Should-Cost & Historical Price Benchmark Helper
Check every incoming quote against what you've actually paid before (and any benchmark data you have) so you can tell at a glance whether a price is high, fair, or suspiciously low - before you award it.
A private internal tool where a buyer loads a quote, sees it matched to your real price history with a traffic-light verdict and a suggested target price, then records an accept / negotiate / reject decision into an audit-logged decision log - exportable as CSV.
Before you start
- A free Supabase account
- A free Vercel account
- A free Resend account (for email alerts)
- A price-history export (CSV or Google Sheet): item/SKU, date, supplier, paid price, quantity, unit
- Optional: benchmark or index figures for some items
The problem this kills
A supplier sends a quote. The item is $4.20 a unit. Is that good? Bad? A rip-off? Most buyers have no idea in the moment - so they accept it, or they push back on the wrong line and let the real overcharge slide through. The reference point you need is sitting in last year's POs and a dozen spreadsheets nobody opens during a negotiation.
So quotes get awarded on gut feel and supplier charm. Prices creep up 3% a year and nobody notices until the contract is signed. The buyer who'd love to say "actually, we paid $3.80 for this six months ago - let's talk" simply doesn't have the number in front of them.
This tool puts that number in front of them. Every time.
What you'll build
A private web app for your buying team. A buyer pastes or uploads an incoming quote (item, supplier, unit price, quantity). The tool matches that line to your historical paid prices for the same or similar item, normalizes the units and currency, weights recent and same-supplier history, and returns a clear verdict:
- Green - in line with what you normally pay.
- Yellow - higher than your reference; here's the gap and a suggested target price.
- Red (high) - well above your benchmark; negotiate hard.
- Red (low) - suspiciously below history; double-check spec, MOQ, and quality before celebrating.
It also flags price creep - when an item has been quietly trending up over time. The buyer reads the verdict, then decides: accept, negotiate, or reject - and that decision is recorded with a reason. The tool advises; the human decides.
What's inside the Implementation Plan
- It starts by interviewing you about your business. This is not a generic template. The very first thing the plan does is ask the AI agent to interview you - your RFQ process, the systems and spreadsheets you live in, how your SKUs and item codes are actually named, your currencies and units, your typical and peak quote volumes, and your real approval rules. It reflects a short tailored spec back to you for a thumbs-up, then builds the tool around your data - not a made-up example.
- A step-by-step build, each step ending in a copy-paste prompt for your AI coding agent.
- The matching and scoring logic: last paid, average paid, best paid, recency-weighting, same-supplier weighting, and price-creep detection.
- Unit and currency normalization so a quote in "per case, EUR" can be compared to history in "per each, USD".
- The decision log, the audit trail, and a CSV export in the exact columns you specify.
- A "No API yet?" fallback so you can build the whole thing today from a Google Sheet or CSV - no integration with your ERP required.
The governance it includes (this is the point)
This is an internal tool that touches sourcing decisions, so it's built to be trustworthy from day one:
- Login so only your team can open it.
- Row-level security so each organization only ever sees its own quotes and price history.
- A complete audit trail - who benchmarked what quote, what verdict it gave, who decided, and when.
- A hard human-in-the-loop gate - the tool never "awards" anything. It produces a verdict and a suggested target; a buyer reviews and records the real decision. Nothing is committed to your records without a person.
- Duplicate guards - the same quote line (SKU + supplier + quote date) can't be benchmarked and logged twice.
Who it's for
Buyers, category managers, and sourcing leads who accept quotes with no reference point for "is this a fair price?" If you've ever signed off on a quote and quietly wondered whether you just overpaid, this is for you. No developer needed - you'll build it by talking to an AI.
You've got this. Open the plan, paste the first prompt, and let the interview tailor the tool to your business.