Why Your AP Team Misses 80% of Overcharges (And How to Catch Them All)

A VP of Procurement at a 300-person manufacturer told us: "We know we are overpaying somewhere. We just do not have enough people to check every line on every invoice against every contract." That gap between knowing and checking is where overcharges live.

The math is straightforward. A manufacturer processing 2,000 invoices per month with an average of 12 line items each generates 24,000 individual price checks per month. Even if your AP team spends 15 seconds per line item, that is 100 hours of manual review. Nobody has that time, so teams spot-check the big invoices and approve the rest.

That is exactly where pricing errors hide. Not in the $80,000 steel order your director reviews personally, but in the $1,200 fastener invoice where the price per box went from $14.50 to $15.80 without anyone noticing. Multiply that $1.30 difference across 200 boxes per month, and you are overpaying $3,120 per year on a single line item from a single vendor.

The Problem with Rules-Based Detection

Most AP automation tools offer some form of anomaly detection. The typical approach: your team writes rules. "Flag any invoice where the unit price exceeds the PO price by more than 5%." "Alert if total spend with a vendor exceeds $50,000 in a month."

This approach has three fatal problems:

1. Thresholds go stale immediately. You set a 5% tolerance for steel coil pricing in January. By March, steel prices have moved 8% on the commodity index. Now every legitimate invoice triggers an alert, your team starts ignoring them, and the one actual overcharge in April sails through.

2. You cannot write rules for things you have not seen. A vendor starts charging a $75 "fuel surcharge" on deliveries that never had one before. There is no rule for that because it never happened before. By the time someone notices and writes a rule, you have paid the surcharge on 40 shipments.

3. Maintenance is a full-time job. A manufacturer with 200 active vendors and 3,000 SKUs would need thousands of individual rules to cover every vendor-item-price combination. Every contract renewal, every new vendor, every product substitution requires rule updates. The team that was supposed to save time on AP is now spending it on rule maintenance.

How Kynthar Learns What Normal Looks Like

Kynthar takes a fundamentally different approach. Instead of asking your team to define what looks wrong, the system learns what looks normal for each vendor and each item, then flags anything that does not fit.

Here is how it works in practice.

Your company buys hydraulic fittings from Pacific Industrial Supply. Over the last 8 months, you have purchased part number HF-3042 (3/4" brass hydraulic coupling) at these prices:

• January: $28.40 per unit (200 units)
• February: $28.40 per unit (150 units)
• April: $28.75 per unit (200 units)
• May: $28.40 per unit (300 units)
• July: $29.00 per unit (250 units)

Kynthar automatically calculates that your average price for this part from this vendor is $28.59, and the normal variation is about $0.25 in either direction. The system now has a baseline.

In September, an invoice arrives from Pacific Industrial for HF-3042 at $31.20 per unit. That is $2.61 above average, roughly 10 times the normal variation. Kynthar immediately flags this line item:

No rule created that flag. No threshold defined by your team. The system learned the pattern from your own purchasing history and surfaced the deviation with specific dollar amounts your team can act on.

Why Smart Thresholds Prevent False Alarms

The single biggest reason rules-based detection fails is false alarm fatigue. Your AP team gets 50 alerts per day, 45 of them are legitimate price changes, and they stop reading the alerts. The 5 real overcharges get approved along with everything else.

Kynthar prevents this with adaptive thresholds. The core idea: the system requires stronger evidence before raising an alert when it has limited information.

With only 1-2 prior purchases from a vendor: Kynthar uses a very wide tolerance. If you have bought gaskets from a new vendor twice at $4.20 and $4.35, and the third invoice says $4.80, that might just be normal variation. The system notes the increase but does not flag it as an anomaly. It needs more data.

With 3-10 purchases: The tolerance narrows. Kynthar now has enough data to distinguish normal variation from genuine deviations. A 15% price jump on a part with consistent pricing will generate a flag.

With 10+ purchases: The system operates at full sensitivity. It knows that this vendor charges between $14.20 and $14.60 for this part, so $15.90 is clearly outside the pattern.

Cross-Vendor Price Intelligence

Single-vendor price checks catch overcharges against historical patterns. Cross-vendor comparison catches overpayment against the market.

Here is a real scenario. Your company buys FR-4 circuit boards from three different vendors:

Three different vendors, three different part numbers, all for the same component: a standard 2-layer FR-4 PCB, 1.6mm thickness. Without item normalization, these look like three unrelated purchases. With it, you can see that Midwest PCB Solutions is charging 24% more than your best available price.

If you are buying 5,000 boards per month from Midwest, that is $4,000/month in potential savings by shifting volume, or at minimum, leverage for a price renegotiation.

Kynthar handles the normalization automatically. When "PCB-FR4-2L-1.6" and "2LAYER-FR4-062" and "FR4-DBL-STD" all refer to the same item, the system connects them across vendors and surfaces the price gap with specific per-unit and per-month dollar amounts.

Five Overcharge Patterns Kynthar Catches Automatically

1. Price Creep

Your packaging vendor charged $0.42/unit for corrugated boxes in January. By October, the price has crept to $0.49/unit through five small increases. No single invoice looked wrong. But across 100,000 units per month, you are now paying $7,000/month more than you were in January, $84,000 annualized.

Kynthar surfaces this as a trend, not just a point-in-time anomaly: "Corrugated box unit price from Allied Packaging has increased 16.7% over 9 months across 5 invoices. Annualized impact at current volume: $84,000."

2. Contract Price Violations

You negotiated a 12-month contract with your MRO supplier at $8.60 per box for nitrile gloves. Six months in, an invoice arrives at $9.15. The vendor updated their price list and your AP team does not have time to check every line item against the contract.

Kynthar cross-references the invoice price against the contract rate: "Nitrile gloves (SKU NG-100L) invoiced at $9.15/box. Contract price: $8.60/box. Overcharge of $0.55/box across 400 boxes = $220.00 on this invoice."

3. Duplicate Invoices

A vendor sends Invoice #8847 for $12,400 in March. It gets paid. In June, the same vendor sends Invoice #8847 again, this time as a PDF with a slightly different format. Your AP system processes it as a new invoice because the file is different.

Kynthar catches duplicates three ways: exact invoice number matches, same-amount invoices from the same vendor within a time window, and near-matches where invoice numbers differ by a single character (catching typos and resubmissions with altered numbers).

4. Cumulative Over-Billing

You issue a PO for $50,000 worth of electronic components. The vendor sends five partial invoices: $12,000, $11,500, $13,000, $9,800, and $8,200. Each individual invoice looks reasonable. But the total is $54,500, which is $4,500 over the PO amount.

Kynthar tracks cumulative invoiced amounts against every PO and flags when the total exceeds the authorized amount: "Total invoiced against PO-2024-0892: $54,500. PO authorized amount: $50,000. Over-billed by $4,500 (9%)."

5. New Vendor Risk Signals

A new vendor submits their first invoice for $38,000. With no history, statistical baselines do not apply. Instead, Kynthar applies graduated scrutiny: first-time vendors with invoices above a dollar threshold receive extra visibility, not because the invoice is wrong, but because there is no established pattern to validate it against.

What Detection Looks Like in Practice

Here is what your AP team actually sees when Kynthar processes an invoice.

An invoice arrives from Consolidated Fastener Corp for $18,430. Within milliseconds, Kynthar has:

• Compared every line item price against historical purchases from this vendor
• Checked every line item against prices from other vendors for the same parts
• Verified the invoice total against the PO authorized amount, including all prior invoices against this PO
• Checked for duplicate invoice numbers and similar-amount invoices from this vendor in the last 90 days
• Compared line item prices against any active contract rates

Your team does not see a spreadsheet of numbers. They see specific, actionable findings:

Each finding includes the specific dollar impact so your team can prioritize which ones to act on. A $130 overcharge on fasteners might not be worth a phone call. A $1,440 annual gap on washers across vendors absolutely is.

Why Speed Matters: Catching Overcharges Before Payment

Traditional detection tools run batch reports. You upload invoices, wait for overnight processing, and review exceptions the next morning. By then, half the invoices have already been paid because your AP team could not wait 24 hours for every batch.

Kynthar runs all detection checks the moment an invoice enters the system. Detection happens in milliseconds. Your team sees anomalies before they approve payment, not after.

This distinction matters enormously. Disputing an invoice before payment is a quick conversation: "Your invoice says $31.20 for HF-3042 but our contract rate is $28.75. Please reissue." Recovering an overpayment after the check has cleared requires credit memos, accounting adjustments, vendor negotiations, and 10-20 times the effort.

Getting Started: No Rules Required

The hardest part of every rules-based system is the setup. Somebody has to define thresholds, map vendor contracts, configure alert routing, and maintain all of it as your vendor base changes.

Kynthar requires none of that. Forward your procurement emails to your Kynthar address, and the system starts building intelligence immediately. Every invoice that flows through adds to the pricing baseline. Every new vendor creates a new profile. Every contract becomes a reference point for future invoice validation.

The system compounds. In month one, it catches obvious anomalies like duplicate invoices and large price jumps. By month three, it has enough history to catch subtle price creep. By month six, it knows your vendor pricing patterns better than anyone on your team because it has cross-referenced every line item on every invoice from every vendor, something no human has time to do.

• Day 1: Duplicate detection and cross-vendor price comparison active immediately
• Week 2: Enough data to flag major price deviations for high-volume items
• Month 3: Full statistical baselines for your top vendors and items, subtle anomalies now detectable
• Month 6+: Deep vendor intelligence, trend analysis, and contract compliance checking across your entire vendor base