Backfilling 3 Years of Receipts in an Afternoon
It's Tuesday. You open your email. There it is. A letter from the IRS. The subject line ends in "...regarding your 2023 tax return."
You stop breathing for a second.
You read it. They're not auditing you. They're "requesting documentation" for the Schedule C deductions you took. Receipts. For 2023, 2024, and 2025. Within 30 days.
Three years of receipts. From an inbox where you've been "filing" them by leaving them in Gmail since 2022.
You pour coffee. You open Gmail. You search has:attachment from:stripe. 142 results. You search has:attachment from:aws. 218 results. You search has:attachment from:notion. 47. You start counting in your head and stop somewhere past 800.
The good news: this is fixable in an afternoon.
Table of Contents
- The math nobody runs
- The afternoon version
- What the result looks like
- Vendors you forgot about
- What backfill cannot do
- What it costs vs. doing it yourself
- How the backfill job actually runs
The math nobody runs
Let's count what's actually in there.
A typical solo founder gets receipts from 15-30 vendors. Some monthly (Stripe, AWS, Notion, Linear, Figma, Cursor). Some annual (domain registrars, Adobe yearly plans, accountant software). Some weird and bursty (Vercel after a usage spike, Cloudflare when you bought a domain you forgot about).
A reasonable average: 50 receipts a month landing in your inbox. Some months 30, some 80. Doesn't matter for the math.
Three years × 12 months × 50 = 1,800 PDFs.
If you do this manually, the workflow is:
- Open email. (3 seconds)
- Right-click attachment, "Save link as." (4 seconds)
- Pick the right Drive folder, or create one if it doesn't exist. (15-30 seconds)
- Rename the file because half of them are called
invoice.pdf. (10 seconds) - Drag to Drive. (5 seconds)
Average per receipt: 30-60 seconds, if you're moving fast and not getting distracted.
1,800 receipts × 45 seconds = 22.5 hours.
That's three full workdays. Of right-clicking. Without a single break, while not getting any other work done.
The afternoon version
The 3-Year Backfill Pack is $59. It walks the inbox once. Lands every PDF in /AutoFileEmail/{vendor}/{YYYY}/{MM}/.
Rough timeline:
- Sign up: 30 seconds
- Connect Gmail: 30 seconds
- Connect Drive: 30 seconds
- Buy the 3-Year Pack: 60 seconds
- Pick the date range (e.g. "Jan 1 2023 to today"): 10 seconds
- Confirm: 1 click
- Wait
The "wait" part depends on inbox size. For 1,800 attachments across three years, expect 30-90 minutes of unattended runtime. Gmail's API rate limits are the bottleneck, not our code. We file roughly 30-60 attachments per minute on a healthy connection.
You start it after lunch. You go reply to the IRS letter, reassure your CPA, and make another coffee. By 4pm the folder is built.
What the result looks like
Open Drive. Navigate to /AutoFileEmail/. You see something like:
Every receipt where it should be. Vendor folder, year, month. Original PDF, prefixed with the email date so it sorts correctly.
The AWS folder will have ~36 PDFs, neatly nested into 36 month folders. The Stripe folder will have more, because Stripe sends multiple email types per month (payouts, invoices, dispute notifications).
This is the structure your accountant wants. We did a whole article on the folder structure if you're curious why this specific layout.
Vendors you forgot about
This is the moment most people don't expect.
When you open the backfilled folder, you will find vendors you forgot you used. Some examples from real users (anonymized):
- An Adobe trial from 2023 that auto-renewed to a $239 annual plan you didn't notice.
- A Webflow charge from 2024 for a project you abandoned.
- Three separate domain registrar accounts (Namecheap, Porkbun, Cloudflare Registrar) because you couldn't decide which one you preferred.
- A Notion AI add-on. A second Notion workspace. A third Notion workspace owned by the founder of a side-project that died.
- The "free" GitHub Copilot trial that became a $19/mo charge after the trial. For 18 months.
- ConvertKit (now Kit). Substack Pro. Beehiiv. Did you really need three newsletter platforms in 2024?
These are deductible expenses that your past self either forgot or never told the accountant about. Backfilling surfaces them.
The flip side: you may also surface charges you should cancel. The audit was annoying, but the audit-prep just saved you ~$200/mo in dead subscriptions.
What backfill cannot do
Honest about the limits.
Receipts that aren't attachments. If a vendor sends an HTML email receipt with no PDF attached, we can't file it. There's no document to file. You'd need to print-to-PDF those manually if your accountant insists on having them as files. Most CPAs accept the email itself as evidence.
Vendors that consolidate. Apple sends a single monthly receipt for all your subscriptions. So your apple.com folder will have ~36 PDFs (one per month, three years), not one per app. Your CPA can handle this, but you'll need to do the breakdown if you want per-app categorization.
Receipts older than your retention. Gmail has no inherent retention limit, but if you delete receipt emails, they're gone. We can only file what's still in the inbox.
Outlook, iCloud, IMAP backfill. Today, backfill works for Gmail. Outlook+M365 is on the roadmap. iCloud Mail too. Sign up and we'll email when those are live.
More than 50,000 emails. The Lifetime Backfill Pack ($99) caps at 50,000 emails for one walk. Solo founders almost never hit this. If you have a marketing inbox with newsletters mixed in, the cap can sneak up on you. The cap is to keep us from accidentally OOM-killing on someone's 200K-email inbox.
What it costs vs. doing it yourself
The honest comparison.
DIY: 22.5 hours × your hourly rate. If you bill yourself at $75/hr, that's $1,687.50 of opportunity cost. Plus the 3-day delay before you can respond to the IRS. Plus the 12% chance you give up halfway through and just send the IRS what you have.
3-Year Pack: $59. One afternoon, mostly unattended. Same result.
Hubdoc + 3-year backfill: Hubdoc charges per receipt for backfill at OCR rates. Plus the $30+/mo subscription. We've seen Hubdoc backfill quotes for solo founders at $300-600 for similar volumes. Then ongoing monthly fees forever.
Receipt Bank/Dext + 3-year backfill: Similar shape. They're built for accounting firms, who pass costs to clients. Solo founders pay retail.
The $59 number isn't pricing magic. It's that we don't OCR. (More on why we don't OCR your receipts.) We're moving files, not reading them. Moving files is cheap. The price reflects the actual work.
How the backfill job actually runs
For the technically curious, the engine in plain English:
- Authenticate with Gmail (read-only scope, encrypted refresh token).
- List message IDs in the date range. Gmail's API returns these in batches of 500. For three years, expect 5,000-30,000 message IDs depending on your inbox.
- For each message, check if it has attachments. If yes, fetch the message metadata and the attachment.
- Pull the sender's registrable domain (
noreply@billing.stripe.com->stripe.com). - Compute the target Drive path:
/AutoFileEmail/stripe.com/2024/03/. - Create the folder if it doesn't exist (cached, so we don't re-create on every file).
- Upload the attachment with a date-prefixed filename.
- Record
(account_id, message_id, filename)in our metadata table for dedup. - Move on.
The dedup index is the part that matters when forward-filing is also running. If a backfill job and a forward-filing job both encounter the same email, only one writes the file. No duplicates.
The IRS angle
If you're reading this article because of an IRS letter, a few things to know:
The IRS does not care about your folder structure. They care that the receipt is genuine, matches a transaction in your bank or card statement, and corresponds to the deduction you took.
What they want to see: original PDFs from the vendor, with the vendor's name and date visible. The PDFs we file are bit-for-bit identical to what the vendor sent. They're as audit-ready as the originals.
What they want NOT to see: edited PDFs, "summary" spreadsheets in place of receipts, OCR'd text without the original. The original PDF is the receipt. Anything derived from it is supplementary.
You don't need to print these. You don't need to physically organize anything. A Drive folder shared with your accountant (who shares it with the IRS via their portal, if requested) is the modern audit-response format.
The bottom line
Three years of receipts is a one-afternoon problem if you have the right tool and a $59 budget. It's a three-day problem if you do it manually.
The receipts already exist in your Gmail. You don't have to find them - they're already there. You just need them to land in the right folders. That's a filing job, not an extraction job, not an OCR job.
If you're staring at an audit letter, or just at January's "this is the year I'll be organized" feeling, pick a Backfill Pack and run it. It's $29 for one tax year, $59 for three years, $99 for everything to ever exist.
For the ongoing job after the backfill finishes: forward filing is free for one inbox, keeps the structure current as new receipts arrive. The freelancer workflow is built around this loop.
Worst case: it's an afternoon. Best case: it's the difference between a calm IRS response and a panic.
The last time you'll dread tax season.
Connect Gmail and Drive, watch the 30-day preview file itself, and never think about new email attachments again. Forward filing is free, forever. When tax season comes, grab a Backfill Pack and we'll sweep the rest of your history.