Start the conversation

Why Saphirion

Because value should be calculated — not guessed.

Saphirion turns your product, performance and price data into transparent, reproducible formulas that calculate the value of your products — so decisions can be explained, defended and scaled.

Built on decades of price-data science — with a level of mathematical rigour no other pricing solution can match. Deployed on premises.

Your data. Your rules.

See the formula at work →

Mathematics superpower

Top six reasons to trust the NLPP formula, not the gut feel.

Built on decades of price-data science — with a level of mathematical rigour no other pricing solution can match.

01

Mathematical precision

Prices are derived from a reliable target price formula — not from experience, negotiation, or gut feeling.

02

Objective and transparent

Results are reproducible and explainable. Decisions rest on clear logic, not black boxes, opaque algorithms or encoded opinions.

03

Rapid comparability

Math is speed. Thousands of parts, variants and offers become simultaneously analysable and comparable — across the entire company.

04

Actionable opportunities

Instantly spot high-value opportunities, see how much to gain, understand how to implement, and be sure where action truly pays off.

05

Measurable impact

Improvements are tracked through to the P&L. The effect is not just visible — it is real and measurable.

06

Fast results

Instead of long transformation projects, short-term measurable results emerge — and a long-term structural competitive advantage.

The dangerous tool is not the one that fails — it is the one that creates the illusion of working.

Every tool returns a number

The question is whether that number deserves your trust.

Visibly wrong

A tool that fails loudly is cheap: you notice, you stop, you move on. An obvious failure costs a demo hour — nothing more.

Plausibly wrong

A tool that always returns a confident-looking number is expensive. Negotiations, savings targets and design decisions get built on results nobody can verify.

Provably sound

Sound mathematics is checkable. The right questions expose the difference between rigor and a convincing user interface — within one demo session.

The danger

Good-enough math fails invisibly.

Market price behavior vs. a good-enough model025050075010000.1 kg0.2 kg0.5 kg1 kg2 kg5 kg10 kg20 kgReal market pricesGood enough modelIllustrative data — the pattern is what matters, not the numbers.Market price behavior vs. a good-enough model

Too high in the middle

Across the middle of the range, the model over-prices — targets your suppliers accept instantly, and money you never see again.

Too low at the ends

At both ends it under-prices — targets the market will not confirm, and negotiations built on numbers that cannot hold.

Unwarned beyond the data

Past the last data point, the error keeps growing — a simplified model extrapolates without ever telling you.

The systematic error never announces itself — each individual number still looks plausible.

The R² illusion

Same R², four different realities.

I0510152005101520I
II02.557.51005101520II
III0510152005101520III
IV0510152005101520IV

I — the only sound case

A genuine linear relation — the one situation R² was designed for.

II — a curve, missed

The relation is non-linear. The fitted line — and its R² — miss it entirely.

III — one outlier steers

A single odd part tilts the whole line. The R² stays just as flattering.

IV — no relation at all

Identical parts plus one exotic point invent a trend that does not exist.

0.67 the same R², the same regression line — for all four. R² cannot tell them apart.

Anscombe's quartet (1973) — the textbook proof that identical statistics can hide entirely different realities.

The R² trap

R² rewards exactly the wrong behavior.

More price drivers: R² rises, real accuracy falls0510152000.250.50.7512345678910Pricing error on new parts, %R² on the fitting dataIllustrative — statistical overfitting: the model memorises noise instead of learning the relation.More price drivers: R² rises, real accuracy falls

More drivers, higher R² — always

R² can only rise as variables are added, relevant or not. It rewards fitting noise and cannot warn you when it happens.

The illusion of precision

An R² of 0.98 prices the parts the model was fitted on. The next RFQ — where money is actually decided — is priced by the red bars.

What NLPP does instead

It finds the best model the data supports, states the safe maximum of drivers, and guarantees valid results — no R² required.

R² is also only defined for linear models — on the non-linear relations found in most real data sets it is not merely weak, it is undefined.

The evaluation guide

Fifteen questions any vendor must survive.

We wrote the exam for our own discipline: the Price & Cost Analysis Evaluation Guide — consisting of fifteen questions. Contact us to get it.

Validity

Results that cannot be wrong

A sound method guarantees valid target prices by construction — in all circumstances, not on average. It never needs to hide, filter or excuse its own output.

Objectivity

The software decides, not opinion

Wherever a user must guess, tune or judge, results turn into opinion. Sound software removes those decisions: two analysts, same data, same result.

Transparency

No black-box pricing

A target price you cannot explain is an opinion with decimals. Sound software shows its work — impact per driver, extrapolation warnings, formula included.

Put us to the test.

Start the conversation and request the full Evaluation Guide. Then bring your own parts data and ask all fifteen questions, live — of us, and of anyone else. We wrote the guide, and we hold ourselves to it.

  • 30-min intro call
  • Full guide on request
  • On your data
?
required ?
?
?