Is Excel immortal?

It might not just be immortal; it might be the most immortal thing of all.

Microsoft Excel, a tremendous machine.

Suppose that you used to run a big cryptocurrency company, and you committed some crimes. Your crimes made you rich and famous at first, and then they got you arrested. All of your friends, who committed a bunch of crimes too, turned on you to save themselves. They said that you told them to commit the crimes, and that you committed all of the really bad crimes on your own. You got tried for your crimes, they testified against you, and you got sent to jail for a very long time. 

The people running the jail decide that, because you committed computer crimes, you can’t be trusted to use a computer. So, while you’re in jail, you aren’t allowed to use the internet, to make phone calls, or watch the news. Your only contact with the outside world is through the occasional music mogul who drops by.¹

One night, early in your sentence, an old man shows up in your cell. Bold entrepreneurs like you don’t belong in jail, he says; you belong in Silicon Valley, building the next internet. The world needs you in Founder Mode, and there is no Founder Mode in jail. He can buy you a pardon, on two conditions. First, you have to give him good terms on your seed round. And second, you have to prove that you’re a true visionary. Show me that you can see into the future, the man says, by answering this one question correctly: Of these twelve things, which ones will no longer exist in 2075?

1. ChatGPT

2. Twitter, Facebook, Instagram, and TikTok

3. Google Search

4. iPhones

5. JPMorgan Chase & Co.

6. Notion: Your connected workspace for wiki, docs & projects

7. NFL football

8. American democracy

9. Tesla, Inc.

10. The United States dollar

11. Living creatures of any kind—livestock and creeping things and all the beasts of the earth

12. Microsoft Excel

What do you choose?²

Obviously, I have no idea what the right answers will be.³ But—weirdly? Disturbingly? Predictably? Comfortingly?—I can imagine vaguely plausible series of events that wipe all twelve options out of existence, except one: Excel. 

OpenAI is on pace to burn $250 billion between now and 2075, and may well implode this week. Social media could get regulated into oblivion, and Gen Z may be happy to send it there. Google Search could get replaced by AI. iPhones could get replaced by AI. JPMorgan Chase is a bank, and banks vaporize themselves all the time. Notion could get bought by Salesforce and be slowly dismembered. Someone’s definitely going to die playing football. The entire country thinks that American democracy has a 50 percent chance of ending in a month. Tesla could bankrupt itself by trying to build a Gigafactory on Mars. The United States dollar could get surpassed by the renminbi, or destroyed by hyperinflation, or replaced by some seasteading crypto utopia, or all three at once. Israel could bomb Lebanon, Iran could bomb Israel, Russia could bomb everybody, ‘til all that remains is the arms of the angels.

But Excel? What’s the story that would explain how Excel goes away? Microsoft gets taken over by Sam Altman, everyone quits, and the whole company is run by one guy worth $8 trillion dollars living out of a refurbished Three Mile Island? The entire concept of money goes away, and nobody needs to keep track of budgets anymore? Somehow, those both seem more likely than a YC company that’s “building a better Excel” actually building a better Excel.

Still, surely there is a timeline out there where Excel isn’t used anymore, and not because nuclear war annihilated everything, or because AI models got installed into our brains and we all now live in the Matrix.⁴ Surely, there is a way in which computers exist and society exists and accounting exists, but Excel does not. 

Anti-patterns

In my head, there are two ongoing soundtracks to my life: Gen Z pop music, and famous calls during sports broadcasts. Someone who’s good at stuff does something good? “The legend…grows!” Something bad happens? “Uh oh..uh oh..oh no! Oh wow!” Something surprising? “I can’t spaek! I can’t spaek!”⁵ A good playlist? “It’s banger after banger for DJ Burns!” A car flys by you on the highway? “That ball’s got some heat on it!” You want something? “You gotta give us a shot!...Why don’t we get a shot?” 

And, in the weirdest mashup of niche interests and entanglements,⁶ whenever I talk to people about their BI tool:

• Me: Do you like your BI tool?

• Them: Yes, I use it frequently.

• Me: Can you show me how you use it?

• Them: I go to this dashboard that someone sent me, and scroll past all of these graphs and charts, and click “Export to Excel.” It is truly a wonderful BI tool.

• Me: Ok, huh, um, why do you want to export everything to Excel?

• Them: So that I can make graphs and charts. 

• Me, in my head: Excel, still invincible!

When “me” was Mode and “they” were a customer, I’d then go off and do the same thing that every other BI vendor does when they this story: I’d try to figure out what we needed to build to keep people in Mode and out of Excel. Because, as we all know, Excel is an anti-pattern; a workaround; a broken workflow. It’s a siloed static file; it’s a knot of opaque formulas and intermingled references; it’s a catastrophic error waiting to happen. It’s a cliché joke about final_v2_final_sept30-DO-NOT-EDIT.xls. It’s a problem, and every other data tool is part of the solution: Modern, cloud-based, live, collaborative, versioned, code-based, scalable, and many other compelling attributes that Excel is not.

Yet, Excel is still there; still undefeated; still immortal (!!!). Which shouldn’t make sense. If using live data is better than badly versioned files, why do we keep downloading CSVs? If freehand formulas are less reliable than well-governed BI tools, why are we constantly futzing with stuff in spreadsheets? If these are actually problems to avoid, why do we refuse to give them up?

The answer—obviously, probably—is that these aren’t bugs of Excel, but features. And if Excel is going to get dethroned between now and 2075, it seems we’ll need someone to build a better version of a tool that supports all of these supposed anti-patterns, and not one that tries to stamp them out. 

Save as…

Here’s a thing that happens in a lot of companies. Every week, the sales team sits down to review all of the deals that people are currently working on. There’s a big list of open sales opportunities, and the sales manager asks each rep about the status of each one. People talk through what’s going on, and then write down at least three estimates for each prospect: The likelihood that they’ll close, the amount they’ll close for, and the date they’ll close. Then, all of these numbers get tallied up, and turned into a forecast. 

It’s hard to overstate how important this process is, especially for big companies that need to tell Wall Street how much money they expect to make each quarter. In 2022, Snap missed a forecast; the next day, the company was worth 25 percent less. Last month, Bumble said they got their forecasts wrong, and their stock immediately fell by almost 30 percent. There’s a reason sales reps are sometimes asked for a “blood number:” Miss it, and you’re dead.

Given how important all of this is, it seems like the sort of thing that should be tightly managed in a SaaS tool, and tightly reported on in some rigorously governed BI tool. The last thing a teetering VP of Sales wants is to bet their job on a bunch of scattered copied and pasted spreadsheets, right?

Well. Despite most companies keeping their lists of prospects in tools like Salesforce, and despite those tools having plenty of flashy (AI-powered!) forecast features, a lot of teams still use Excel.⁷ 

Sales reps, it turns out, aren’t reliable forecasters of their own deals. Some are chronically optimistic.⁸ Some sandbag. Some miss telltale signs that a deal is about to fall apart. So, sales managers will often make their own adjustments, based on what they know about the rep and the deal. They’ll manually tinker with things: This deal will probably come in at half of what is forecasted; this deal probably will slip to next quarter; this deal is likely to close, even though the rep is nervous about it. And they don’t want to publish all this offroading in some canonical dashboard that’s broadcast to the entire office—it is the hidden thumb on the scale, for executives’ eyes only. 

Moreover, sales leaders want versions of their forecast. They don’t want to see the numbers in Salesforce, as they are today; they want a saved, immutable sept_27_forecast_pessimistic.xls, to compare to this weeks’s oct_4_forecast_without_jetblue.xls.

Live, mysteriously-changing data is bad. Collaboration is bad. Restrictions on how to manipulate data are bad. Any sort of abstraction that forces sales leaders to work through some drag-and-drop visualization builder that muddies how, exactly, the fields for “deal amount,” “opportunity ARR,” and “win-probability-adjusted revenue forecast” are summed up is bad. Excel—orphaned from all other realities, saved to desktop, and barbarically updated via emails and saved as…—is good.  

But could something be better?

Excel 2075

Of course, if I knew how to build a better Excel, I wouldn’t be writing a blog; I’d be worth $5.2 billion dollars and be living in outer space. But because I write a blog, I have opinions about what a better Excel would need:

—

It would need to be able to open local files. If someone sends you a CSV, you could double-click on it and open it in Excel 2075. Nothing will ever replace Excel if you have to log in to use it, or you have to upload—or worse, connect to—data to it. Excel is a lot of people’s default data tool because it’s literally their default data tool: Their computer defaults to it when they try to open a file of data. This implies, I suspect, that any true Excel replacement will have to be a desktop app.⁹ 

—

One way to think about Excel is as a database and a BI tool, all packaged into a single file. When you share an Excel file, you’re sharing raw data and your manipulation and presentation of that data. Most BI tools are just the latter: They’re a prism, but the light that’s shining through them comes from somewhere else.

This is an underrated benefit of Excel. It’s a fixed snapshot, a sandbox for coloring outside the lines, the original modern data stack in a box. So long as people want to play around with their data—and no matter how much we tut-tut about it, there will always be a sales forecast to sharpie over—any Excel replacement will have to be an all-in-one tool like this.

—

Why hasn’t Tableau replaced Excel? Mostly, because it doesn’t ship with 70 percent of the world’s computers. But I’d guess it has a second problem: You can’t quite manipulate data directly in Tableau. You can’t look at a bunch of raw numbers, highlight the ones you want to make a chart of, and see a graph of exactly that. There’s a layer of abstraction between your charts and your data, and that abstraction can be unnerving. Does my configuration for this chart do what I want it to? Does it aggregate everything up in the way that I want it to? Is there some missing row somewhere, or malformed date, or null value that I can’t see? Unless you know how Tableau works, you can’t truly know if the chart you see is telling you what you think it’s telling you.

Excel doesn’t put anything between you and the raw data. You can look at a table of numbers, look at the chart you made, and it immediately makes sense how you got from one to the other. No doubt, Excel has a learning curve, but because of closeness between data and result, it’s a much more intuitive learning curve than other tools. And a better Excel would probably have to be just as direct.

—

Still, for all their benefits, spreadsheets aren’t the only way to manipulate data. Sometimes, Tableau’s fancy charts are useful. Sometimes, a SQL join would be better than a hairy VLOOKUP. 

So far, most attempts to make a polyglot data tool stick notebooks in BI tools or stick spreadsheets in notebooks. But if you want to beat Excel, putting a notebook in a spreadsheet might be a better way to go. 

—

Finally, an obvious thing that a better Excel would be better at than Excel is scale. Excel still has relatively breachable computational limits. It can only handle so much data, and workbooks can only be so complex. If Excel could offload some of its work to the cloud—not by being a SaaS product, but by transparently pushing data out of the desktop app to some remote execution engine—its caps would be much higher. 

Moreover, a better Excel could also sync itself to the cloud. Very roughly, it could work like a branch of code: You work with your data locally, in your own sandbox. But then you could push your sandbox back to a central repository, or, if the central repository has changed, update your sandbox to include those changes. 

All together, none of this would feel that different than Excel does today. You download an application. Double click on a CSV, and Excel 2075 opens it up. You can manipulate everything directly in a spreadsheet. You can add more CSVs, by typing them in, or by dropping them in your workspace. Join them together; play with them with R; make some charts. Save the whole thing, and email it, as one file. If the data gets too big, or the computation gets too complex, push the file to the cloud, and use the desktop app to do your work on some giant computer in AWS. Let the app be smart about what to run locally, and what to run remotely. 

The thing is, this kind of exists. If you squint at it enough, MotherDuck, the hosted version of DuckDB,¹⁰ has a similar technical architecture.¹¹ Through DuckDB, MotherDuck supports a hybrid execution engine that lets you query data locally and remotely. It’s a sandbox and a synced service, depending on which you need. 

All they have to do now is build a MotherDuck desktop app, make it look like Excel, add charts, and reposition the entire marketing pitch to appeal to 24-year old investment bankers. 

Would it work? I mean, no, almost certainly not. But it might. And sometimes, the longest shot (!!!) wins the biggest race of them all.

1. It's very hard to find anything good to say about the New York Department of Correction, but whoever did this—I mean, wow. Someone arrest Michael Lewis and put him in that cell too.

2. Surely there’s a whole category of scam like this, right? Like:

• Offer people a chance to bet on something that won’t get resolved for many years. 

• Entice them to make the bet by giving them what seems like good odds. 

• Take their money and carefully note their bets in ways that seem very official.

• Spend their money on watches and cars and Turkish Airline business class upgrades.

• Everyone forgets? Everyone dies?

Parents seem to scam toddlers like this all the time: “We’ll get you that toy when you turn eight,” which is some distant age unimaginably far into the future. There’s no way there aren’t dozens of crypto companies doing the same thing to adults, by putting bets about 2050 “on chain,” and only requiring a small downpayment to lock in your odds. Perhaps this is why you are in jail.

Of course, it is also possible that you’re in jail because of Excel.

3. Though I guess that VC Jafar does? VC Jafar will judge your answer based on how much your opinions line up with his. A true VC, VC Jafar. 

4. But even then, there will be Excel. Look at this office. You can’t tell me these people don’t live in Excel. Neo—the Chosen One!—uses a numpad! 

5. Michael Van Gerwen, the bald dart player in this video, has a Wikipedia page that is 19,000 words long and includes 275 citations. Babe Ruth’s Wikipedia page is 20,000 words long, and has 259 citations.

6. I must stress, despite the overwhelming evidence to the contrary, that BI is not an “interest.”

7. I mean, citation needed. But this is a blog post, not a Wikipedia page about a dart player. 

8. This is also what makes them good sales reps.

9. This is the biggest problem with Google Sheets. It’s basically Excel, in a browser. There are nice things about that, but it makes it clumsy to open a file with it.

10. And the second modern data stack in a box.

11. I’m a personal investor in MotherDuck, and they definitely do not endorse this pitch.