Why Arden starts with a file, not a bank login.

Every budgeting app built in the last ten years asks for your bank login on the first screen. Mint did. Monarch does. Rocket Money does. YNAB optionally does. Empower insists on it. Copilot requires it. The assumption is so baked in that “personal finance app” and “give us your Chase password” have become the same thing in most people's minds.

Arden is built on a different assumption: that the bank login is the most fragile part of the entire category, and making it mandatory is how a whole generation of money software quietly turned into data-resale businesses.

The bank login as failure mode

Anyone who's used Mint remembers the 3 a.m. email: “We couldn't connect to your bank. Please re-authenticate.” That email arrived because bank APIs aren't real APIs. They scrape. They're fragile. They break when the bank redesigns its login page. They break when multi-factor auth prompts change. They break when the bank deploys a new CAPTCHA. Every aggregator — Plaid, Yodlee, MX, Finicity — spends half its engineering capacity treating this as a first-class problem, because it is.

You experience this as: the app that was supposed to show you your money stops showing you your money, usually at the exact moment you wanted to check. The fix is re-linking. You re-link. Four days later it breaks again.

If the only way to use your budgeting app is through a pipe that breaks monthly, the budgeting app breaks monthly too.

Worse: the moment the bank login is mandatory, “we need Plaid to work or our product is useless” becomes an internal argument that wins every time. It shapes the roadmap. Features that don't depend on the pipe get deprioritized. The product ossifies around the dependency. Users who can't connect their credit union (it happens) get abandoned as acceptable churn.

What changes when the file is first

Every bank, credit union, brokerage, and retirement provider in the United States lets you export your transactions as a file. They have to. It's a regulatory and competitive commitment that predates the Plaid era and will survive it. CSV, XLSX, OFX, QFX — pick one, they all work.

Arden treats that file as the primary input. Drag it into the import screen, we parse the columns, you confirm, we load. No bank credentials ever leave your device, because we never asked for them. No re-authentication loop. No 3 a.m. email. No “Chase is temporarily unavailable” banner.

What it looks like in practice:

• You download a CSV from Chase, Capital One, Navy Federal, Schwab, or your local credit union. Takes 10 seconds.
• You drop the file into Arden. Our parser handles the odd shapes — prefix rows from Schwab and TD, the debits-as-positive convention some banks use, columns named “Trans” or “Ref” that sibling apps can't interpret.
• You confirm column mappings on a two-row preview. Adjust any that look wrong.
• We import. The transactions are now in your budget, ready to be assigned.

The whole flow takes 30 seconds to a minute, including the file download. It works at 3 a.m. It works when Plaid is down. It works when you're offline with a CSV on your laptop. It works with the one credit union nobody else supports.

Plaid is an option, not a gate

This isn't an anti-Plaid post. Plaid is an excellent aggregator and we integrate it — for people who want automation. Once your production Plaid connection is approved, you click a button, log in once to your bank, and transactions sync in the background until you revoke the connection.

The difference is the word optional. Every feature in Arden — envelope budgeting, net worth tracking, investment holdings, reports, insights, milestones — works end-to-end with CSV imports alone. You don't lose anything by never connecting a bank. You can try the whole product before you decide whether to hand anyone your bank credentials.

A lot of people will make that decision and come down on “no,” and that's fine. Arden is the first money app designed to respect that decision rather than work around it.

The exit door

There's a subtler reason for the CSV-first architecture. If you can put CSVs in, you can also get them out. Arden exports everything — every transaction, every category, every tracking account, every investment holding — as a clean CSV (and a JSON dump for the machines). Even after you cancel. Especially after you cancel.

This isn't something we bolted on to meet a regulatory requirement. It's the same primitive as import, running in reverse. Building the product around the file meant the exit door was structurally free.

The business implications are clarifying. We can't hold your data hostage; the door is always open. So we have to keep being worth the $12 a month. So we get up every day and try to ship a better version of Arden. The incentives align because they were forced to align at the architecture layer.

What we traded for it

Being honest: CSV-first costs something. A truly automated flow — where transactions appear without you doing anything — is a genuinely good experience, and we don't have it by default. You have to download a file once a month, or once a week if you're obsessive. Plaid users get the automated version; CSV users opt into a manual rhythm.

What we think we got back:

• No dependency on fragile infrastructure. Arden works whether Plaid is up, down, or on fire.
• No data-resale pressure.Without the bank pipe as our moat, we can't monetize transaction data as a side business even if we wanted to. Paid subscription is the only viable model, which aligns our incentives with yours.
• A real answer for privacy-minded users.Not marketing copy — an actual architecture where your bank credentials don't exist in our system.
• Portability as a structural property. Your money history never lives exclusively in Arden. The file it came from is still on your disk.

The philosophical version

Most money software treats the bank connection as the core capability and everything else as a view layer. Arden inverts that: the file is the core primitive, everything else — the budget, the net worth page, the investment tracker — is a way of seeing the file. Plaid becomes one of many ways to produce files instead of the one mandatory pipe the whole product depends on.

This is the difference between “aggregator with a budgeting UI bolted on” and “personal finance software that respects the fact you already have your data.” It sounds like a small semantic distinction. It changes what gets built.

← More writing · The longer manifesto →