Exception handling in e-commerce AP: what actually causes delays (and how long fixes really take)

If you run AP for an e-commerce brand (or you’re the bookkeeper cleaning it up), you already know the feeling.

An invoice arrives. It looks normal for about 1.7 seconds. Then you notice:

• the totals don’t match the PO

• the freight lines have mixed GST

• the supplier billed the subscription in USD again

• the invoice is a duplicate… maybe

• someone split-shipped the order so the quantities are “sort of” correct

And suddenly that invoice isn’t an invoice anymore. It’s a small investigation.

This is what exception handling actually is: turning “something’s off” into a clean decision you can post, approve, and pay.

The part most people miss: delays aren’t caused by the exception. They’re caused by the chase.

In e-commerce, the exception itself is often simple. What’s hard is:

1. figuring out who can resolve it (AP? buyer? warehouse? 3PL? founder?)

2. getting an answer back fast enough to keep the invoice moving

Every extra handoff adds time. Every unclear rule adds more time.

That’s why manual invoice processing timelines can stretch into weeks, while automated teams tend to process much faster on average. Benchmarks vary, but it’s common to see manual workflows taking roughly 1 to 3 weeks, while automation pushes it down to days.

Now let’s get specific.

The e-commerce AP exception list (with real-world fix times)

Below are the exceptions that cause the most delays in e-commerce AP, plus the actual root cause and the usual time-to-fix range.

Important note: “time to fix” isn’t the time to notice it. It’s the time from “flagged” to “resolved and ready to post/pay”. The same exception can be 10 minutes or 10 days depending on who you need to respond.

1) Missing PO (or “wrong PO”, or “PO doesn’t exist”)

What causes it in e-commerce

• Purchasing happened via email/phone/Slack because stock was urgent

• Shopify ops bought something “quickly” on a card

• PO exists but supplier didn’t print it on the invoice

• Multiple POs got consolidated into one invoice (classic)

Why it delays

• AP can’t confidently code it

• Approver won’t approve without a PO reference

• You need the buyer to confirm what this relates to

Typical time to fix

• Fast: 10 to 30 minutes if it’s an internal lookup and the buyer replies quickly

• Normal: 1 to 2 business days (waiting for the right person)

• Painful: 3 to 7 business days (supplier reissues invoice or buyer is MIA)

Missing or incorrect PO details are one of the most common triggers for invoice exceptions broadly.

2) Price mismatch (invoice price ≠ PO price ≠ agreed terms)

What causes it

• Old price lists (supplier updated, your PO didn’t)

• “Freight added later” as a separate line that wasn’t agreed

• Promo/volume discounts not applied

• FX conversions done inconsistently

Why it delays

• Someone must confirm the right price

• Often requires a credit note or invoice reissue

• Approvers hate approving “wrong totals”

Typical time to fix

• Fast: same day (if it’s a known price change and buyer confirms)

• Normal: 2 to 5 business days (supplier issues credit note)

• Long: 1 to 3+ weeks if it becomes a dispute

Invoice discrepancies like incorrect pricing are one of the standard categories that block straight-through processing.

3) Quantity mismatch (invoice qty ≠ received qty)

What causes it

• Partial shipments (common with stock-outs)

• 3PL receiving lag (received physically, not updated in system)

• Backorders and substitutions

• “Supplier invoiced full order upfront” but you only received part

Why it delays

• You need warehouse/3PL confirmation

• Sometimes you need to wait for the second shipment

• If your process needs 2-way or 3-way matching, this blocks approval

Typical time to fix

• Fast: same day if receiving data is clean

• Normal: 2 to 4 business days if you’re waiting on receiving confirmation

• Long: 1 to 3 weeks if the mismatch only resolves when the next shipment arrives

4) Duplicate invoice (or “is this a duplicate?”)

What causes it

• Supplier resends invoice because they didn’t get a reply

• Same invoice comes via PDF email + portal download

• Invoice number changes slightly (INV-103 vs INV103)

• Credit note arrives and someone mistakes it for a fresh invoice

Why it delays

• Paying duplicates is a nightmare, so teams get cautious

• Manual systems can’t reliably detect near-duplicates

• You end up searching threads, inboxes, and Xero/MYOB history

Typical time to fix

• Fast: 5 to 15 minutes if your system flags duplicates properly

• Normal: 30 to 90 minutes when it’s a “looks similar” case

• Long: 1 to 3 business days if you must confirm with supplier

5) Tax or GST coding issues (especially mixed-tax invoices)

What causes it

• Freight invoices: customs, duties, fees, fuel levy, tolls, admin fees (mixed treatment)

• International suppliers (GST-free imports vs taxable local charges)

• Supplier uses the wrong tax code

• Multi-state or marketplace fees with confusing tax breakdowns

Why it delays

• The reviewer has to interpret tax treatment line-by-line

• If it posts wrong, BAS clean-up later is worse

• Often requires a senior bookkeeper review, not junior AP

Typical time to fix

• Fast: 10 to 25 minutes if you’ve seen that supplier’s format before

• Normal: 30 to 120 minutes if it needs research and splitting

• Long: 2 to 5 business days if you need supplier clarification or revised invoice

This is one of those areas where “invoice automation” is supposed to help, but only if the tool can handle line-level detail and mixed tax lines, not just header capture.

6) Multi-account allocations (aka “just split it later”)

What causes it

• One supplier invoice spans multiple cost centres (ads + packaging + freight)

• Bundled services (retainer + reimbursables + software)

• Amazon, Shopify, PayPal-type invoices that touch multiple GLs

• Brand has multiple entities (AU entity, NZ entity, holding entity) and invoices get sent to the wrong one

Why it delays

• Someone needs context to split correctly

• People procrastinate because it’s fiddly

• It creates a second round of review later (month-end surprise)

Typical time to fix

• Fast: 10 to 20 minutes if the rules are already defined

• Normal: 30 to 90 minutes if it’s manual splitting and checking

• Long: Days, not because it takes days to do, but because it gets parked until month-end

This is where bookkeeping AI and accounting AI can be genuinely useful, but only if it learns the split pattern and flags low-confidence lines instead of forcing humans to re-check everything.

7) Approval delays (invoice is fine, humans are slow)

What causes it

• Approver gets it in email and forgets

• Approval is ambiguous (“is this marketing or ops?”)

• Founders are travelling, buyers are on the floor, everyone’s busy

• Too many invoices routed to the same person

Why it delays

• Waiting time dominates work time

• The actual approval click is 10 seconds… two days later

Typical time to fix

• Fast: under 4 hours (same-day approvals)

• Normal: 1 to 3 business days

• Long: 1 to 2 weeks (usually when there’s no escalation path)

This is also why AP cycle-time benchmarks improve so sharply when approvals are automated and tracked properly.

8) Vendor statement mismatches and missing credits

What causes it

• Credit note issued but not applied

• Returns/refunds not reconciled back to AP

• Supplier statement shows overdue invoices you never received (or vice versa)

Why it delays

• Requires statement reconciliation

• Often involves searching old emails and portal downloads

• It’s multi-step: confirm, request documents, apply credits, re-age

Typical time to fix

• Fast: 30 to 60 minutes if you have the docs already

• Normal: 3 to 7 business days (supplier back-and-forth)

• Long: 2 to 4 weeks if it spans multiple months and multiple credits

9) Data capture issues (the invoice is readable to a human, not to a system)

What causes it

• Weird formats (scanned, skewed, faint)

• Missing invoice number/date

• Supplier changes layout suddenly

• Line items aren’t structured (everything in one description blob)

Why it delays

• Someone must manually correct fields

• If you don’t correct it, everything downstream breaks (matching, coding, approvals)

Typical time to fix

• Fast: 2 to 10 minutes for light corrections

• Normal: 10 to 25 minutes for messy docs

• Long: 1 to 2 business days if you must request a cleaner copy

A simple “ageing ladder” for exceptions (so you can predict the blowouts)

Here’s a blunt but useful rule:

• Minutes: anything you can solve without leaving your system (duplicate check, known coding rules)

• Hours: anything that needs internal confirmation from someone who’s responsive

• Days: anything that needs supplier action (credit note, reissue, statement docs)

• Weeks: anything that mixes dispute + receiving + terms ambiguity

That’s it. That’s the ladder.

So when someone asks, “Why is AP slow?” the answer usually isn’t “because the team is slow.”

It’s because too many invoices sit in the days and weeks buckets.

Why e-commerce has a higher exception load (and why it feels constant)

E-commerce AP is exception-heavy because your suppliers and charges are inherently messy:

• freight and landed cost allocations

• ad platforms and marketplaces

• refunds/returns creating credit chaos

• fast purchasing when stock is at risk

• lots of subscription tools billed in odd cycles

Industry reports often show exception rates staying stubbornly high (think: a meaningful chunk of invoices flagged), and automation tends to lower it materially.

Also, some invoices genuinely require rework. One cited industry stat puts rework at around 12.5% of invoices on average.

So no, you’re not imagining it.

Where invoice automation software actually cuts delay (without pretending humans disappear)

Good invoice processing automation doesn’t “eliminate exceptions”.

It does something more valuable: it turns exceptions into smaller, faster decisions.

What changes when it’s done well

• Fewer “false exceptions” (system catches missing fields, duplicates, basic checks early)

• Cleaner routing (the invoice goes to the one person who can answer, not five people who can’t)

• Confidence-based review (90% can go through quickly, 10% gets human attention)

• Clear audit trail (no more “who said this was okay?”)

That’s the real promise of bookkeeping AI and accounting AI in AP: not replacing judgement, but reducing the amount of judgement you need per invoice.

And yes, tools like Dext and Hubdoc help with capture and organisation, especially for simpler invoices. But e-commerce exceptions often live at the line-item level (freight, mixed GST, allocations, PO matching). That’s where workflow-native AP tools (like what you’re building with Pulsify) tend to win: fewer re-touches, fewer “fix it later” invoices, faster approvals because the bill is already coded properly.

The punchline: most delays come from three root causes

If you want a neat summary, it’s this:

1. Missing context (no PO, unclear allocation, unknown terms)

2. Waiting on humans (approvals, receiving confirmation, buyer sign-off)

3. Waiting on suppliers (credit notes, reissues, statement docs)

If your workflow reduces those three, exception handling stops feeling like a daily fire drill.