#146 Do Spreadsheets Need a Rethink? With Hjalmar Gislason, CEO of GRID

Hi, I'm Maggie, VP of Curriculum at datacamp, and I know firsthand how overwhelming it can be to keep up with the rapid changes in data and AI.

But staying ahead doesn't have to be hard.

That's why our curriculum team has curated the best learning content to help you and your team stay ahead, covering everything from data and AI literacy for busy executives to cutting edge generative AI for developers, and everything in between. Whether you're upskilling yourself or your team, we make learning practical, engaging and flexible. It just works. Power your data and AI transformation journey on datacamp.com the software industry has been very busy building fantastic tools for data scientists and other data experts. But the everyday knowledge worker that still has to do quite a lot of work with data and numbers and visualization calculations and so on has been left with the Humble spreadsheet for about 40 years. And what spreadsheet users often don't realize is that they are writing software. They are. Spreadsheets are code, they're just encoding relationship between data that lives in cells instead of writing kind of lines of code that get executed one after another.

Welcome to Data Framed, this is Ritchie. We talk a lot about Python, SQL and BI tools on this show, but there's one other data analysis tool that's been wildly popular for four decades now, the Humble Spreadsheet. In truth, there's not been a great deal of innovation in this platform in the last decade or so from the market leaders, but that's left things wide open for a new generation of spreadsheet companies to push things forward. Today I'm speaking to Hjalmar Gislason, the CEO of Grid is His company's been building a modern spreadsheet platform, so I want to know what's left to innovate in spreadsheets, how AI will change the platform, and how spreadsheets fit in amongst other modern data analysis tools. Let's hear what he has to say. Hi Hjalmar, really glad to have you on the show. I'm going to dive straight into the tricky question, which is you're trying to build a company around spreadsheets, but for such a long time when people think of spreadsheets, it's either Excel or we, more recently Google Sheets. So what made you try and decide to take on these big competitors?

The long and short of it is that in my previous job as VP of Product Management at Qlik, one of the kind of VA companies, I realized that while we were making this accessible software for business people to do analysis and look at their Operational data. There were still these were power user tools. There were only a handful of people in each organization that really knew how to use them. And then they had a much larger viewership obviously. And everybody would go to these qlik people or tableau people within their organization to have anything done with new dashboards being built out or analysis or something like that. But then at the same time, it's not like the people that they were servicing and that were mainly viewing the BI tools weren't working with data and numbers, but they were just doing it in spreadsheets. And that kind of led me on a six year journey or so to dive into that market and try to really understand it. But I think to sum it up, I came to the conclusion that the software industry has been very busy building fantastic tools for data scientists and other data experts. But the everyday knowledge worker that still has to do quite a lot of work with data and numbers and visualization calculations and so on has been left with the humble spreadsheet for about 40 years. And while it has evolved, there hasn't been like any big step change in the tooling we give to everyday knowledge workers. So that's kind of how we came to this market. I should then also be careful that we're not taking Excel and Google Sheets heads on. We think of ourselves as the numbers tool for a new generation. And what we mean by that is broad strokes. What the market looks like is enterprise companies, larger companies, older companies, somewhat generally speaking, older people use Excel. Then you have companies started in the last maybe 15 years. They will have started on Gmail, they will be using the Google Suite now, Google Workspace and over the last maybe five years Google Sheets has in those organizations almost entirely replaced Excel. But then you have kind of the newest generation of companies maybe started in the last five years. They will probably also have started on Gmail and they're buying the Google Suite that will be their anchor. But they are also buying best of breed tools like Notion, like Canva, like Airtable and the like. And those are our people. So we have the best numbers tool for that target audience. Excel people and Google Sheets people can use Grid as well for their benefit. But our sweet spot is that kind of upcoming generation that is using cloud based productivity tools that aren't necessarily a part of either the Microsoft or the Google stack.

That's really interesting that you said you've got these BI tools of power, BI and Qlik and Tableau and all this sort of thing and that actually even though they're purporting to make data easy to use for everyone. They're still too technical for a lot of people. And so actually this is the reason spreadsheets still exist. So you also said spreadsheets been around since the 1980s. So what are the sort of innovations left? Is everything not invented already?

Not quite, but it's definitely been an innovation that has stuck around. So I was lucky enough when I was living in Boston to get to know Dan Brecklin, the creator of VisiCalc, the very first spreadsheet. So that was released back in 79 and came to the market and it was really, it was the first business software ever made. Like it was the first reason anybody would bring a PC into an office. Before that, there was no reason to. There was nothing you could do at work with a PC, a personal computer. So VisiCalc was made for Apple II. So Apple II were the first computers that made their way onto desks in people's offices. And the reason for that was VisiCalc. So it was very transformational. And immediately what people started doing wasn't only working with what I think we typically think of when people are, when we think of spreadsheets are calculations and financial models and the like. But we also probably all know that people use spreadsheets as small databases. They stand up almost like small business applications and so on. And this happened right away, like people were making CRMM systems in VisiCalc before CRM was probably decades before CRM was even a term. And then now that's broken off and I think it's $150 billion industry, the kind of the customer relationship management. And then you've more or less seen this with kind of entire categories of business software that they start off as something. And even when people are starting their businesses, they may even today, I think many companies first CRM system is a spreadsheet and then they figure out maybe when they have 100 customers that there might be a better solution to that. But the reason they've stuck around is that they are so flexible. That's one of the biggest strength and probably also one of their biggest weakness. But it's a very kind of essentially just this concept of being able to have a two dimensional text editor. You know, if it was nothing else, that is very compelling because there aren't any. Like you only have certain level of guidance and actually very low levels of guidance in terms of having to define your columns or however you can just type anywhere. And that allows for a lot of freedom and it lowers the barrier to entry. A lot and then the waves that we've seen. So visical on Apple II, the doors based PC came out in the early 80s and then Lotus One Two, Three went on the rise. You could open Visical Documents or Visical spreadsheets and Lotus 123, but Lotus 1 2, 3 quickly replaced VisiCalc because they were on this more powerful platform that became more ubiquitous. Then you go into the 90s, then you have the rise of Windows and the windowed way of working and Excel was first there and made use of that. And then in the 2000s you have a lot of work moving to the browser, moving to the cloud, and then Google Sheets comes in and takes advantage of that. So what you've essentially seen is you've seen these four waves. Now Google Sheets still is a contender, Excel is still the 500 or 800 pound gorilla in that market, but you have these kind of four generations there, each of them happening with a major wave in computing and making some use of that, but not fundamentally changing the way we approach this work. And I think that's for the good. But the way we think of it at Grid is first of all, we're focused exclusively on the numbers side of what people do in spreadsheet. Others like Airtable for example, have done a great job of being there for you when you need something more modern and maybe a little bit more sophisticated if you're using a spreadsheet as a database. But if you're doing numbers work and you kind of realize that the traditional spreadsheets, meaning Excel and Google Sheets, aren't quite pulling it off for you, because you are in this new generation you're using, you have your data in a notion database and you want to visualize it. Like that's just not easy to do with Google Sheets, let alone with Excel. And that's where we come in. Hook up with your notion data, allow you to easily visualize that, embed it back, everything is live, connected, you can do calculations on top of it and so on. So it's partly just being more mindful of this new technological stack. The other big innovation we bring to the table is that if you think like a computer scientist, you will often think about the data layer, the logic layer and the presentation layer. And the spreadsheet is interesting in the way that it combines all three. You know, type in two numbers, that's data, add them up, that's logic. And then you bold the result, that's presentation. And obviously spreadsheet users don't think of it that way, but this Again, is both a strength and a weakness. But while the spreadsheet has moved into the browser as a user interface, it hasn't really made use of the browser as kind of the media rich interactive UI that the web allows you to do. And that's kind of what we do. We separate, we give you a new presentation layer. And so that when you have done your thinking and pulled together your data and you've done your logic in the spreadsheet, you can then, as a modeler or as whoever pulled the data together, you can now give it a guided narrative, allow your audience to get the right context, interact with the right things without them being able to ruin your model or looking at the wrong things and so on. So that's where we come in.

That's really interesting. There's a lot to unpack there. I just like to get into a bit more about the combining of the different layers between, like, data and logic and presentation, because that's one thing where certainly if you've done any programming, you're like, ooh, that's a, it's a, it's a very weird paradigm to switch to with, with spreadsheets. But if you're a spreadsheet user, it's very, very natural. So what sort of work have you done to try and separate these things or, or make them distinguished?

So. So very early on, and actually, I think the moment when I realized that not only is there a big opportunity just in this general space, but here is a pain point, is when two things came together. First of all, I came across a survey that said that 88% of spreadsheet users on a regular basis have to share what they have pulled together in a spreadsheet with someone else. At the same time, I was hearing, as I was talking to people, a lot of uncertainty around exactly that moment. People don't like to send their Excel files or share full access to their Google Sheets models with others because they know that it can ruin the model, they can look at the wrong things and so on. So what people do instead is they copy, paste things out of spreadsheets into PowerPoints and PDFs and emails and so on to give it some narrative, write some text around it. But these are tedious to make, they're tedious to update. If you have to change the numbers, people will have different versions of things available to them and so on. So this combination of, here you have this fantastic tool for doing the thinking and the, let's say, the exploration and the thinking on the data and the logic. But it doesn't really have a proper presentation layer. People are always moving somewhere else when they need to present their findings. Isn't that something that we can combine more? Because even though people are putting charts into their spreadsheets and they may be bolding some lines to help guide the eye to the right things and so on, those are mainly for themselves and also for the visuals that they end up copying out of their spreadsheets and pasting into something else.

I think, like, everyone's experienced the horror of like copying pasting bits of spreadsheets into a report. So yeah, I'm certainly glad if that's going away.

Absolutely. And you probably also kind of. People probably also recognize this moment in the meeting where somebody asks, but what if, like something was different? And then the answer is, oh, let's rerun the numbers and I'll get back to you. Wouldn't it be great if you were able to. Because the model is there. Just going to change that around in the meeting and answer the question on the spot and kind of move on with it. So that's the other thing that we give. We allow people to interact with the model without them having the full ability to change things or ruin it or change things that you don't want people to be playing around with.

That's actually very cool. It does seem like that's one of the big problems with doing data analysis is that quite often by the time you've answered the question, the person who asked it doesn't care anymore because there's that sort of lag. And being able to do real time updates to your analyses is pretty important a lot of the time. So I'd like to get into generative AI since it's such a huge topic and this is something you've been building into Grid. So first of all, can you just tell me in what ways do you think AI can improve spreadsheets?

So maybe first, for the audience's perspective. So Grid offers a spreadsheet solution where you can edit spreadsheets in a natural way and then you can build these grid documents on top of them to present the data. So you can do that either on top of spreadsheets that are built inside of Grid, or you can pull in spreadsheets from Excel or Google Sheets or even data from Notion and App Table and other sources. So in that context, we have been thinking a lot about, like everybody in the world. We've been thinking a lot about kind of what does generative AI mean for this market and for our space? And one of the things that are fairly obvious. We are always looking for opportunities to make working with numbers less intimidating and do that by making visualization super easy, making data work very approachable, and so on. And one of the things that especially people that aren't already advanced spreadsheet users struggle with is writing formulas. So you know that there are formulas that can do a lot of different things, but you don't know the names of the functions, you don't know exactly, you know the syntax of it, and so on. So there's quite a lot of discovery that tends to happen either in the documentation or just Googling the web for solutions. And how do I write a formula that does X? We decided kind of the first step we decided to take was to help this user by integrating formula assistance into grid. So instead of writing, we actually say kind of slash, slash is the new equals. So you can start the formula by typing in or hitting the formula assistant button. And then you just type in natural language what you want your formula to do, and it will come back with a suggested formula in the right syntax and so on. And this works really, really well. And I'll be candid when I say first, when we kind of started playing with this, I thought this would be a great demo gimmick. When we had integrated it and I saw how powerful it was, I realized that this is great for novice users. This is not just a demo gimmick. This is actually useful for early spreadsheet adopters. And then I find myself using it all the time myself. Even though I consider myself fairly advanced spreadsheet user, I find myself using it all the time because it's just faster than remembering what that function was called or what exactly the syntax looks like when you're doing the lookups or the sorting or some of the more kind of intermediate complexity things that you tend to do in spreadsheets. So it's really a fantastic addition to the spreadsheet and something that I think will become table stakes in spreadsheet software within too long. At the same time, I think there are a few other areas where generative AI will play. It will probably play in the analytics industry. More generally speaking, I think it will play in several different places. But if we think about a spreadsheet in particular, you already see people doing data enrichment using generative AI. So essentially you can draw a skeleton of some data you want to fill in. Most of the demos I've seen are you have countries in the rows, and then you may have countries, capital population, those types of things. And as to the columns and then you have generative AI filling them in. This is fairly good. At the same time, most often people are working with their own proprietary data and kind of things that you would be looking up from internal systems. So at least OpenAI's models that are trained on the open web will not be able to answer those types of questions. But it hints at what could be done if you feed such models with your own business data and how much work that could save. The third area that I would like to touch upon, and this is, I think true of a lot of things that people are applying generative AI to these days. They are great assistants, they can save you a lot of work, they can be very creative, meaning they can help you explore a space that you may not go into parts of, kind of an exploration space where you might have had a blind spot before. But I think we should look at them as assistants and not as experts. So, meaning that the difference between if you have an assistant, they may do a lot of your work, they may save you a lot of time, but in the end you are responsible for their work. If you turn to an expert, you actually expect them to know more than yourself about the domain and come back with answers that you will trust almost without a second thought. And with generative AI, I think that is dangerous. We want people to be looking at them as assistants and applying them their own domain, domain knowledge and so on to their work before submitting it or accepting it. But I think one of the hardest things will be the right expectation management to explain to people that you can't just throw in a question and get a perfect answer back and don't have to think about it. And this is where in the spreadsheet world, where I think this applies is I think there is some expectation that soon financial modeler's job will be able to be solved by a generative model. You'll be able to prompt it and say, make me a five year budget for a company that is so and so and explain the characteristics of your business and that is not within the realm of the possible of the current generation of AI. And I'll just be bold and state that out loud.

That's really interesting. And I think one of the big fears with using AI is that it sometimes gets the wrong answer. And if you make mistake, particularly if it comes up with a formula in a financial spreadsheet and the number's wrong, then that could have huge implications for your business. So do you have any recommendations for how you deal with like potential Wrongness. I mean you mentioned the idea of it being an assistant, but do you have any other advice?

Well, it's not like we haven't had these problems in the past. There have. Huge errors have been made by humans making the wrong formulas. And I think in many ways the same things apply here. Apply a second set of eyes on anything that matters, try to build in checks wherever possible and so on. And there's actually, when it comes to these kind of business critical type of spreadsheets, there's a whole school of best practices in those types of modeling. And often as we know there is a whole category of just financial planning software that kind of goes, that moves that out of the spreadsheet into something a little bit more rigid, which may often be the right solution. But in general I think that it comes down to, I think this is probably the biggest thing we have to solve with generative AI generally is just how do we make sure people double check, how do we make sure that people don't trust these things blindly. And that applies inside of the spreadsheet as it does anywhere else. What we've been trying to do is build in little hints. We don't submit the formula. We show you the formula formula and the potential result so you can check it before you accept it. But there's probably more discovery needed there to just understand what's the right level of what should we say, what, what's the right level of confidence we should tell people to have in this. That's actually one of the things that generative AI doesn't do terribly well. It, it doesn't know itself how confident it is in the answers it, it returns.

Okay, so one other area where it seems like AI could be useful is in the explanation of results. So have you put in any thought into how you go from results to some kind of interpretation if you want to do reporting?

Absolutely. There have been experiments in the more broadly in the analytics industry for I want to say almost a decade now with just generated narratives around data. Before we had the current generation of generative AI, these were often very dry, like they were often templated where you had pre written text and then maybe you changed some adjectives based on if the number was positive or negative or things like that. But there were certainly things that were slightly more sophisticated than that. I find this area fascinating and I think there are definitely things there. Microsoft is making some inroads and at least the way they talk about and demo the copilot that they are introducing for the entire suite, they say it can be applied to write a narrative around your data. I'm curious to see how good it is. And I think that once again we will have to look at it as an assistant, that we will have to be very critical of the work it returns and then make sure that we read it through nicely. But it's also interesting and I've always been fascinated by how much value we perceive in the text that explains what we see in the data. So I remember I was talking to an analyst at one of the big research companies and his area of expertise was financial solutions or fintech. Essentially they offered two products. You could buy one or the other or you could buy a bundle of the two. And one product was a tracker. It tracked products in the market and showed you month by month their market share and how much it sold and things like that. And. And then there was two pagers that came out monthly that essentially made charts of the data that you could subscribe to and then explained in text what the chart was showing you. And the much more popular product was the one that had less data in it. In fact, it had less information. Like there were fewer data points in there, but it had the text explanation. And maybe obviously in some cases he was putting some perspective, like a historical perspective or like it's typical of this company to do so and so and so on. So there's a value in that as well. That was actually more popular than even the bundle of the two. So I think there's a good reason people are exploring this because numbers aren't something we're born to work with. It's a very abstract way of thinking. But words and language is something that we have an innate skill to work with. And therefore translating insights into text is a huge area. And this is something that we, we've been exploring. The interesting thing is how do you inject? Because spreadsheets themselves have very little semantic context. They have like their individual cells. They do have some relations between them. There are some labels. The labels can be put pretty much anywhere. How do you teach generative AI to read the right level of semantics into that and be able to understand. This is actually the total line and this is how the components of the. Of the revenue number and things like that. This is going to be an interesting area to keep exploring.

That's really fascinating. I like your example about the sort of the trading sort of newsletter where it's like actually having just a really short text summary was more useful than just vast amounts of data to a lot of people. I suppose if you Think about it. Well, if you're trading stuff, you already got a choice of like, buy stuff, hold stuff, sell stuff. And so getting closer to that is like, is actually pretty useful. You mentioned that you're quite an advanced spreadsheet user yourself. And I'm wondering, are the uses for AI different if you are an advanced user compared to if it's a very casual spreadsheet user or a beginner?

So in what we have implemented, I think that the highest, the biggest value comes to the novice users, the user that has maybe always been afraid to even get started because they don't even know where to begin. So that's on the kind of formula assistant side. I, however, think that more broadly speaking, generative AI will be an amplifier to pretty much every knowledge worker's job. So the more time you spend doing something, the more time you spend in spreadsheets today, the more value you will get out of generative AI once we learn how to properly apply it, because there's just a larger number to multiply. So I think that that's where we will see the most impactful solutions in this space come it's to help the people that live and breathe spreadsheets day in, day out, already do maybe five or ten times more in every productive hour that they get and what they can do today. But at the same time, we don't see ourselves as the AI spreadsheet company. We are just looking at AI as one of the tools that we can bring to make working with numbers less intimidating to the everyday knowledge worker, especially the young professional who is getting started in, in their career and isn't already an advanced spreadsheet user. So it's not AI isn't central to what we do. It's one of the tools and now a very interesting new tool that we can take a look at and say, how can we use this to help our user base? And that's actually how I think most of the software industry should be thinking about AI. You already have valuable software in a given space where you are a domain expert. How can you apply AI in that space and to the great software that's already delivering value to your target audience, rather than rethinking entirely. How can AI make this radically different? That will be the role of startups that will be big in 10 years. And yes, you may want to keep an eye on it, but your immediate opportunity as to, how can I bring this into what we already have and make that better?

That seems like really good advice. I mean, certainly that's something we've been thinking about a lot at datacamp is like we're not going to pivot to becoming an AI company, but where can we build AI into data camp just to help people learn faster? You mentioned AI is not the only thing you've been working on. And one thing I'd like to talk about is integrations. So you mentioned airtable before and that airtable sort of big thing was it's a spreadsheet, but it also helps you work alongside data in a database. And of course spreadsheets are only going to be like one tool of many if you're working with data. So how does Grid think about integrating with other software?

We try to understand the tool stack of our target audience really well. Airtable is one of the tools we see there are in this category of companies started in the last five, maybe up to 10 years and maybe especially the tech companies and companies that are on the innovative side of the spectrum. There are companies that have built almost their entire IT infrastructure on airtable. We talked about CRM before. They have their CRM in Airtable, they may have some of their billing in airtable, they may have some of the financialists in airtable and so on. So it's really for the companies that really embrace it. It's this really big and important thing. And yes, airtable has used the positioning they took was they used get rid of the spreadsheet and things like that. But they were really just talking about the database side of using spreadsheets, meaning when people are using the fact that you have this two dimensional grid to type things into as a way to store contacts or other kind of tabular data. Like I said before, our whole spiel is to be there for people when they are on the number side of things. So the way we think about integrations in that case is for people that are in airtable, they will probably also want to do projections, they will also want to do calculations, they will also want to visualize their data and report out on it, like I said before in a guided narrative in kind of an approachable way. So Grid is a fantastic tool to layer on top of airtable, having your data in there, creating charts and narratives in Grid and the reporting out on what you have in airtable and similarly notion, which has been pushing their databases quite a lot and been super successful on the wiki long form document side of things for this demographic. Like they don't have anything in terms of data visualizations, let alone interactivity and calculations. So that's where we come in and we try to plug that as well as we can so that data can flow freely between. You know, if you're. What we often see is the tool stack may be your most advanced spreadsheets may be in Google Sheets that can flow in through Grid and straight into your notion wiki page. Some of the data is in a notion database that can be combined on the same dashboard. And then maybe some data in airtable flows nicely in there, but it can also flow into Miro or into kind of one of the more visual thinking tools that you're working with. And all of this is really a breeze using Grid, but really cumbersome if you're using Google Sheets, let alone Excel.

Certainly I would agree that with Excel in Google Sheets, trying to get them to interact with other bits of software, it does take a little bit of effort to get set up. So it's nice that something you're thinking about. One other integration I'd like to talk about is with things like Python r even SQL. I mean, you mentioned databases, but as a sort of data scientist, my sort of tool of choice is going to be like Python or Rich. So how do those integrate with Grid?

I have a lot of opinions, but our main target audience is somebody that is not quite that technical. So they will probably run away screaming if they see something that looks like code. So that's not kind of our our sweet spot. However, obviously being able to take a Python function and making that available inside of a spreadsheet so that you can do calculations on the data that you have in a spreadsheet using an advanced function coming from something else is interesting and is really valuable. There are others that are taking that on much more directly. Where we will probably end up in this hierarchy is we just by making APIs available so that you can do these calls both ways and you can both write to Grid and you can also call out to external systems from Grid is probably how we will enable this. Because again, the technical sophistication of the people that are writing the Python code is probably they're going to be fine and probably even prefer having kind of a restful API that they can interact with, whether that is actually using Grid to call out to a function they've written in Python or having Python reach out to Logic that has been written in a spreadsheet. And I'll drive that example home with an example. We've seen quite a lot of people create pricing calculators using Grid. There's no doubt that the marketing department or kind of the product organization inside of A company will have a spreadsheet with their pricing plan somewhere, and then they realized it may be a little bit complicated. So we want to get it out on our web page. And up until Grid. The way to do that is you hire the web agency. The web agency came in, they rewrote the entire logic from your spreadsheet, and then they would put up a calculator that customers could use on the website, and that was that. And then there was a change to the pricing, and you would have to go back to the web agency for them to react. Do the logic that was already redone in the spreadsheet with Grid. As soon as you have made the change in the spreadsheet and you have saved that to the right place, then that is reflected in the calculator that you've already built. And if you want to do something more advanced. So, for example, if your web department wants to plug into that or build custom UI on top of it or something like that, you would be able to use an API to still use the logic that the product of the marketing organization was well suited to put into their spreadsheet, but they would never consider themselves coders or be able to write that piece of code that would come to the same results. I hope that helps to understand how we bridge here. We have this business side that is very well versed in their spreadsheet, and that's how they work. And then you have the more technical side that wants to work in code, but you can bridge between the two worlds using something like Grid.

Okay, that seems pretty useful. So in that case, what's the sort of middle case where you have business analysts who are working with Power bi, Tableau, click. All these sort of BI tools, how do they work with Grid as well?

So there's definitely an overlap between what you can do in Grid and what you can do in these other tools. And we are not here to replace Tableau or Power BI or Qlik. However, we are here to enable an audience that up until now has not been able to take the time it takes to learn these tools to do some of the same things. Not necessarily quite as sophisticated, but do things that are more akin to something that only people that knew these tools would have been doing before. The other big distinction I'll make is that BI tools are about analyzing records in databases, essentially records of things that have happened in the past. So a bi, if you want to be. You want to be quick about it, BI is a way to filter, sort, aggregate and limit results that give you views on records of things that have happened in the past. But if a business is dreaming about the future, it will be doing so in a spreadsheet model. So businesses analyze the past in bi, but they dream about the future in spreadsheets.

One maybe criticism of spreadsheets is that sometimes it has problems. Particularly there was like a case a few years back where the genetics community had to rename some genes because Excel just kept like changing the name to a date. I think it was like the 3-1- gene or something. And so I think a lot of people, well, AI is cool, but I just want to have a spreadsheet that doesn't corrupt my data. So how do you think about data quality and data integrity?

Yeah, so I agree with. I mean, all of these criticisms are there for a reason. They're all correct. And they come as a side effect of these tools trying to be very open and flexible to different types of data entry and things like that. Maybe in some cases, like in the case of like, yeah, I think it was like March 1 and Sep9 were the two that were often getting confused with the gene names. Maybe it was being too lenient. So maybe you can have like targeted criticism about exactly what it did in those cases. But more broadly speaking, I think we often forget to talk about the flip side of it, which is how much spreadsheets enable and empower business users to do that they would otherwise not be able to do. And the way I encourage people to think about that is the reason people turn to spreadsheets so much is that you're faced with a new thing. You're faced with like, we were talking about the database thing before. Maybe you have, there are a hundred customers you have to call, like for some reason, so some service broke and you have to call 100 customers or reach out to them. So you need the names, maybe a little bit of details about the incident that happened, times, phone numbers, contact information, those types of things. And you have two choices. You can either ask it to have a meeting where they can filter this view out of your CRM and maybe add some fields in the database where you can check if you've gotten a hold of them or not, and so on. An it may be willing, if you have a priority project, they may be willing to meet with you next week and then they will start the project, and then four weeks later they will be able to give you what you want. So your alternative is, I can do that. I can go kind of the official route and five weeks later be able to start my job. Or I can fire up a blank spreadsheet, maybe pull down the contact information from the CRM and just do it in a spreadsheet. And I can do that before noon, I can start calling them in the afternoon. So it's obvious where the choice comes. So when it comes to we should be careful of spreadsheets and we should know about their limitations. But we should also not underestimate the empowerment that they are before kind of everyday business workers to take care of a lot of their everyday IT needs where they themselves are able to solve for something that otherwise they would need much more technical help to do.

Okay, yeah, so there is going to be a trade off between a spreadsheet being helpful for you and trying to like fix your data and giving you that extra productivity. But occasionally it's going to do something wrong. So you've got to be careful about like checking the results. So.

Okay. And I guess we would say the same of like our IT systems. I mean, they sometimes have errors and those errors can have big problems. We are just, in software development, we tend to be, or at least we want to be much more disciplined when it comes to testing and automated test and quality assurance. But things still happen, as we know.

All right, so plotting is maybe the second area where traditionally spreadsheets have been a little bit poor, like just trying to do a histogram or a heat map, which are maybe like not the most popular plots, but they're still fairly mainstream. Very difficult to do in Excel or Google sheets. So what's grid doing around improving visualization?

So we have a bunch of visualization options and we are trying to. Our approach is we want to make sure that the basics are super easy. So that just put a lot of work into tuning our column charts, line charts and tables like these are. If you take these three visualizations together, you can portray pretty much anything. Sometimes you want to reach outside of that range to bring a particular point home, for example, to heatmaps or dot plots or something like that. So we've been implementing those as well. At the same time, the way I see this moving forward is we can probably never fulfill everybody's visualization needs with something that we built into the product. So here is where some point those more technically capable should have access to APIs where if you want to go outside of what the tool offers out of the box, you would be able to enable that and maybe you would be able to build that in a way that not only you can use it, but you can make that available to other users. Of the tool as well. So essentially an extension type model there. But this is something we think dearly about and given our target audience. Also, there's sometimes a tendency, especially when people are getting started with visualizations and so on, that they, they want to move on to something that is non traditional, when maybe a column chart is the best way for them to actually portray the point they want to get across. So we also don't want to limit them in what they can do, but we want to guide them to knowing when to use something other than just some of the basics, because they are often the right choice.

It does amaze me how far you can get with business analysis just by doing line charts and bar plots. But occasionally, yeah, you do need other plot types. So there's like this long tail of different visualizations you need to use. So that sort of extension model does seem quite reasonable. Now, the third area where I think spreadsheets struggle is with debugging. So I think everyone's come across the case where there's like there's a weird error somewhere and they end up clicking through cells to try and find out exactly which formula went wrong somewhere. So how do you think about improving the debugging experience for spreadsheets?

So this is a fascinating area, so I'll take a dive. So behind every spreadsheet is a dependency graph. And you will be surprised how complex that dependency graph can become very quickly. Meaning you make a cell, I made an example before, where you type a number in cell A1 and B1 and then C1, you add them up. Like that dependency tree is only one level deep. But then as soon as you have, like, maybe you have a financial model with 10 assumptions and you're calculating that for monthly, for five years or something like that, by now you'll have a couple of thousand nodes and the number of links between those nodes will be in the tens of thousands and the depths of the graph will probably be, I don't know, 25, 30 levels deep easily. So it's obvious that it's easy to get lost in that somewhere. And it's also obvious that it may be hard to trace exactly where in all those kind of dependencies. Going up 25 levels to hundreds of cells to come to a conclusion in a single cell can be hard. One of the privileges of the job I have is that we had to build our own spreadsheet engine from scratch. So we are by far the most powerful spreadsheet engine that can run entirely in JavaScript and thereby in your browser. That was needed to be able to make fast interactions happen with large spreadsheet models in a browser. So therefore we had to take a deep dive here and there are. We have only just begun to scratch the surface of making this available to our users, but the tools we have internally to kind of debug just as we are developing the spreadsheet engine are fascinating. So when you're able to see the intermediate results and how they flow and exactly where. So for example, if you have an error to be able to trace from the cell where the error manifests itself to where it actually first started propagating, which is something that Excel or Google Sheets do nothing to help you help you do, you will have to just realize which one of all these cells that are now throwing an error are at the root of the dependencies for what you're looking for. And I think there are a lot of opportunities to do more there. And again, we have to be mindful of what's useful to our target audience aren't necessarily the most sophisticated ones. But I've seen a major opportunity also in just applying that to a much more advanced audience where we can throw at them analysis of their spreadsheet and maybe also point out non obvious errors like omissions in cell ranges and things like that that happen all the time where you've added up all but the last two months of something and therefore you have the wrong sum somewhere. Excel tries to help you in Google Sheets as well, but if you aren't like visually looking at the right place in the model, there will be no queue for you to look at. So it's an absolutely fascinating area. And the dependency graphs obviously come up in all sorts of computer science solutions. But a calculation dependency graph like spreadsheets is a super fascinating area both to try to optimize. There are so many places where you can optimize for speed and memory and all sorts of things, but also just in the way that it manifests essentially a large algorithm. What you have in a dependency graph is an algorithm. You could write traditional code that would come to the same conclusion. Now good luck debugging that. But you have these business users that are creating these complex algorithms just by essentially offloading their thinking cell by cell or line by line, row by row into a spreadsheet model.

I find it fascinating because we normally talk about dependency graphs in the context of data engineering and scheduling, like when your different analyses are going to run and the fact that actually the spreadsheet like the engine, is using the same graphs to run the calculations and it's going to be helpful for debugging. That's a really interesting sort of coincidence.

For those that are interested in those types of things in the audience, I think they'll be fascinated to know that dependency graphs were not there in the very first spreadsheets. So the way they would calculate is they would just first calculate cell A1 and then B1 and C1 and then it would go row by row. So you could only do calculations that would refer up and to the right, up and to the left. And if the calculation was referring either down or to the right, you'd have to rerun the model a few times before you got a persistent result.

Wow, that seems pretty inefficient. Yeah, so I guess I'm glad we have graph based engines now. All right, so we talk a lot about data literacy on this podcast. And so how do you think spreadsheets can be used to help novices get started with data?

I would go as far as say spreadsheets are probably where novices start with data. So I mean it's probably the first data tool. Most of us are at least first free form data tool or like pre exploration data exploration tool is maybe the way to put it that we come across. I actually I have been surprised as I've taken a dive into this, how bad the first time user experience is with spreadsheets. They have been able to rely on a lot of ingrained knowledge inside of businesses and universities and everywhere else. But when people are starting to use spreadsheets, they can turn to somebody that already knows how to use them. That's how most people get a big chunk of their initial learning there. I think that first time user experience in general, and maybe as I alluded to before, being better about best practices, teaching people best practices while they are doing things for the first time, like okay, you want a pie chart that can be good, but here is when a pie chart is a good solution for you and here are some of the reasons you might not not want to use them. And then, oh, I, I learned something. I should be using a different type of visualization to get my point across. Whereas today we just throw people in and they have, there's no guidance neither on properly how to use the functions of the tool nor of what good data analysis and presentation of data results looks like.

Yeah, I do agree that spreadsheets are like a really good sort of starting point for putting data. And so it does seem like some level of spreadsheet competency is pretty important in almost every data role. But do you have a sense of like, what the most important spreadsheet skills are? Like, do you just need to be really good at formulas or figuring out macros or what's the deal?

I mean, obviously it depends a little bit on your role. One of the interesting things about MySpace and maybe in some ways in talking to your audience, is that the people that are least likely to rely on spreadsheets in their work are techies. Because we have other tools at our disposal to do the things that normal people have to do in spreadsheets. So we are a little bit blind to often how much they just drive, they run the world. Like the whole business side of our organizations are driven by spreadsheets. That this decision making, many of the processes, like a lot of things are just driven off of spreadsheets that we've never seen and we are largely unaware that they even exist. So I think that I'm very much kind of the numbers side of things. So I think that kind of the. Yes, getting the basic kind of formula skills right is a very important way to start. And the funny thing there is you can master. I'll give you a statistic. So the top 30 spreadsheet functions, meaning the 30 most used functions in spreadsheets, they actually cover more than 98% of all spreadsheets in the world. So only 2% of spreadsheets use functions that are not one of these kind of 30 most used ones. So kind of master the most commonly used ones and you are going to be well ahead of a lot of your colleagues. I think also after having gotten enough skill to get the job done, meaning to get your thinking out and doing the analysis, or you're getting to the results that you were looking for, then I think the skill after that is how do I communicate this? Like, how do I properly tell the story? What is the data telling me and how should I present that to someone so it makes an impact? And that is definitely something that spreadsheets don't help you with. That is something that somebody like us, that's our kind of bread and butter helping you with the narrative and the presentation. And that is, I think, the skill that comes after. First you need to learn how to work with data and think. And maybe the part I overlooked is the amount of data cleansing and manipulation you often have to do before you can start your actual work. So learning some tips and tricks there is definitely useful. But then on the other side of that, if you're not showing this to someone, it's worthless. It's Just in your head. And usually, like I said, 88% of the time people find themselves presenting what they have pulled together in a spreadsheet to someone else and that is an important skill to have.

That's kind of interesting, the idea that I think you said there's only like 30 commonly used functions or something, so. And now think about it. Yeah, you probably can get a long way with just being able to do like sum and average and then maybe the occasional if statement.

Yeah, absolutely.

So yeah, actually the hardest bit is being able to communicate what your results are to other people. Yeah, I can certainly agree with that. All right, so I know talking about the future is a bit of a mugs game, but I would like your opinions on what you think the long term future is for spreadsheets.

I think spreadsheets and the spreadsheets, the way we think about them, when I say the word right now, are probably going to be around for a long time. I'm willing to wager a bet for 20 years and I wouldn't be surprised if if it's a lot longer. There are several reasons for that. First of all, they are a pretty well proven way of doing a lot of things. Like they're very generic, open ended tools. And while ever since kind of the visical days, people have been chipping off major use cases and making proprietary or purpose built software to better do something that people have been doing in spreadsheets, new needs arise every single day as well and people turn to spreadsheets to solve them. And that is going to be like spreadsheet usage hasn't gone down with more and more proprietary and purpose built software being built. For some of these use cases it's actually gone up. And the reason is for everything you chop off, there's just more added to the long tail that kind of needs to be solved as well. The other is that there are just so many processes, so much knowledge, so many assets that already rely on things within businesses. And I think we, especially us on the kind of, I imagine a lot of people listening here are very much on the early adopters end of the curve. I think we tend to underestimate how sticky things at work can be. I've seen spreadsheets that have been updated weekly for 20 years. I'm not joking. Like there's a row added to a spreadsheet every week and that spreadsheet started 20 years ago. So these types of things happen and they happen because it works and it's what the business user has at the disposal. And there's no reason to change it if it works. Wow.

The 20 year spreadsheet. I mean, I guess having long term compatibility is amazing. It also seems like slightly terrifying that that thing exists.

No, absolutely. And the interesting thing here is that nothing interesting has happened in this space without being backwards compatible with what came before it. Some of the decisions made by Dan Brecklin, who I referred to before the maker of Visical in his dorm room in Harvard in 1978, are still the way we write formulas in spreadsheet today. Because Lotus123 had to be backwards compatible with Visical, Excel had to be backwards compatible with Lotus 123. Google Sheets was backwards compatible with Excel from day one. And even though each and every one of us are adding on some new functionality, whatever paradigm comes next will have to be backwards compatible with what we have today.

Do you have any final advice for spreadsheet users?

Yeah, I mean, obviously I'd tell them to go and try out Grid. It will add to your life, it will help you better present the things that you have pulled together. But more generally speaking, I think that maybe having a conversation with somebody that works in creating software, because what spreadsheet users often don't realize is that they are writing software. Spreadsheets are code, they're just encoding relationship between data that lives in cells instead of writing kind of lines of code that get executed one after another. But the spreadsheet world has a lot to learn from some of the discipline and even just very simple things like having a checksum somewhere, like somewhere where you know that if everything works, the result in this cell should be a certain number. Having those kind of checks in place, even though it isn't anything more than that, will often save you a lot of pain somewhere. So have a beer conversation with a computer scientist and see what you can learn from each other.

That's brilliant. I love the idea of spreadsheets as being like a stealth way to teach people programming with that. Thank you, Kjelma for being on the show. I hope you enjoyed the experience with.

This was great fun. Thank you, Richie.