Why We'll Never Charge Per Email Scanned

Every receipt-management tool on the market eventually meters you.

Hubdoc bundles you into a QuickBooks plan that climbs with usage. Dext has tiers based on document count. AutoEntry is literally per-document above the entry tier. Even Expensify - the supposedly cheap one - caps submissions and charges per active user.

The pattern is universal. The reason is universal too: OCR has a per-document GPU cost, and per-document cost forces per-document pricing.

We do not OCR. So we will never meter you. This is not generosity. It's just a different cost structure.

Here's the math.

Table of Contents

• Why every OCR-based receipt app meters
• Why we can be free for one inbox forever
• So what do we charge for
• Forward filing is steady-state pennies
• "But what about your servers / engineers / VC?"
• Why cap exists at all
• The "no per-email metering" version of the contract
• Why this isn't generosity
• TL;DR

Why every OCR-based receipt app meters

Modern OCR runs on GPUs. Per page, a high-quality OCR pass costs somewhere between half a cent and three cents in raw cloud compute, depending on the engine (Google Document AI, Amazon Textract, Microsoft Form Recognizer, Tesseract on managed GPUs).

Add the SaaS company's overhead - storage, indexing, search, the dashboards, support - and the marginal cost of "process one more receipt" lands around 5 to 12 cents per document.

If you process 500 receipts a year on Hubdoc, that's $25 to $60 in compute costs alone. Add gross margin (usually 75%+), and you can back out why the entry tier is $35 a month: they need that revenue to cover the compute on the heaviest users while still paying for the rest of the business.

This is honest accounting. It's not predatory. It's just what OCR-as-a-service looks like.

The consequence is that every OCR app eventually faces a choice:

1. Cap submissions. Free or cheap tier with a hard limit (Expensify's "active user" model).
2. Tier by document count. Pay more as you process more (Dext, AutoEntry).
3. Bundle into a bigger product. Make OCR a feature of accounting software you already pay for (Hubdoc inside QuickBooks).

All three paths route to the same place: you pay more as you do more. Per-receipt pricing is the inevitable shape of an OCR-based business.

Why we can be free for one inbox forever

We don't OCR. So our cost structure is wildly different.

What does AutoFileEmail actually do per filed attachment?

1. One Gmail API call to fetch the message and attachment.
2. One Drive API call to upload the file into the right folder.
3. Two database writes (one to mark the message processed, one to record the filed document).

That's it. No OCR. No image processing. No structured-data extraction. No search index.

Cost per filing? Roughly:

• Gmail API call: included in Google's free tier up to 250 quota units per second per user. We use ~5 quota units per filed attachment. Effectively free at our scale.
• Drive API call: same story; included in Google's free tier.
• Database writes: small fractions of a cent on Supabase.
• Our compute: a Cloudflare Worker invocation, ~$0.50 per million invocations.

Round numbers: we pay something like 0.001 to 0.005 cents per filed attachment. Three or four orders of magnitude cheaper than OCR.

500 attachments a year for one user costs us maybe 2 cents in marginal cost.

At 2 cents a year, "free" is not a marketing concession. It's just the actual cost.

So what do we charge for

Two things, both honest.

Backfill Packs ($29 / $59 / $99, one-time, no refunds.)

Backfill is the bursty part. When a user signs up and asks us to file 3 years of historical email, that's potentially 30,000 messages processed in an afternoon. Even though our per-message cost is tiny, the Gmail API has rate limits and quota costs that show up at this scale, plus we need to allocate compute slots to keep the user's queue moving without blocking other users.

The Tax Year Pack ($29) covers ~1 year of history. The 3-Year Pack ($59) covers up to 3 years. The Lifetime Pack ($99) covers everything in your inbox up to a 50,000-email cap.

That cap is honest fair use, not a meter. We're not going to hit you with overage charges if you blow past it; we'd rather have a hard stop you can predict than a per-email creep that surprises you.

For the full breakdown of when each Pack makes sense, see backfilling 3 years of receipts in an afternoon.

Pro tier ($14.99/mo or $149/yr) - coming post-MVP.

When we ship multi-inbox support, the value-add justifies a real subscription. 5 inboxes, all clouds, advanced filtering. The 1-inbox tier stays free.

That's the whole price card. No per-receipt fee, no overage, no document-count tier, no "active user" cap.

Forward filing is steady-state pennies

The day-to-day "new email arrives, file it" workflow is the cheapest part of our infrastructure. Every 30 seconds we poll your connected inbox; if there's a new attachment, we file it.

Per user per month: maybe 50,000 polling calls (mostly returning "no new messages") and maybe 50 actual filings. The polling is essentially free; the filings are essentially free.

For one user that's under $0.01 a month in marginal infrastructure cost. For 10,000 users? Around $100. For 100,000 users? Around $1000.

This is why "free for one inbox forever" composes. The unit economics aren't strained by free users. The fixed costs (engineering, support, marketing) get amortized across paying users (Backfill Packs, eventual Pro). Free users don't break the model; they're just cheap users.

Compare to Hubdoc, where every free user would cost them real GPU dollars on every receipt. They literally cannot offer "free for one user forever." We literally can.

"But what about your servers / engineers / VC?"

Reasonable question. Here's the answer.

Servers: Cloudflare Workers + Supabase. Total monthly cost at our current load is under $100. Storage is essentially zero because we don't store the PDFs - they live in your Drive.

Engineers: small team. Lifestyle business, not a VC growth-at-all-costs play. Our ramp is 2-3 years to sustainability via Backfill Packs and eventual Pro subscriptions, not 18 months to Series A.

VC: none. No pressure to push pricing or features that would compromise the "we don't read your stuff" contract.

This last one matters. A lot of "free forever" pitches from VC-backed startups end with the company running out of runway and either jacking up prices, monetizing your data, or shutting down. We're optimizing for none of those failure modes.

The pricing model has to actually work without funny business. Ours does because the cost structure is structurally different from OCR-based competitors. If we ever drift toward OCR, the structural advantage disappears and the pricing shifts. So we won't drift. The architecture is the moat.

Why cap exists at all

The Lifetime Pack has a 50,000-email cap. People sometimes ask why.

It's there for one reason: protecting steady-state economics from the long-tail "I have 17 years of email and want all of it filed" user. That user exists. They're rare, but at zero variable cost we'd quietly subsidize them forever.

50,000 emails covers basically everyone. Mid-size freelancer with 10 years of inbox history? ~5,000 receipts. Heavy SaaS founder with vendor sprawl? ~15,000. The cap is there to bound the worst case, not to extract money on the median.

If you somehow hit 50,000 and need more, email us. We'll figure something out. There's no overage charge.

The "no per-email metering" version of the contract

Here's what "we don't meter" actually means in practice:

• Connect 1 inbox - free, forever.
• Forward filing of new attachments - free, forever, no count cap.
• Drive folder structure - free, no folder cap.
• 30-day historical preview on signup - free.
• Backfill beyond 30 days - one-time Backfill Pack.
• 5 inboxes (when Pro ships) - $14.99/mo or $149/yr, fixed.

That's the entire pricing surface. There is no number-of-receipts variable in any of those line items. We have no API endpoint that increments a per-user receipt counter for billing purposes.

For the side-by-side with the OCR-everything tools, see Hubdoc vs us. The price differences are not subtle once you actually do the math at 500 receipts a year.

Why this isn't generosity

Founders sometimes pitch "free forever" tiers as a generosity flex. We don't.

Free for one inbox is what the cost structure naturally allows. We'd be price-gouging if we charged $5/mo for the 1-inbox tier when the marginal cost is fractions of a cent. The pricing reflects the architecture; the architecture reflects the philosophy ("we don't read your stuff"); the philosophy is the product.

OCR costs money. We don't OCR. So we don't charge for the thing we don't do.

That's it.

If we ever started OCR'ing - even quietly - the cost structure would flip and we'd have to meter. Which is one of the strongest reasons we'll stay anti-OCR forever: meter-by-receipt is bad UX, bad privacy, and we don't want to be there.

TL;DR

• Per-receipt pricing exists because OCR has per-receipt GPU costs.
• We don't OCR, so per-receipt costs don't exist for us.
• 1 inbox + forward filing = free, forever, no caps.
• Backfill Packs ($29 / $59 / $99) cover the bursty one-time backfill load.
• Pro ($14.99/mo, post-MVP) covers multi-inbox.
• 50,000-email Lifetime cap is honest fair use, not a meter.
• This isn't generosity, it's just the architecture.

Free is sustainable when reading isn't the product. AutoFileEmail is free for one inbox at autofile.email. Connect Gmail and Drive, walk away. No GPU bill, no per-receipt counter, no overage email at the end of the month.

We don't read your receipts. We just file them.