and try and pull out what it thinks are interesting trends and insights that it will surface to you, within the ui. 

[00:23:33] Dara: This one got me, I mean, this, this one isn’t new. Well, this bit’s new, which is on the detail pages, but the insights on the overview pages have been around for, for a while.

[00:23:44] Dara: and it’s a bit hit and miss. Sometimes it picks up stuff that maybe you don’t really need to care about. but it is a massive time saver. ’cause when I read the description on this one, it just got me, reminded me, this is a bit where I’m gonna be, you know, talk about the old days. but I used to do, you know, you used to, this is a big part of the work, if you, you know, a client would have a problem.

[00:24:06] Dara: They’d say, oh, our, you know, our conversion rates dropped year on year. Or, we’ve got, you know, people having an issue getting through the checkout or whatever. You know, one of these typical problems. And they would have. And the process was always the same. You’d go into GA, you’d narrow it down, you’d look at a bunch of dimensions and metrics and you’d try and through the process of elimination, you might use segments as well, and you’d narrow it down to try and figure out what exactly was causing the problem.

[00:24:33] Dara: Sounds great, but in reality, that took forever because sound playing, especially with a bigger account, you’d sampling, you’d load times, it was a pain, you know, to actually answer that question. It could take you, depending on the size of the website and whatever else, could take you hours to figure that out.

[00:24:48] Dara: And this just does that in the background and then pops it up with a little note saying, did you know that user is from Paris who are in this age demographic? You know, their conversion rates dropped on this date. and you just don’t have to go in and do that anymore. and even just picking up things that maybe would’ve gone unnoticed before as well.

[00:25:07] Dara: So I remember when this came out, it was kind of scoffed at a little bit like, oh, that, you know, the suggestions are a bit rubbish or whatever. I think as it improves, this is gonna save a huge amount of time where you would’ve had to manually go through that process in the past. 

[00:25:22] Matt: Which is what you hope AI is gonna do for us is, is like remove the tedious and leave behind the, the d to a, to a certain extent. Because this is like, to be specific, I’m just looking at the announcement this, this was from the 2nd of April. 

[00:25:39] Matt: They specifically that it’s, it, it specifically that its generating insights are now shown directly within detailed reports in Google Analytics. So I dunno if they were, they were, I know I have seen them elsewhere and I know almost at the very beginning of GA four there was like in the dropdown, you could ask sort of simple questions and it would try and pull the relevant data.

[00:26:00] Matt: But I think these are more, it takes an active role in trying to pull out insights and it’s useful. We were talking about the cam canvas earlier. and generally across BigQuery, they’ve peppered in these little LLM pieces. So within any table within BigQuery, there is a generated insights section and it will do exactly the same thing.

[00:26:20] Matt: It’ll look at that data set and it will try to suggest a list of questions you could ask of that data and, and then get it to run queries to return the data that answers those questions. And it’s just helping just to, to jump to the jump to something that you can then begin to interrogate and, and confirm and dig into reports, shop templates.

[00:26:43] Dara: I think these features are useful for certain users of ga. It’s like anything that kind of customizes the interface and saves you having to go and deep dive anything. But I think you said you don’t use it much at all. I, well, I don’t these days, but I used to, although it wasn’t called GA four, then it was.

[00:27:01] Dara: It was universal. but you know, you want to go in a deep dive. You don’t want one of these prepackaged kinds of snapshots or reports, but that’s basically what this is. It’s just for, I guess, more, users who just want a higher level view and they just wanna go in and get some of the basic metrics without having to go and dig around and, and find them.

[00:27:19] Dara: So, I get it. It’s useful for some people. It’s not something that I think probably many people, even people who listen to this podcast probably wouldn’t be particularly wowed by that feature. So, yeah. Onwards. 

[00:27:31] Matt: Yeah. And tell it, it maybe you shout at, shout at your wireless again. I’m gonna, I’m gonna stick with wireless.

[00:27:37] Matt: yeah. Let us know if we’re talking nonsense. And it’s a feature that you absolutely love. and the other thing I had was around clo, not cloud code. Sorry. Cloud code. so cloud code is a visual studio. Code extension, Visual Studio Code being like an IDE that you can use on your, on your locum machine.

[00:28:01] Matt: But what they’ve done to Cloud code, cloud code is a Google released extension that interacts with Google Cloud in loads of different ways. They’ve added tons of BigQuery features. So essentially you can develop and work with and interrogate and manipulate tables and data sets and query data and everything all from your local machine.

[00:28:20] Matt: Now, with not having to do it directly within the, the BigQuery ui, that, that might be beyond the pale for a few for, for some people it might maybe be, you know, the, they’re not necessarily used to that world of, of living in the, the, the IDE, but certainly, for example, with data form, I found that working with the, the.

[00:28:44] Matt: Data, form data locally in an IDE and using the extensions that are built around that add loads of quality of life, things that maybe Google doesn’t, doesn’t include like auto completion and, and split windows and all these different bits. I think both of those things I just mentioned, Google does happen in the ui, so they were terrible examples.

[00:29:04] Matt: But trust me, there’s, there’s, there’s interesting things you could do with it and it, and it allows you to quickly and easily even move things around back, stuff up in GitHub and all the good things that come with, with sort of working in a software engineering type way with this type of stuff. So I haven’t played again, there’s a theme.

[00:29:20] Matt: I haven’t played with it yet ’cause I, I only, I only got announced earlier today, but, I certainly will be, I think I’ll be, adopting that pretty heavily into my practice. 

[00:29:31] Dara: Yes. Maybe I shouldn’t be so impressed by this. Maybe this is what we should just expect, but it does seem like, not, not just Google, but all of the kind of big tech companies, they are, they’re not, they, you know, they’re, they’re, they seem to be fairly equally pushing out the developments on both the technical side.

[00:29:48] Dara: So with things like IDs, but then also the less technical side where they’re making it a lot more accessible. And it gives me a headache almost thinking about trying to tackle both of those two ends of the audience. You know, in parallel, they’re making it, they’re making it better for. Developers, but they’re also making it better and easier for people who don’t have the really hardcore technical skills.

[00:30:12] Dara: Yeah, yeah. True. That’s a good point. I had a sneaky GA four update. I say it’s not, it’s not, it’s not sneaky. It just didn’t, you, you obviously weren’t very interested in it, so you, it didn’t make it to my really good GA four section reports. What were they even called? Reports snapshots. So I’ll review snapshots.

[00:30:32] Dara: the aggregate identifiers came out as well, in April, which is the solution for us. It’s a solution to the problem that was happening with some misattribution of page Google traffic being misattributed to organic. And I think it’s, well, one of the reasons that was causing it was if the user has ads, I never remember the name of it, but the ad consent cookie, whatever that one’s called.

[00:30:58] Dara: If they haven’t given consent for that. Then it wasn’t passing through the information from the GCLID parameter and it was therefore misattributing that traffic to organic. So they’re using these aggregate identifiers now, which is a way for them to identify that it is Google page search traffic.

[00:31:19] Dara: But you obviously, it obviously still doesn’t track anything. It’s not supposed to track, but, you know, as opposed to being tracked in the wrong bucket, at least it’s being tracked just coming from the right traffic source now. So that’s quite a big feature if you are, if you’ve been having headaches from this issue, which a lot of people would’ve been.

[00:31:36] Matt: Yeah. So we thought we’d be good, would be to talk around how to make sure that your data is as robust as possible when it’s sitting in BigQuery. So we talked last week about all of these new features that are making data within BigQuery more and more accessible. So what practical things can you be doing across that data life cycle to, to make sure that it’s as robust as it can be, that it is.

[00:32:00] Matt: You understand when things are going wrong, that it’s secure, that there’s that, that all of these boxes are ticked. We’re not gonna go into aching detail, but just to, just to point out points where we think, you know, practical things need to be thought about and, and actions taken. 

[00:32:17] Dara: My takeaway from last week was that you don’t need to worry about any of that. Did I miss the meaning, did I misunderstand? 

[00:32:24] Matt: There is that, you can certainly take that approach that’s called the ostrich tactic and you just sort of bury your head in the sand and just plug away and, and whatever numbers it spits out at you, it’s fine. Just, yeah, just present those. 

[00:32:38] Dara: But, but no, no, but, but I, I, obviously, I’m, I’m joking, but it’s kind of, it’s, it’s, well it’s not even half true, but there’s something in the, in, in the joke in that Google and all the other kind of w well we talked largely on Google last week I see was ’cause of next 25, but.

[00:32:56] Dara: They have to do this, but they are making it look like everything just works perfectly. So there’s gonna be a lot of people out there who probably will just go away now and think, oh, this is great. Everything’s automated. Everything you say, I, we can just press buttons and things will happen and it’s all happy days.

[00:33:13] Dara: But you’re absolutely right. It’s not the case. And you did make that clear last week, just because there’s a lot of AI in BigQuery now doesn’t mean that you can just trust that everything that’s in there is gonna be correct. ’cause it’s the whole garbage in, garbage out thing, isn’t it? 

[00:33:30] Matt: Yeah, yeah, exactly. And that goes, that goes right away, right the way at the very beginning from where the data’s being generated, right the way through to whatever you, those tools that we’re describing that, that are essentially activation tools. so it seems sensible to start on that far left hand side in terms of, of generation.

[00:33:53] Matt: So to my mind. I really, well, let, let, let’s talk in the, the, the world and the cadence of, of marketing analytics, potentially, like e-commerce and, and sites like that have a really well thought out on well structured data layer. So you, so you understand all of the data that’s coming out of it, it’s documented and, and it, and you ensure that everything that is being tracked is correct and well designed and well, well thought out in the first instance, right to the very beginning. 

[00:34:20] Dara: So, I feel like I might, I might ask you this question too, at, at every, every section of this, but, so if you are, if you are a, if you’re an organization and you’ve got the right people to do all these different bits of the, of the process all well and good, and everyone talks to each other and collaborates, then great.

[00:34:38] Dara: That’s often not the case. So who, whose, whose role typically would it be to make sure that that’s done or So if you are a, let’s say you are, a. You’re new to the role in the company and you work in the marketing team, and you want to make sure that you’ve got good data to use to inform your campaigns.

[00:34:57] Dara: You are worried that maybe the data layer isn’t right or the right data’s not being collected. I mean, we’re, when we go, in fact, I won’t ask you this question now as we go through this. Let’s, let’s maybe call out who would typically be responsible for that and maybe you know who you might need to speak to as well.

[00:35:15] Dara: So it could be like, it might be a third party agency or it could be another team, or it could be another person with a particular skillset. Because this is one of the complications, isn’t it? It’s like this becomes more and more broad between, you know, from collecting the data right the way through modeling and, and activation, the skill sets increase probably exponentially over time as well.

[00:35:38] Dara: So it might be just worth us kind of trying to highlight that as we go through this process as well. As much as we can. 

[00:35:45] Matt: Yeah. I think with that point, the, the, the, the, the data generators, typically they’re gonna be your, your Debs, the people who are, who are putting the sites together and creating whatever logging or whatever events that are firing out of, of your site.

[00:36:00] Matt: So that’s the most important third party that you’re gonna be interacting with when it comes to data layer specs and, and getting your, your preferred tracking set up. I think, unfortunately, like you say, a lot of marketing teams currently don’t necessarily have a data engineer or an implementation expert sitting on staff that’s gonna be able to, to just know off the top of the head how to set something like this up and, and run with it.

[00:36:29] Matt: So in that case, you may reach out to a third party measure lab, or somebody similar, some other consultancy or agency that can help design those things. But there is also a. You know, there’s Google isn’t great at documentation, but I think, I think there’s, there’s tons of, of resources out there now between, you know, CMOs written on things on, on this and there’s loads of other, well, it’s, well TRO and ground data layers, and TA four data layers and e-commerce data layers and the things you need to include in them and the structure of things.

[00:37:05] Matt: So I think most people will be able to come to a, come to something and, and figure it out with some, with enough reading, and time put into it or, or reach out to an expert. 

[00:37:14] Dara: I guess one thing that often doesn’t happen though is that, ’cause we, we find this, don’t we, where sometimes these things are considered after the fact.

[00:37:21] Dara: And I think one universal piece of advice would be that the people who are ultimately gonna use the data and act on it need to have a conversation with whoever it is that’s responsible for. Collecting that data, producing that data in the first place. Otherwise, you could have somebody who’s very, very capable and knows how to track everything under the sun.

[00:37:41] Dara: But if they’re not gonna ultimately use the data, then you know, there needs to be a, a marrying up of what is actually collected, what it means, and then how that data’s gonna be used by the people who are gonna run campaigns or make changes in the website or whatever. 

[00:37:55] Matt: Absolutely. And I think the next point is that it’s, we’re talking about a typical, say, a typical Google stack here.

[00:38:01] Matt: So I think a lot of, a lot of the tracking, et cetera, and the tools that will be in the hands of, of the marketing team would exist within Google Tag Manager. So setting up your various pixels and Google Analytics tag tracking and crucially what I’m about to come onto in terms of consent, and understanding and ensuring that all of your consent and consent mode and, and all of these things are properly configured because.

[00:38:27] Matt: If it isn’t, and I’m not gonna go into detail of how to do that, ’cause that that could be the subject of a podcast in of itself. But if your consent isn’t configured correctly, that can be leading to all sorts of, not just, not just policy and, and GDPR and, and breach issues, but really fundamental tracking problems that may not be immediately obvious to you within when it’s, when the data’s reaching GA four and ultimately BigQuery.

[00:38:54] Matt: So ensure that that is all set up, it’s robust, it’s, it’s following the right, the right updates and consent defaults and stages of which it’s firing. It’s really, it’s really crucial as well. 

[00:39:09] Dara: Yeah. And again, on that point around collaboration, I guess you, we advise that people would check that, ’cause there’s obviously a legal component to that.

[00:39:18] Dara: So what’s actually being gained in terms of consent and how the data’s being used would involve making sure that that’s okay from her. Legal standpoint and any disclaimers that are made or cookie policies, that kind of thing. But then also the technical understanding of what that consent will do. So the status of that, of that consent will do in terms of the data that ultimately is, is tracked and used.

[00:39:38] Dara: So again, there’s probably multiple teams or multiple people that need to be involved in that step of the process as well. 

[00:39:47] Matt: And the other important thing to consider here is making sure that you have some sort of monitoring or some sort of way of understanding when something’s gone wrong. So you, you’ve got, you’ve got tools like, for example, tracking plan, which can be really useful way of, of something alerting you when something’s changed or something’s broken or some metric doesn’t look right and it, and it can, it can give you a poke in the ribs to say, you need to come and investigate what’s, what’s happened here.

[00:40:14] Matt: Equally, a relatively new GA four feature. See, I do know some features that appear in GA four if the. Annotations. I say it’s a new feature. It’s a feature that existed before in every other analytics platform ever, but GA four added back in. so be able to understand key changes and things that have happened within a development lifecycle.

[00:40:36] Matt: Maybe you’ve added a new pixel, maybe you’ve updated a tag, maybe you’ve implemented a new piece of consent tracking. Being able to see within your G four data when something happens and being able to correlate those two things, it’s gonna be really important. Ultimately it’s about maintaining trust that in, in the data, once it reaches that source, you, you’re really happy and everyone’s happy that the data is correct and being able to point to reasons for, for changes is, is crucial as well. 

[00:41:04] Dara: Yeah, I think the monitoring is such a key point. ’cause even if you’ve done a brilliant job of implementing everything in the first place, you just, you just don’t know when something’s gonna go wrong on the website. It’s outta your control or there’s gonna be a change to.

[00:41:17] Dara: I dunno, some, some, some browsers change even that, that, that, that affects traffic on the site. So anything like that is going to mess up your, you know, the data that you eventually end up reporting on or, or, or using to activate campaigns or whatever. the documentation as well. I mean, I, I know people were crying out for annotations to go back into GA four.

[00:41:39] Dara: I wonder if you got any, like what would you advise in terms of keeping that documentation across everything, so, you know, ’cause obviously GA four is not gonna be your only data source typically. So would you, would you suggest that people maybe use something like, I dunno, like a system like Confluence or something like that to keep track of the broader kind of history of, of what’s, what’s gone wrong and what key dates and things like that?

[00:42:07] Matt: Yeah, I think, I think it all depends on what you have as an organization and how, how you like to, to document things. Part of the problem is when you have these big all encompassing documents, they tend to just sort of die on the vine and, and lose their use. So, where you can pepper in automation, like for example, there’s some really cool GitHub repos that people have created that will automatically add an annotation in GA four when you publish a new version of GTM.

[00:42:36] Matt: So that is, that’s completely hands off and it will just add a little annotation, like something changed here and you don’t have to do anything. The more you can automate documentation and, and make it not have to be thought about the better. But yeah, like you say, I think, I think key, key additions of other data sources, understanding where these are placed and what data they’re pulling in and all that stuff and, and having it available to people.

[00:42:59] Matt: metadata is, is, is, is really important. I think on the GA four side as well, understanding. The settings and the, and the, the, the attribution settings and the modeling and, and all these sorts of things that are going on within, within that platform. It can be really easy to toggle on modeling data and not really understand that that’s what you’re doing and you’ve got consent mode set up and the data you’re seeing within GA four may not actually be real numbers, which is maybe the future.

[00:43:31] Matt: Maybe that’s just how, analytics works now, but it, it can be very confusing, especially if you’ve then got a raw output into BigQuery and it doesn’t match up with GA four and GA four doesn’t match up with another analytics platform you’ve got and, and you’re doing testing. You’ve gotta figure out why.

[00:43:46] Matt: So understanding those features and, and the admin of all of these things with enough, with enough depth to, to, to know what you’re turning on and off is really useful. 

[00:43:57] Dara: And that’s an area where governance, I guess, is really important because you’ve, it’s, it’s. One thing if people are querying data without really knowing exactly what they’re looking at, but it’s another, if you have, too many people who shouldn’t have access to admin, you know, admin features on these, on these tools, because you can go in and, you know, make changes that you don’t fully understand the implications of.

[00:44:22] Dara: So having a tight control over who’s actually able to make those kinds of feature changes or setting changes is really key as well, isn’t it? 

[00:44:29] Matt: Yeah, and I, I’m loathed to, to a certain extent, to, to recommend stringent access policies on things like GTM because on one hand it kind of gave a bit more power to the, to the marketing analytics and the marketing teams who historically were just locked out of, of making their own changes.

[00:44:48] Matt: But I think it is important to, to. Ringfence that, and, and think about who has access and who’s doing what and working in workspaces and keeping things protected and, and working in development and, and staging, development and production environments to, to keep things ringfenced in that way, just to protect your data, as much as you can.

[00:45:09] Matt: To that end, get the data out. And again, I’m just gonna stick with GA four for now, but say, getting that data out into BigQuery, all of all of these features. So, so right now we’ve got, you know, we know our consents, right? We know our, we know our data layer, like the back of our hand. We’ve got everything well documented.

[00:45:26] Matt: We understand all the settings and, and things that are turned on and off. In GA four, we are extracting our data out into BigQuery, building out from their reporting tables and trying to pull that data apart to answer the specific questions we want to answer, which I think is really important as well.

[00:45:43] Matt: Combining that with other data sources that may also be within BigQuery and not just pointing out and talking to raw tables is gonna. Save you a lot of money for one thing, but it’s also going to make it much more useful for all of these new features that Google are stuffing into BigQuery.

[00:46:02] Matt: And unfortunately, well, I say unfortunately, that includes layering out your metadata and adding in column descriptions and, and making the data as rich as possible for people to understand what it, what’s in it, and what, what it means to you as a company. Again, governance. But Google is automating a lot of that with upcoming creatures, like automatic metadata generation and, and, and alike.

[00:46:31] Matt: That’s gonna make it, make the, the LLMs and, and yourself and your staff members and anyone else in your organization, the organization, understand what’s actually in there. help with a lot more confidence. What you’ll find often is. People when it gets to a certain critical mass, people just keep pulling in the same data over and over again because they don’t know what’s already in there.

[00:46:52] Matt: And that ultimately leads to what I like to term the derelicts data warehouse. but, but just, just something that’s become a bloated mess. No one knows where anything is, so they just keep pulling more and more which ultimately is costing you a lot of money. And for all of these features, that Google is adding in probably produces five different answers to the same question ’cause it ’cause people are talking to different sources. 

[00:47:18] Dara: yeah, I was gonna ask you about that or, or, or kind of drill into that a bit more. ’cause you, you’ve, you’ve, you’ve probably got like a few reasons for that to happen. You’ve got the discrepancies you might have between the source data in the source platform versus what you eventually end up using.

[00:47:34] Dara: And you are a nice clean BigQuery table though. And, and even then it might be different again when you get it back into. Some kind of bi tool visualization tool where you’ve maybe added filters and whatever. Nevermind the fact that, as you said, you can also then even in BigQuery, have multiple copies of the same table.

[00:47:52] Dara: So what’s the, this is, there’s probably no simple answer to this, but what would you suggest is the right approach there in terms of making sure that people aren’t all looking at different versions of the sensor? Slightly different versions of the same data. 

[00:48:08] Matt: Yeah. I think governance, again, making sure that’s, that what’s in there is well documented, labeled metadata is created.

[00:48:14] Matt: One aspect that I guess we can jump to now, which, which was gonna be my next point is, is around access levels and, and security and make, just sort of having an idea of who needs access to what, where does that live, putting that into different data sets with different accesses so people can go in there and work in there and, and, query that data.

[00:48:35] Matt: It’s tricky to have a really. Overly strong arm of centralized governance now, I think because it, it, it’s slow and, and can be a bit of a bottleneck, particularly at the speed of everything’s moving. So I think it’s going to be perfectly reasonable in my mind for organizations to have more federated teams that are working on their own sources of data, potentially that spring off of a centralized, data source.

[00:49:05] Matt: But they need to make sure they maintain the correct governance and make sure that everyone understands what data’s already in there and that people aren’t able to just keep adding in new data willy-nilly and getting different, different access to different questions. So yeah, documentation, access, limitations, clearly labeled.

[00:49:24] Dara: Yeah, I think that’s all you can do, isn’t it? ’cause it’s hard. ’cause you, on the one hand, want to make the data accessible and the different, different tools accessible, but you can’t just have a free for all either. So you do. I guess, yeah, governance is a whole, it’s a whole topic on its own.

[00:49:40] Dara: So, I think what you said there, it sums it up as a kind of general view. 

[00:49:45] Matt: And I think also, and this goes right, right way from from the left again, but, and again, I guess it comes down to governance, but having joined up understanding of things like, naming conventions and, and ways of labeling that are, that are certainly team-wide, but hopefully company-wide so that people can understand what’s in that data.

[00:50:13] Matt: ’cause even it, it could be easy to label some, some column this one and then have a really numb bit of a crap description in there. And that may be exactly with what people are after, that may be the piece of data that they really need. But they’re just gonna go and try and answer the question again and go for a while.

[00:50:32] Matt: Process of pulling the data in and exploring Mandana analyzing. And the same goes in GTM. You see some really big like tag manager containers that are full of 20 different naming conventions that it’s almost impossible to find if somebody is already sort of set up a piece of tracking you can just glob onto and, and adapt to your own lens.

[00:50:53] Matt: So understanding your language as a company in those terms is gonna be really useful. Not to mention that it also makes better naming conventions, it’s gonna make it more useful for LMS as well. and then those features, 

[00:51:06] Dara: And like, I guess with any, any individual part of this process or journey or whatever we’re calling it, you would want to do periodic audits of each one anyway.

[00:51:16] Dara: I wouldn’t use so with things like naming convention, because. Best of intentions. These things tend to slip over time because different people work on different things and, and, and, and, you know, things shift away from, they might start out nice and consistent and then they’ll gradually just drift off.

[00:51:33] Dara: So, you know, whether it’s done every six months or once a year or whatever, looking at each aspect of this and actually checking that everything is still consistent. Getting rid of any duplication, getting rid of any old tables or tracking or whatever it is. Old reports could be anything. If things aren’t being used, then they’re just taking up space on basic.

[00:51:52] Dara: Just going through and doing a good spring clean every now and again, and getting rid of some of the old stuff that’s just making things confusing. 

[00:52:01] Matt: but yeah, you’re right. I think that there’s a lot of features, especially when you’ve got the data into Google Cloud. There’s tons of features around monitoring that you could be setting up to.

[00:52:11] Matt: People on everything from assure people in. People aren’t querying too big data sets. But you could also get as granular probably as being alerted when people are setting up new tables, being alerted when, potentially something doesn’t fit a particular naming convention you want to adhere to. And you can build out dashboards within Google Cloud monitoring that look at when tables were updated, how long ago they were updated as this, as that has been sitting there not doing anything for however long, that even has graphs, like the number of tables that are being created on a daily basis.

[00:52:48] Matt: You can see how that’s going. So being aware of that, setting up alerts through email if you, if you’re really desperate to be interrupted on a Sunday afternoon, you can get it sent to your WhatsApp, I believe. so, so, deploying techniques like that are really useful as well. 

[00:53:04] Dara: The aid, you know, is gonna become more and more useful for this as well. Like, you know, where they’re gonna tell you they found these inconsistencies in, whether it’s naming and conventions or we, you know, we suspect this table, table is duplicate or whatever. it’s gonna get easier and easier as long as people don’t do that thing. 

[00:53:25] Dara: Where if it’s, if it’s too trigger happy and then you just ignore everything it tells you, it needs to be pitched just at the right level to say, you know, there’s, there’s something that’s not consistent here, or there’s something that looks like it’s redundant that you could consider getting rid of, that you can become blind to that if it’s, if it’s triggering, especially if you’re getting on a WhatsApp on a, I think I’d be throwing my phone out the window if I started getting WhatsApp messages about big query tables.

[00:53:51] Matt: Yeah. And I’m, I’m guilty of that. I set one up for our, I set one up for our website a while ago. I. I was trying to gauge what size of service side GTM container we need. So I set an alert off telling me if there hadn’t been traffic for a certain amount of time, and I get it looked like it’s just on my emails, not could be to a text, but it’s like three times a day. Like you haven’t had traffic for 20 minutes, don’t care. and I need to turn that off. 

[00:54:19] Matt: But yeah, so, so, all those things, the, the, the, the thing to think about in terms of, of security as well, from a pay people perspective is the, the principles of least privilege. And that goes for, for people, that goes for service accounts, that goes for anything that’s accessing the data.

[00:54:35] Matt: So that is to give anyone only the access they need in order to perform whatever role, whatever piece of analysis, whatever, whatever the application may need to, to perform its function. Because anything wider than that is, is a security concern and could, could cause problems. 

[00:54:53] Dara: Yeah, and it’s always better to give less access and have the people who really need more come and give you a case for why, as opposed to, you know, just blanketly giving out too much access to, to everyone. Like it’s, it’s obvious, isn’t it? I say it’s obvious. He sees it in practice. 

[00:55:10] Matt: It’s tempting. Yeah, it’s very tempting because sometimes, I often describe it as permissions, whack-a-mole. It can feel like you, particularly when you’re working with new, with new cloud projects, often you get these accesses and then it’s like, oh no, I, I also need this very obscure access. 

[00:55:30] Matt: And like, knock that one down and then it, then it highlights another access you need, you knock that one down and it can be quite a bit of back and forth and the temptation is there. 

[00:55:40] Matt: Yeah. Yeah. Yeah. And, maybe that can be okay if for, for somebody working within a company just to get things set up in the first instance, as long as they have one eye on rolling that back to what is needed. 

[00:55:54] Dara: And that’s often, that’s often the challenge, isn’t it? It’s like, oh, if I just give you edit access now and I’ll take it away once you’re done.

[00:56:00] Dara: But you’ve got to, someone’s gotta actually remember to do that. So again, I guess this is where our thoughts are. 

[00:56:05] Matt: Yeah, yeah, yeah. And the same goes for GTM auditing, accesses auditing the, the, the tags. Same goes for GA four auditing access. I mean, I would get, I would challenge anybody listening to this now to go and look at, say, their GA four user accounts and, and tell me that everyone still works at their organization or they know that, that there, there’s probably about 10 agencies.

[00:56:29] Dara: Yeah. Yeah. At least. Yeah. And, and say, oh, you see all sorts, you see, you see personal email addresses of people that left the company two years ago, and you just think this is, you know, get, get that out of there now. It’s crazy. So yeah, periodically check all these things, make sure they’re still the way you expect them to be.

[00:56:47] Matt: Yeah. And I guess the final one, and this is, this is probably more slightly on the advanced end, but making things repeatable where you can, if you’ve got a, if you’ve got a medium sized team, then it’s so easy for everyone to have four different ways of doing something. And to, and for differences to begin to creep in.

[00:57:08] Matt: If, if you can automate things, touch, so maybe you’ve got a particular way you like to set up Tag manager containers, take that out and, and be able to import that again as a, as a template. And starting point. Similarly with Google Cloud, if you can leverage things like Terraform, which is like infrastructure as code, just to have a best practice.

[00:57:32] Matt: It already has your naming conventions baked in your alerts, common alerting setup, and all of these kind of things that can just be deployed and set up and right from the off and, and you can include things like granular permissions and stuff like that within that, that kind of, that kind of tooling.

[00:57:47] Matt: It’s just that it’s not necessarily, it’s, it’s, it’s pretty code heavy. It’s not too taxing, but it is, it is. It is code, Python code. 

[00:57:58] Dara: Does Terraform only work with BigQuery? Does or does it work with other tools as well? 

[00:58:03] Matt: Yeah, it works. It works for pretty much any cloud provider. Really, it’s infrastructure as coach’s open source, but it’s supported by Google, Amazon, all, all the other cloud providers, you can use it to just deploy infrastructure.

[00:58:21] Matt: Okay, cool. Cloud infrastructure. Yeah. But yeah, I think that’s, I mean, it was not exhaustive. It was just more of a pointing out a few places along the journey where, where there may be things to consider and best practice to consider to make the most outta of these tools. And like I said, we could’ve gone probably into podcast depth and any one of them, really. 

[00:58:40] Dara: And, and, and maybe, maybe we will. I think that was the aim for this one, wasn’t it? What it is to give a bit of, a bit of a kind of broad brush strokes, bit of a whistle stop kind of tour and say, look, here are some of the things to look out for, but it, you know, it, this, this could become a theme where we do actually dive a little bit deeper into.

[00:58:57] Dara: Because I think we both felt when we were watching all the next 25 talks that like Google do, do a very good job of saying, you know, all of this is just taken care of. Everything works straight outta the box, which we know it doesn’t. So I think data quality, making sure you know what’s actually going into this in the first place, is such an important theme.

[00:59:17] Dara: So we probably will end up. Diving a bit deeper, I’d say on, in future, you know, and maybe with some of the guests as well. It might be something that we, you know, basically expect to hear us talk a lot about, about data quality ’cause it’s so fundamental to all of this. 

[00:59:33] Matt: Yeah. And, next time we talk about GA4 in any length, maybe we’ll bring someone on, from Measurelab who actually knows what they’re talking about.

[00:59:42] Dara: This is probably a good point actually to mention something that we’re gonna try out. So if, if you’ve got a specific question, a problem with some data that you’re working with or something you just can’t get your head around, the name is to be confirmed. We’re, we’re, we’re kind of tying the idea of calling this like data agony t or something like that. But, we might change our mind on that. 

[01:00:04] Matt: Yeah, because I, I wonder if as well, this is clearly an invention, a fly between us here, but, I was interested like just, just general audience participation. If anybody’s got some hot takes or some thoughts on something as well as problems , if they just wanna send it our way and we may or may not talk about it, that would be cool.

[01:00:26] Matt: So I dunno if Adi amp necessarily works. 

[01:00:28] Dara: Yeah, you’re right, actually. Let’s not restrict it. I think, if we, if we, if we mention something that you disagree with, we’d love to hear that. If you’ve got something you’d like us to talk about, then brilliant. But I think we’re open to questions, feedback, whatever. and then as you said, and I like this, we either will talk about it all we want, depending on how we feel. 

[01:00:52] Matt: Yeah this may be the last time you ever hear about this idea or segment of the podcast, and it just disappears forever. Top topic. 

[01:01:01] Dara: I, I think if anything it’s gonna spark feedback. It’s gonna be our pinch of salt news where we just basically make things up every week 

[01:01:07] Matt: that’s stuck as well now, pinch of salt news is 

[01:01:10] Dara: Yeah, I think so. Yeah. That’s a keeper. That’s a keeper. 

[01:01:13] Matt: And it, and to be fair, if anybody’s reached this point with the podcast, they’re probably the people, most likely people to send us two interactions, send us some questions. 

[01:01:21] Dara: Okay. But I think for now, for today, that’s probably a good point to draw, to draw to a conclusion.

[01:01:28] Matt: Yeah, I think so. We’ve got touch wood, quite a few cool guests lined up for some future episodes, so look forward to that. 

[01:01:36] Dara: That’s it for this week’s episode of the Measure Pods. We hope you enjoyed it and picked up something useful along the way. If you haven’t already, make sure to subscribe on whatever platform you’re listening on so you don’t miss future episodes.

[01:01:48] Matt: And if you’re enjoying the show, we’d really appreciate it if you left us a quick review. It really helps more people discover the pod and keeps us motivated to bring back more. So thanks for listening and we’ll catch you next time.

• Spotify
• YouTube
• RSS Feed