Paperless Expense Reporting: Build the Monthly Report From Receipt Data, Not Paper
Paperless expense reporting means the period report is assembled from data you already have instead of retyped from paper. Each receipt has been captured and read into a date, a total, and a merchant. The corporate card provider sends a statement file. Mileage and per diem lines are calculated from a trip log. The report is what you get when you join those three sources for one employee and one period, flag what does not line up, and export the result in a format your accounting system can import. Nobody staples anything, and the approver reviews a short list of exceptions instead of a stack of slips.
This article covers that report layer. How a single receipt gets photographed, run through OCR, checked, and turned into a claim line is covered in our paperless expense management guide, so it is only summarized here. What follows starts at the point where you have a pile of extracted receipt lines and a card statement and need a report that finance will accept.
What a paperless expense report contains
A useful expense report is a table with one row per expense and enough columns to answer four questions: what was spent, how it was paid, what proves it, and where it should be booked. In practice that means:
- Line ID, date, merchant, and amount
- Payment method: corporate card or out of pocket
- The matching card transaction ID, when there is one
- Category, plus project or cost center
- A link to the receipt image or PDF, or a note that no receipt exists
- Business purpose, written by the employee
- Flags from the checks described below
The receipt fields come from OCR. The card fields come from the statement. Category, project, and purpose come from the employee, because no receipt prints them. A report that keeps those three origins in separate columns is much easier to audit than one where somebody typed a single “amount” from wherever it happened to be visible.
On the US side, IRS Publication 463 describes what an employee hands over under an accountable plan: a statement of expense, account book, diary, or similar record with each expense entered at or near the time it happened, plus documentary evidence such as receipts. It also says a record prepared on a computer counts as an adequate record. A well-built expense report is that statement of expense.
Step 1: collect the period’s lines from three sources
Start every report by pulling these together for one employee and one period, usually a calendar month or a trip:
- Receipt lines. After receipt OCR, each one carries
receipt_date,total_amount, and optionallyseller_namefrom OCRskill’s/ocr.jsonendpoint, plus the category, project, purpose, and payment method the employee entered at capture time. - Card statement lines. Most card programs let an administrator export transactions as CSV. If all you have is the PDF statement, OCRskill’s
/ocrendpoint can return it as Markdown per page with tables preserved, which you can parse into rows. Check the parsed totals against the statement total before trusting them. - Calculated lines. Mileage, per diem, and other allowances have no receipt. They come from a trip log and a rate table.
Keep the period boundary strict. A receipt dated on the 30th that posts to the card on the 2nd belongs to one period only, and your matching window has to allow for that gap or the transaction will show up as a missing receipt next month.
Step 2: reconcile receipts against the card statement
Corporate card reconciliation is where most of the time in a manual report goes, and most of it is mechanical, which makes it the part of expense report automation that pays off first. For each card transaction you want exactly one of three outcomes: a matching receipt, a missing-receipt note, or a dispute. For each card-paid receipt you want a matching transaction.
The rules that work in practice:
- Match on amount first. An exact match within a few days is almost always right.
- Allow a date window. Card posting dates trail purchase dates, and hotel charges often post at checkout, days after the folio date. Three days is a common starting point.
- Allow a tolerance only where it makes sense. A restaurant card charge is often higher than the itemized bill because the tip was added on the slip. Allow that for meals, within a cap, and flag the difference instead of hiding it.
- Use the merchant name as a tie-breaker, not as the main key. Card descriptors look like
SQ *EXAMPLE BISTROorMETRO CAB CO 4471, while the receipt says “Example Bistro”. Token overlap after dropping processor prefixes is enough to choose between two equal amounts. - Match one to one. Once a receipt is paired with a transaction, neither can be used again. Without this rule, two coffees of the same price on the same day both claim the same card line.
Currency is the one thing the receipt cannot settle. A receipt in euros and a card line in dollars will never match on amount. Take the settled amount from the card statement and keep the receipt amount and currency next to it, as the expense management guide already recommends.
Step 3: handle missing receipts and out-of-pocket lines
Two kinds of line fall out of reconciliation.
A card transaction with no receipt needs either a receipt or a written explanation. Publication 463 says documentary evidence is not needed for an expense, other than lodging, that is under $75, or for a transportation expense where a receipt is not readily available. You still need the record of amount, date, place, and business purpose. So a sensible report rule is: under $75, accept a missing-receipt note with a purpose; at $75 or more, or for any lodging, chase the receipt. Your own policy can be stricter, and many are.
When the employee says they only have a card slip for a meal, that is a different problem: the receipt exists but proves too little. The non-itemized receipt post shows how to catch that case at capture time.
A receipt marked “card” with no card transaction usually means one of three things: the charge has not posted yet, it was paid with a different card, or the employee actually paid personally. Leave it in an exceptions state and recheck it when the next statement arrives, instead of guessing.
Out-of-pocket receipts skip reconciliation entirely. They go into the reimbursement total, and the duplicate check from the capture stage is what stops them from also being claimed on the card.
Step 4: add mileage and per diem lines
Calculated lines are part of the report but need different evidence. For a personal car used for business, the line is miles multiplied by a rate. The IRS standard mileage rates page lists 72.5 cents per mile for January 1 to June 30, 2026 and 76 cents per mile for July 1 to December 31, 2026. A mid-year change like that is easy to get wrong in a spreadsheet, so pick the rate from the trip date, not from the report date.
Per diem works the same way: a daily rate for the locality times the number of days, with partial days prorated by your policy. Publication 463 says a per diem or car allowance satisfies the adequate accounting requirement for the amount of the expenses only when the allowance is not more than the federal rate and the employee proves the time, place, and business purpose within a reasonable period. It also notes that documentary evidence is not needed for meals or lodging covered by a per diem allowance under an accountable plan. Federal rates by location are on the GSA per diem page. The report therefore needs the trip dates, destination, and purpose for each allowance line, even though there is no receipt to attach.
Step 5: run report-level policy checks
Some checks only make sense once the whole period is in one place:
| Check | Why it needs the full report |
|---|---|
| Late lines | Publication 463 treats accounting within 60 days after the expense was paid or incurred as a reasonable period under an accountable plan. Compare each line date with the report submission date. |
| Missing receipts at or above your threshold | Only visible after reconciliation. |
| Card amount above receipt amount | Tips, extras, or a wrong match. A person decides which. |
| Duplicates across periods | Same merchant, date, and amount in last month’s report, or claimed as both card and out of pocket. |
| Per-day or per-trip caps | A meal limit is often per day, which a single receipt cannot test. |
| Personal spend on the card | Lines the employee marks as personal become an amount owed back. |
Publication 463 also treats returning excess reimbursement within 120 days as a reasonable period, as well as answering within 120 days a periodic statement, issued at least quarterly, that asks the employee to return or account for outstanding advances. If you pay travel advances, put the outstanding advance on the report so it is netted against the reimbursement instead of tracked in a separate spreadsheet.
A working example: receipts plus card statement in, report CSV out
This Python script takes extracted receipt lines as JSON, a card statement as CSV, and a report date. It matches receipts to card lines with the rules above, adds mileage lines at the rate for their date, flags missing receipts and late lines, writes a report CSV, and prints totals by project and category. It uses only the standard library.
The receipt file is what the capture step produces: the three OCR fields plus the employee’s inputs. All values here are fictional.
[
{
"id": "R5",
"type": "receipt",
"receipt_date": "2026-09-04",
"total_amount": 38.0,
"seller_name": "Quick Print Shop",
"category": "supplies",
"project": "ACME-ROLLOUT",
"paid_by": "personal"
},
{
"id": "R1",
"type": "receipt",
"receipt_date": "2026-10-06",
"total_amount": 186.4,
"seller_name": "Example Bistro",
"category": "meals",
"project": "ACME-ROLLOUT",
"paid_by": "card"
},
{
"id": "R2",
"type": "receipt",
"receipt_date": "2026-10-06",
"total_amount": 412.0,
"seller_name": "Harbor View Hotel",
"category": "lodging",
"project": "ACME-ROLLOUT",
"paid_by": "card"
},
{
"id": "R3",
"type": "receipt",
"receipt_date": "2026-10-08",
"total_amount": 23.5,
"seller_name": "Metro Cab Co",
"category": "transport",
"project": "ACME-ROLLOUT",
"paid_by": "card"
},
{
"id": "R4",
"type": "receipt",
"receipt_date": "2026-10-09",
"total_amount": 14.2,
"seller_name": "Corner Office Supply",
"category": "supplies",
"project": "INTERNAL",
"paid_by": "personal"
},
{
"id": "M1",
"type": "mileage",
"date": "2026-10-12",
"miles": 84,
"project": "ACME-ROLLOUT"
}
]
The card export:
txn_id,date,merchant,amount
T901,2026-10-07,SQ *EXAMPLE BISTRO,214.40
T902,2026-10-08,HARBOR VIEW HOTEL,412.00
T903,2026-10-08,METRO CAB CO 4471,23.50
T904,2026-10-10,AIRPORT PARKING GARAGE,96.00
T905,2026-10-11,COFFEE CORNER,6.75
The script, saved as build_report.py:
import csv
import json
import re
import sys
from datetime import date
from decimal import Decimal
DATE_WINDOW = 3 # days between receipt date and card date
TIP_TOLERANCE = Decimal("0.25") # card may exceed a meal receipt by up to 25%
SUBMIT_WINDOW = 60 # days, from your own policy
RECEIPT_REQUIRED_FROM = Decimal("75.00")
MILEAGE_RATES = [ # IRS business rate, USD per mile
(date(2026, 1, 1), Decimal("0.725")),
(date(2026, 7, 1), Decimal("0.76")),
]
CENT = Decimal("0.01")
NOISE = {"sq", "tst", "srl", "inc", "llc", "ltd", "the", "pos"}
def tokens(text):
return {t for t in re.findall(r"[a-z0-9]+", text.lower()) if t not in NOISE}
def mileage_rate(day):
rate = None
for start, value in MILEAGE_RATES:
if day >= start:
rate = value
return rate
def candidates(receipt, txn):
r_amt, c_amt = receipt["amount"], txn["amount"]
gap = abs((receipt["date"] - txn["date"]).days)
if gap > DATE_WINDOW:
return None
if r_amt == c_amt:
kind = "matched"
elif receipt["category"] == "meals" and r_amt < c_amt <= r_amt * (1 + TIP_TOLERANCE):
kind = "matched, card higher (tip?)"
else:
return None
overlap = len(tokens(receipt["merchant"]) & tokens(txn["merchant"]))
return (abs(c_amt - r_amt), gap, -overlap), kind
def build(receipts_path, card_path, report_date):
with open(receipts_path) as f:
lines = json.load(f)
with open(card_path, newline="") as f:
card = [dict(row, amount=Decimal(row["amount"]),
date=date.fromisoformat(row["date"])) for row in csv.DictReader(f)]
receipts, rows = [], []
for line in lines:
if line["type"] == "mileage":
day = date.fromisoformat(line["date"])
rate = mileage_rate(day)
amount = (Decimal(str(line["miles"])) * rate).quantize(CENT)
rows.append({"line": line["id"], "date": day, "merchant": "mileage",
"category": "mileage", "project": line["project"],
"amount": amount, "receipt_amount": "",
"status": f"{line['miles']} mi at {rate}",
"card_txn": "", "flags": []})
else:
receipts.append({"id": line["id"],
"date": date.fromisoformat(line["receipt_date"]),
"amount": Decimal(str(line["total_amount"])).quantize(CENT),
"merchant": line.get("seller_name", ""),
"category": line["category"], "project": line["project"],
"paid_by": line["paid_by"]})
pairs = []
for r in receipts:
if r["paid_by"] != "card":
continue
for t in card:
found = candidates(r, t)
if found:
pairs.append((found[0], r["id"], t["txn_id"], found[1]))
pairs.sort()
used_r, used_t, match = set(), set(), {}
for _, rid, tid, kind in pairs:
if rid not in used_r and tid not in used_t:
used_r.add(rid)
used_t.add(tid)
match[rid] = (tid, kind)
by_txn = {t["txn_id"]: t for t in card}
for r in receipts:
row = {"line": r["id"], "date": r["date"], "merchant": r["merchant"],
"category": r["category"], "project": r["project"],
"amount": r["amount"], "receipt_amount": r["amount"],
"card_txn": "", "flags": []}
if r["paid_by"] != "card":
row["status"] = "out of pocket"
elif r["id"] in match:
tid, kind = match[r["id"]]
row.update(status=kind, card_txn=tid, amount=by_txn[tid]["amount"])
if row["amount"] != r["amount"]:
row["flags"].append(f"card {row['amount'] - r['amount']} above receipt")
else:
row["status"] = "no card line found"
row["flags"].append("check the statement or mark as out of pocket")
rows.append(row)
for t in card:
if t["txn_id"] not in used_t:
flags = ["missing receipt"]
if t["amount"] >= RECEIPT_REQUIRED_FROM:
flags.append(f"{RECEIPT_REQUIRED_FROM} or more, receipt needed")
rows.append({"line": "", "date": t["date"], "merchant": t["merchant"],
"category": "", "project": "", "amount": t["amount"],
"receipt_amount": "",
"status": "card line without receipt",
"card_txn": t["txn_id"], "flags": flags})
for row in rows:
if (report_date - row["date"]).days > SUBMIT_WINDOW:
row["flags"].append("older than submit window")
rows.sort(key=lambda r: (r["date"], r["line"]))
return rows
if __name__ == "__main__":
receipts_path, card_path, report_day, out_path = sys.argv[1:5]
rows = build(receipts_path, card_path, date.fromisoformat(report_day))
with open(out_path, "w", newline="") as f:
out = csv.writer(f)
out.writerow(["line", "date", "merchant", "category", "project", "amount",
"receipt_amount", "status", "card_txn", "flags"])
for r in rows:
out.writerow([r["line"], r["date"], r["merchant"], r["category"],
r["project"], r["amount"], r["receipt_amount"],
r["status"], r["card_txn"], "; ".join(r["flags"])])
totals = {}
for r in rows:
if r["category"]:
key = (r["project"], r["category"])
totals[key] = totals.get(key, Decimal("0")) + r["amount"]
for (project, category), amount in sorted(totals.items()):
print(f"{project:<14} {category:<10} {amount:>8}")
open_items = [r for r in rows if r["flags"]]
print(f"{len(rows)} lines, {len(open_items)} with flags")
Run it for a report submitted on November 10, 2026:
python3 build_report.py receipts.json card.csv 2026-11-10 report.csv
It prints the totals:
ACME-ROLLOUT lodging 412.00
ACME-ROLLOUT meals 214.40
ACME-ROLLOUT mileage 63.84
ACME-ROLLOUT supplies 38.00
ACME-ROLLOUT transport 23.50
INTERNAL supplies 14.20
8 lines, 4 with flags
And report.csv contains:
line,date,merchant,category,project,amount,receipt_amount,status,card_txn,flags
R5,2026-09-04,Quick Print Shop,supplies,ACME-ROLLOUT,38.00,38.00,out of pocket,,older than submit window
R1,2026-10-06,Example Bistro,meals,ACME-ROLLOUT,214.40,186.40,"matched, card higher (tip?)",T901,card 28.00 above receipt
R2,2026-10-06,Harbor View Hotel,lodging,ACME-ROLLOUT,412.00,412.00,matched,T902,
R3,2026-10-08,Metro Cab Co,transport,ACME-ROLLOUT,23.50,23.50,matched,T903,
R4,2026-10-09,Corner Office Supply,supplies,INTERNAL,14.20,14.20,out of pocket,,
,2026-10-10,AIRPORT PARKING GARAGE,,,96.00,,card line without receipt,T904,"missing receipt; 75.00 or more, receipt needed"
,2026-10-11,COFFEE CORNER,,,6.75,,card line without receipt,T905,missing receipt
M1,2026-10-12,mileage,mileage,ACME-ROLLOUT,63.84,,84 mi at 0.76,,
Reading the result row by row:
- The bistro receipt says 186.40 and the card line says 214.40. Because it is a meal and the difference is within 25%, the script pairs them and flags the 28.00 gap. The approver checks the slip for a written tip.
- The hotel matches exactly even though the card posted two days after the folio date.
- The parking charge of 96.00 has no receipt and is above the threshold, so it is flagged as needing one. The 6.75 coffee only needs a note with a purpose.
- The print shop receipt from September 4 is 67 days old on the report date, which is past the 60-day window.
- The 84 business miles on October 12 are priced at 0.76 because the trip falls after July 1.
The meal totals use the card amount, which includes the tip. If your policy reimburses tips, that is what you want. If not, switch the meal rows to receipt_amount.
Step 6: export for accounting and finance reporting
The report CSV is not yet a journal entry, but it is close. To post it, map each category to a general ledger account and each project to a cost center or dimension in your accounting system, then aggregate. Most accounting tools can import a CSV journal or expense batch; build the file in the column layout your system’s import documentation specifies, rather than inventing one.
Card lines and out-of-pocket lines book differently. Card-paid expenses clear the card liability account. Out-of-pocket lines and mileage create an amount owed to the employee, paid through payroll or AP. Keeping the card_txn column in the export lets finance tie each posted line back to the statement during the month-end card reconciliation.
Once every report lands as structured rows, finance-side reporting is a query rather than a project. Spend by category per month, travel cost per client project, the share of card lines with missing receipts per team, and the average days from purchase to report all come from the same table. The last two are worth watching: they tell you whether the capture process works before an auditor does.
Step 7: keep the report and its evidence together
Store the approved report with everything it was built from: the report CSV, the receipt images and PDFs, the extracted fields, the card statement file, and the approval record with who approved and when. The IRS accepts electronic records when the system meets the conditions in Rev. Proc. 97-22, which include an audit trail between the general ledger and the source documents. The card_txn and line ID columns are that trail.
For how long, the IRS page on how long to keep records gives three years after filing as the general period, with longer periods in specific situations and at least four years for employment tax records. Publication 463 adds that employees who hand their records to their employer and are reimbursed generally do not need to keep copies, so the employer’s archive is the one that has to hold up. Outside the US, retention periods are set nationally, so check the rule in each country where you operate.
Limitations
- The example assumes one currency. Multi-currency reports need the card’s settled amount and an FX field from the statement export.
- Matching by amount and date fails when an employee has several identical charges on the same day, such as repeated transit fares. Those need either the merchant tie-breaker or a person.
- Split charges, refunds, and partial credits break one-to-one matching. Treat negative card lines as their own category and match them by hand until you see how your card provider reports them.
- OCR gives you the date, total, and merchant. It does not give you category, project, purpose, currency, or VAT through the typed fields. Those come from the employee, the card file, or your own parsing of the
/ocrMarkdown. - The thresholds in the script (three days, 25%, $75, 60 days) are starting points. The $75 and 60-day figures come from Publication 463 for US accountable plans; your policy and your tax advisor decide what applies to you.
Conclusion
The slow part of a paper expense report was never adding up the totals. It was finding which slip belongs to which card line, chasing the receipts nobody kept, and pricing mileage at whatever rate someone remembered. Once receipts arrive as typed fields and the card provider sends a file, the report becomes a join with a short list of exceptions: tip differences, missing receipts above the threshold, late lines, and card lines that never got a receipt. Those exceptions are also the best measure of your process. If the same team produces most of the missing-receipt flags every month, the fix belongs in capture, not in the approval queue.
To try the receipt side, get a test key from get-key.json, run last month’s receipts through /ocr.json with receipt_date,total_amount,seller_name?, and feed the output and your card export into the script above. Product details and pricing are at ocrskill.com.
