Use case

Quote a machined part in 60 seconds —
and get more accurate with every job you ship.

Werk24 extracts every dimension, tolerance, and feature from your 2D drawings. Saphirion — a Swiss analytics firm specializing in manufacturing intelligence — turns your historical pricing into an auditable formula. Together they replace estimator gut‑feel with measured cost.

Laserhub: 1,000+ parts/month quoted, 98% accuracy on price‑driving features.Read the case study →

Why manual quoting costs you margin

Manual reading of 2D drawings and gut‑feel costing make quotes slow and inconsistent. Estimators must predict cycle time, setups, scrap, and risk from memory. Under pressure or with mixed experience levels, you see quote variability, margin slippage, and long cycle times.

Is this for you?

Best fit

  • Manufacturers producing a small set of part families (e.g., turned shafts, milled plates, common subassemblies).
  • Recurring RFQs where features repeat and learning compounds.
  • Desire to codify tribal knowledge and standardize quoting.

Not ideal

  • Every part looks different (pure job‑shop variety with no clustering).
  • No access to actual production costs, or unwillingness to measure them.
Illustration of recurring part families versus one‑offs
Part-family focusGrid highlighting recurring families A and B among mixed parts.Best fitNot idealFamily AFamily BOne-offs

What the extraction looks like

Werk24 doesn't just OCR text — it understands engineering drawings. Here is a real excerpt of the structured JSON returned for a single part. Every field includes units and a confidence score.

{
  "titleBlock": {
    "partNumber": "W24-2847-REV-C",
    "material": { "standard": "EN 10088-3", "grade": "1.4404 (316L)", "confidence": 0.97 },
    "generalTolerance": "ISO 2768-mK"
  },
  "dimensions": [
    { "type": "diameter", "nominal": 18.0, "tolPlus": 0.018, "tolMinus": 0.0, "unit": "mm", "feature": "H1", "confidence": 0.99 },
    { "type": "linear", "nominal": 45.0, "tolPlus": 0.05, "tolMinus": -0.05, "unit": "mm", "confidence": 0.98 }
  ],
  "gdt": [
    { "characteristic": "position", "value": 0.1, "unit": "mm", "datumRefs": ["A","B"], "confidence": 0.96 },
    { "characteristic": "perpendicularity", "value": 0.02, "unit": "mm", "datumRefs": ["B"], "confidence": 0.95 }
  ],
  "threads": [
    { "designation": "M12x1.75-6H", "type": "metric", "depth": 20.0, "unit": "mm", "confidence": 0.98 }
  ],
  "surfaceFinish": [
    { "parameter": "Ra", "value": 1.6, "unit": "µm", "area": "bore H1", "confidence": 0.94 }
  ]
}

This is what sets Werk24 apart from generic OCR — structured, typed, confidence‑scored data ready for downstream pricing, feasibility, and ERP.

Your formula reflects how you build parts

Two shops can manufacture the same part with different setups, machines, and QA routines. A generic cost model ignores those differences. Your formula — or Saphirion's model — weights the drivers that matter in your environment (setup vs. cycle time vs. tooling vs. QA), so price reflects how you actually build parts.

In most shops today, quoting is still based on estimates and intuition. Actual costs are only reviewed much later by cost control, meaning valuable insights arrive too late to improve the next quote.

A closed loop changes that. Every finished job delivers hard data — machine time, tooling, QA, even scrap — which flows directly back into your pricing model. The system self‑calibrates continuously, so each new part makes the next quote more accurate.

  • Measure actual production cost at job close (time tickets, machine data, tooling, QA, scrap).
  • Feed back automatically into the next quote for similar parts — your model self‑calibrates.
  • Reduce reliance on guesswork: the system learns from real costs instead of only expert intuition.
  • Continuously improve pricing and operations: the same signals that tune price reveal process bottlenecks.
Closed‑loop pricing: quote → production → measured cost → learning
Today vs. Tomorrow — Closing the LoopTop: The estimator predicts costs and cost control provides annual feedback. Bottom: Each newly manufactured part updates pricing automatically.TodayEstimatingProductionRecord actual costsCost controlPredictRecorded after productionAnnual feedbackTomorrowPricing modelProductionActual cost (per job)Auto-learn & updateEach new partupdates immediatelyAutomatic / ContinuousManual / Delayed feedback

How it works

Three phases. You provide the data, we build the formula, then production keeps it honest.

  1. What you give us You

    ~100 representative drawings (PDF or scan) and matching historical prices or actual production costs. Use stable, comparable SKUs — avoid bundled or one‑off specials.

  2. What we do Werk24 + Saphirion

    Extract every dimension, tolerance, GD&T frame, material, and finish from your drawings → Fit an interpretable pricing formula with error bands and driver importance → Flag outliers, missing drivers, and extraction anomalies → Iterate with your cost engineers until KPIs are met.

  3. After go‑live

    Feed actual production costs after each job close. The model self‑calibrates and improves every subsequent quote — automatically, through your existing tools.

Outcomes

Fewer gut‑feel guesses, tighter margins, full traceability.

  • Automated quoting for standard parts

    Generate quotes in seconds for parts that match your trained families. Estimators focus on exceptions and complex geometry — not routine work.

  • Self‑calibrating prices

    Every shipped job feeds measured cost back into the model. The quote‑to‑actual gap shrinks automatically over time.

  • Traceable rationale for every price

    Every quote is explainable: driver weights, confidence scores, and a full audit trail. No more "the senior estimator just knows."

Metrics that matter

Track the signals that show quoting is not only faster, but also more predictable, competitive, and profitable.

  • Speed: Time‑to‑quote (P50/P90) to prove cycle‑time improvements.
  • Competitiveness: Win rate and price realization vs. target margins.
  • Discipline: Gross‑margin variance across estimators, sites, or teams.
  • Accuracy: Quote‑to‑actual cost delta — should steadily trend toward zero.
  • Model health: Forecast accuracy (MAPE) and human override rate.

FAQ

Is this suitable if every part is unique?
It works best when you repeatedly quote a few part families. If every part is a one‑off with no clustering, the model cannot learn effectively. We’ll help you assess family structure in your RFQ portfolio.
What closes the loop?
Instrument your production process to capture actual costs per job. That data flows back to pricing for the next similar part. Over time, the quote‑to‑actual gap shrinks and confidence rises.
What is Werkflow?
Werkflow is a free online demo where you can upload your own drawings and see Werk24's structured extraction output in real time — no API key or signup required.
Where does the data live?
Within your tenant. Data is encrypted in transit and at rest. Access is role‑based and auditable.