The take

LAMBDA, MAP, REDUCE, SCAN, BYROW. These are not new Excel functions. They are functional programming primitives, shipped one by one into the same spreadsheet your accounts team has been using since 2015, and the cumulative effect is that Excel is now a programming language with a spreadsheet UI bolted onto it. Most users are still writing nested IFs because nobody has told them the shape of the tool has changed underneath them.

Why people don't see it

The Excel UX hasn't changed. The grid still looks like the grid. The ribbon still has the same Home tab. New functions arrive without fanfare, usually announced in a blog post nobody reads, and they land next to VLOOKUP and SUMIF as if they belong to the same family. They don't.

There's also no inflection point. Microsoft did not stand up at a keynote and announce "Excel is a programming language now." They shipped LAMBDA in March 2021, then MAP/REDUCE/SCAN/BYROW/BYCOL through 2022, then named LAMBDAs in 2023, then LET, REGEX functions, and the LAMBDA Helper Functions as a coordinated set. Each release looked incremental. The cumulative shape did not.

And reasonably, most users don't have a benchmark. If your job is to track invoices and you've used Excel for a decade, the new functions look like more of the same. Nobody in the office is sending you a Loom video saying "by the way, the spreadsheet is now Lisp."

Why it's a problem

Two failure modes.

The first is that work continues to be done the old way. People build workbooks with thousands of nested-IF formulas, helper columns, and dragged-down VLOOKUPs, when the modern equivalent is one formula, defined once, called from anywhere. The old way isn't wrong, exactly. It's just five times the maintenance burden, harder to audit, and slower to change. The cost is paid by every person who inherits that workbook.

The second failure mode is harder to see. A generation of Excel users is internalising the platform's old mental model at the exact moment that mental model is becoming obsolete. They will arrive at the modern functions later, on the back foot, having to unlearn habits they could have skipped entirely. The cost shows up as "Excel feels harder now" when it has actually become enormously more capable.

The third failure mode is that the people who have noticed are quietly building things their colleagues can't read. A workbook with named LAMBDAs and MAP/REDUCE/SCAN chains is the modern equivalent of a codebase. If you hand it to someone who is still writing nested IFs, they cannot maintain it. The skill gap inside the same product is now larger than it has ever been.

The better way

Treat Excel like the language it is. Learn the LAMBDA family the same way you'd learn the function-definition syntax in any other language. The mental model is identical: a function takes arguments, does something to them, and returns a result. You name it, you call it, and the call site doesn't need to know how the function works.

Then learn the array operators. MAP for transformations. FILTER for selection. REDUCE for aggregations. SCAN for cumulative work. The pattern is the same as map, filter, and reduce in every functional language, because that's what Microsoft modelled them on.

The full mental model takes one weekend to learn and changes every formula you write afterwards. The Jun 1 Magic Monday on MAP is one entry point. The May 17 Deep End on the modern Excel landscape is the longer survey of how all the pieces fit together.

Frequently Asked Questions

Do I still need VBA?

For almost nothing. VBA still wins for tasks that genuinely need to escape the spreadsheet, like manipulating the file system or controlling another application. For everything that lives inside Excel itself, the LAMBDA family and the array functions cover the ground VBA used to cover, without the security warnings, the macro-enabled file extensions, or the version-compatibility tax.

Will these functions exist in older versions of Excel?

No. LAMBDA, MAP, REDUCE, SCAN, and the spilled-array functions require Microsoft 365 or Excel 2024. Workbooks built on them return #NAME? errors in Excel 2019 or older. If you share files across versions, this is the constraint to plan around, and it is the single biggest reason adoption is slower than it should be.

Is Excel actually a Turing-complete language now?

Yes, with the LAMBDA family it formally is. You can implement recursion, conditional branching, and arbitrary data transformations using only worksheet functions. The practical implication isn't "build operating systems in Excel," it's that the boundary between "spreadsheet" and "programming environment" has effectively dissolved.

What's the single most important new function to learn?

LAMBDA. Every other function in the modern set depends on it. MAP, REDUCE, SCAN, BYROW, and BYCOL all take a LAMBDA as their argument. If you don't have the LAMBDA mental model, the rest of the kit is unusable. Once you do, the rest is a small extension of the same idea.

Is this just a niche power-user feature, or does it actually matter for regular Excel work?

It matters for regular work. The clearest sign is that the new functions remove maintenance burden, not just complexity. A workbook with one named LAMBDA called from 200 cells is more reliable than a workbook with 200 copies of the same nested IF. Every regular Excel job that involves repeated logic benefits, whether the user thinks of it as "programming" or not.

The spreadsheet you learned in 2015 has been quietly replaced by something much more powerful, and the fact that nobody told you is the most Microsoft thing about the whole situation.