Evidence-first website QA for industrial teams

Catch conflicting product specs before buyers do.

SpecDrift finds hard-fact conflicts for the same model across product pages, PDFs, and page sections—then gives your team the exact source evidence to resolve them.

One relevant finding first. No full audit, spam, or payment.

  • Public pages only
  • No integration
  • Human-reviewed findings
PUBLIC EVIDENCE EXAMPLEFinding SD-0042
High confidence
MODELEM2RS-A882Stepper drive
!
CONFLICT DETECTED

Supply voltage & output current

01
PRODUCT PAGEHeader summary
20–80 VDC · 7.0 A
02
SAME PAGETechnical Data
18–80 VAC / 30–110 VDC · 8.2 A
03
OFFICIAL PDFUser manual
20–80 VAC / 30–100 VDC · 8.2 A

We flag contradictions. Your product team confirms the correct value.

BUILT FOR SPEC-DENSE CATALOGS
  • HTML
  • TEXT PDFs
  • MODEL-AWARE
  • UNIT NORMALIZATION
  • SOURCE EVIDENCE

01 / THE PROBLEM

One product.
Three public truths.

Industrial catalogs are maintained across templates, PDFs, languages, and years. A single copied value can quietly split into several versions while every page still looks “finished.”

A

Product page

Marketing summaries are fast to update—and easy to detach from technical data.

60 MPa
B
PDF

Datasheet

Old files stay indexed and downloadable long after a new revision is published.

100 MPa
C
EN

Localized page

Model names, units, and negative signs can drift during manual localization.

Unclear
THE RESULT

Wrong-fit inquiries, avoidable engineering back-and-forth, and product data buyers cannot confidently quote.

02 / THE WORKFLOW

From domain to defensible evidence.

No PIM migration. No CMS plugin. No generic score out of 100.

  1. 01
    yourfactory.com

    Crawl public assets

    Discover product pages, downloadable PDFs, and linked manuals from your domain.

  2. 02
    HTMLSKUPDF

    Group by model

    Match assets only when model identity is clear. Ambiguous families are not forced together.

  3. 03
    1000 kPa=1 MPa

    Normalize & compare

    Compare hard facts after normalizing units, ranges, symbols, and common engineering notation.

  4. 04
    URLPDF p.12

    Deliver proof

    Every finding retains both values, source locations, scan date, and a confidence label.

03 / FOCUSED BY DESIGN

Hard facts only.
No vague AI critique.

The first version checks facts that affect selection and quoting. It does not grade your translation, redesign your website, or pretend to know which public value is correct.

See the pilot scope

FIELDS WE CHECK

Model identityRange limitsVoltageCurrentPowerTemperaturePressureFlowDimensionsDN / NPSOutput interfaceIP rating
OUT OF SCOPE

SEO scoring · translation style · visual redesign · certification validity

04 / EARLY EVIDENCE

The problem is observable in public catalogs.

A directed validation scan looked at industrial sites with dense product catalogs and multiple public assets. It proves the problem exists—not how common it is across the whole industry.

20

industrial sites reviewed

Automation, instrumentation, electrical equipment, pumps, and valves.

11

with a high-confidence conflict or asset mismatch

Each supported by two public statements that could be independently checked.

EXAMPLE / MPM280
One table showed 40–125°C; the page copy showed−40–125°C.

A missing negative sign can change whether a product appears suitable for low-temperature use.

Directed, non-random sample. These numbers must not be interpreted as an industry-wide incidence rate.

05 / FOUNDING PILOT

Start with one finding.
Expand only if it matters.

Send us your domain. If we can identify a relevant, reviewable conflict, we’ll share one example before discussing a wider scan.

$49pilot scan
after sample review

FOUNDING PILOT INCLUDES

  • Up to 30 products
  • Product-page internal checks
  • HTML-to-PDF comparison
  • Source URLs and exact values
  • Confidence level for every finding
No checkout today. We confirm scope and fit by email first.

06 / FAQ

Questions your product team will ask.

What counts as a specification conflict?

Two public statements tied to the same model that cannot both hold after unit normalization—for example, different maximum pressures or voltage ranges.

What if the values describe different configurations?

If a configuration boundary can explain the difference, the finding is downgraded or withdrawn. Unclear grouping is reported as an identity or configuration risk, not a confirmed conflict.

Do you decide which specification is correct?

No. SpecDrift proves that public statements disagree and preserves the evidence. Your product or engineering owner confirms the authoritative value.

Do you need access to our CMS or PIM?

No. The pilot uses only public pages and text-based PDFs reachable from your domain. Private-source integration is outside the validation scope.

Is this translation or SEO quality assurance?

No. We focus on model identity and hard technical facts. Writing style, keyword scores, visual design, and certification validity are intentionally excluded.