Skip to content

Glossary

Match score in AI product matching: what it means on a quote line

A match score in AI product matching is a number that shows how well the product the software proposed fits the line the customer wrote. Some tools call it a confidence score. In ERP Agent every quote or order line carries a match score from 0 to 100, and lines under 80 are marked, so the rep checks those first.

Definition

Match score (confidence score)

A number on each line of an automatically prepared quote or sales order showing how well the proposed product fits the customer’s request. It tells the reviewer where to look. It does not approve the line.

What does a match score mean in wholesale distribution?

Enquiries arrive in the customer’s words, not your item codes: “M10 nyloc A2”, a competitor’s part number, a customer part number, a size with no unit. Some lines leave no doubt. Others leave a choice open: which length, which material, angled or straight.

The match score tells the two apart. Without one, a rep checks every line and saves little, or trusts every line until a wrong product reaches a customer. With one, the rep reads the doubtful lines closely and skims the rest. The score drives the review step in AI quote automation for distributors and order entry automation.

What moves a match score up or down?

The evidence that usually moves it:

What moves a match score up or down?
Evidence on the lineTypical effectExample
Your product code, or a customer or competitor code with a known cross-referenceHigh“4471205”, “our ref 22-0187”
Every stated attribute agrees with one productHigh“M10 x 40 hex set screw A2”
An attribute is missing and several products fitLower“M8 washers”, no material
A stated attribute contradicts the candidateMuch lower“24 V” against a 230 V product
Vague wording, or an item outside your rangeLow, or no match“fixings as last time”

ERP Agent pushes down candidates that contradict a stated size, voltage or rating class. A line with no plausible match goes on the organisation’s placeholder code rather than a forced guess.

Example: match scores on a fixings enquiry

A fastener distributor gets this enquiry from a contractor’s buyer. Prices are left out.

Customer requestIllustrative example
Email from a contractor’s buyer
  1. 200 x M10 x 40 hex set screw A2
  2. 200 x M10 nyloc A2
  3. 50 x our ref 22-0187
  4. 500 x M8 washers
  5. 10 x M12 anchors
Proposed lines · Illustrative example
Customer wroteProposed productMatchRep
M10 x 40 hex set screw A2Hex set screw DIN 933 M10 × 40, A297Approves
M10 nyloc A2Nylon insert lock nut DIN 985 M10, A293Approves
our ref 22-0187Socket cap screw ISO 4762 M6 × 20, A291Approves: same product as last time
M8 washersFlat washer DIN 125 M8, zinc plated71Marked: takes the A2 alternative
M12 anchorsThrough bolt anchor M12 × 110, zinc plated48Marked: asks for type and length

Illustrative example with invented data and scores. In ERP Agent, lines under 80 are marked for review.

ERP Agent quote review interface with sample data
Sample data: match scores beside the customer’s original wording. Open full size

How do you read a confidence score on a quote line?

  • Start with the marked lines, under 80 in ERP Agent. Open one to read the agent’s reasoning, then take an alternative or change the line in quote chat.
  • Read it as confidence in the product. Quantity, unit and price still need a glance, especially on lines sold by the box, by length or by weight.
  • The score guides the review; your team approves. Your team stays in control: by default a person reviews before anything reaches your ERP, CRM or other system.
  • Do not compare scores across tools. Each vendor calculates its own.

Is a match score the same as accuracy?

No. A match score is an estimate for one line before anyone checks it. Accuracy is a result: the share of lines that turned out right, measured afterwards over many lines. At LVI-WaBeK, ERP Agent automates quote lines with over 95% accuracy.

In a pilot, compare the agent’s lines with what the rep sent, and count the lines that scored high and were still wrong. The AI quote agent pilot checklist shows how.

Frequently asked questions

What is a confidence score on a quote line?

A confidence score is another name for a match score: how sure the software is that the proposed product is the one the customer meant. In ERP Agent it runs from 0 to 100.

What is a good match score?

In ERP Agent, lines under 80 are marked for review. What matters more is calibration: lines at 90 should be right far more often than lines at 60. A tool that shows the same score on every line tells the rep nothing.

Can a high-scoring line go out unchecked?

No. ERP Agent sends nothing to customers on its own: a person approves the draft, and the score only decides where the rep looks first.

Does fixing a low-scoring line help next time?

Yes. The agent learns from the lines your team sends: the next time a customer writes the same thing, it picks the same product.

Let’s look at your quote workflow

Bring an enquiry your team found awkward

We’ll walk through how the agent would score it on the review screen, reasoning and alternatives included.

  • Your requests and product catalogue
  • Your pricing and review process
  • A clear scope for the next step
Lauri Pelkonen

Lauri Pelkonen

Founder & CEO

Book a 30-minute call Request a demo by email

Or use the address directly