Product - B2B Enterprise 08 0 to 1 B2B Industrial

Building a custom PIM from scratch
for a mission-critical industrial manufacturer

PrecisionSense GmbH managed 3,000+ SKUs across four regulated industries using a combination of ERP exports, email threads, and Slack messages. One person spent half their working week just keeping the data coherent. I led the product definition and delivery of a custom PIM system that replaced the entire manual stack.

3,000+SKUs managed
-18 hrsManual work per week
-91%Data update time
4Regulated industries

PrecisionSense GmbH is a fictional industrial sensor manufacturer based in Germany with 8-figure ARR. Their product portfolio spans 600 base products and 3,000+ SKUs, serving four highly regulated sectors: automotive EV battery monitoring, medical diagnostic equipment, renewable energy pitch control systems, and ruggedized aerospace sensors.

In each of these sectors, an incorrect technical specification is not a customer service problem - it is a legal and safety liability. The stakes around data accuracy were unusually high.

PrecisionSense relied on their ERP system for transactional data: pricing, inventory, and orders. But ERP was never designed to carry descriptive technical data, compliance certifications, or marketing assets. The gap between what ERP could store and what customers actually needed had been filled entirely by manual human effort.

Manual overhead 20 hrs

Per week spent by one person manually gathering data from ERP, email, and Slack to update product records. 50% of a full-time role.

Update speed 2-3 min

To identify, extract, and update a single product request - creating a queue backlog that grew faster than it could be cleared.

Data risk 0 audit

No audit trail existed for who changed what and when. In medical and aerospace contexts, this was a compliance exposure.

SKU complexity 3,000+

Variants managed in rigid ERP tables that had no concept of parent-child relationships or attribute inheritance.

"The ERP knows what we sell and how much we have. It has no idea what the product actually is, what it does, or whether it is certified for use in a medical device."

The deeper issue was that product data lived in three separate systems that never talked to each other: engineering data in a PLM system, transactional data in the ERP, and commercial data in spreadsheets and email. Every customer-facing output - datasheets, portal listings, catalog entries - required someone to manually reconcile all three. That person had become a critical single point of failure.

Before scoping a custom build, I evaluated three established SaaS PIM platforms. All three failed on requirements that were non-negotiable for PrecisionSense.

01
Pricing at scale. Standard SaaS PIMs charge per SKU. At 3,000+ SKUs with a roadmap to 10,000+, the annual licensing cost exceeded the custom build cost within 18 months. The client would have been paying more to use worse software.
02
Workflow rigidity. Compliance approval flows in medical and aerospace contexts require custom logic: multi-tier sign-off, external partner review stages, and product-type-specific validation rules. No off-the-shelf tool supported this without expensive professional services that reintroduced the vendor dependency we were trying to escape.
03
ERP integration depth. The client ran a highly customized ERP instance. Standard connectors offered shallow, one-way sync. The business required real-time bidirectional data flow between ERP, PLM, and the new PIM. Only a custom API layer could deliver this reliably.
04
Data sovereignty. Storing mission-critical technical specs and compliance documents on a third-party SaaS platform was not acceptable to the client's legal team. A self-hosted custom system gave them complete control over their intellectual property.

I led product definition end-to-end: translating the client's operational pain into a functional spec, making the build-vs-buy decision, defining the technical architecture priorities, scoping each feature against the core use cases, and owning the delivery timeline in partnership with the engineering team.

The architecture philosophy I set from the start was "vanilla first" - minimal external dependencies, prioritizing auditability and long-term maintainability over feature richness. For a lean industrial team with no dedicated software ops function, a system that is easy to understand and secure is more valuable than one that is feature-complete but opaque.

Five core modules, each mapped directly to a specific operational failure in the current state.

Module Problem it solved Key technical decision
ETL data onboarding Migrating 3,000+ SKUs from ERP exports and Salesforce CRM without months of manual re-entry Custom Extract-Transform-Load engine supporting Excel and Salesforce; even with manual cleaning required, migration time dropped from months to days
Parent-child data model Managing 3,000+ variants that share 80% of their attributes with a base product Attribute inheritance: changes to a parent spec propagate instantly to all child variants, guaranteeing 100% consistency across the SKU tree
RBAC and audit trail No visibility into who changed what; unverified edits to technical specs in regulated industries create legal exposure Granular view/edit permissions by role, admin approval workflow for core spec changes, and an immutable change log for compliance audits
Technical asset management BOMs, RoHS/REACH certifications, and CAD files living in email chains and Slack - impossible to version or locate reliably Direct attachment of engineering files to specific SKU variants; always-current link between the physical engineering asset and the digital product record
Full data export Avoiding vendor lock-in and giving the client legal sovereignty over their own product data Structured .zip export of all data at any time; the client can leave, migrate, or audit without asking permission from a vendor

Architecture decision worth noting: the system was intentionally built with minimal external library dependencies. This was not a technical constraint - it was a product decision. An 8-figure industrial client with no dedicated software ops team needs a system where a security audit is straightforward and where there is no npm-style dependency chain to maintain. Simplicity was a feature.

Data sources
ERP (custom instance) Salesforce CRM PLM system Excel/CSV exports
ETL layer
Custom ETL engine Data validation rules Transform mappings
Core PIM
Parent-child data model Attribute inheritance RBAC permissions Audit trail Asset management
Output channels
B2B customer portal E-commerce product feed Print catalog export Structured .zip export
Before
  • 20 hours/week of manual data reconciliation per person
  • 2-3 minutes to update a single product record
  • Technical specs, certifications, and CAD files scattered across email and Slack
  • No audit trail; anyone with ERP access could change any field
  • ERP, PLM, and sales data siloed with no automated sync
  • Variant management in flat ERP tables with no inheritance
After
  • 2 hours/week for oversight and exception handling
  • Under 15 seconds to update a product record
  • All assets attached directly to the relevant SKU variant, versioned
  • Full audit trail with role-based approvals for regulated fields
  • Real-time bidirectional sync between ERP, PLM, and PIM
  • Parent-child model with attribute inheritance across the full SKU tree
-91% Update time

From 2-3 minutes per product record to under 15 seconds. Queue backlog cleared within the first week post-launch.

18 hrs Recovered per week

The role that spent 50% of its time on data reconciliation now spends less than 10% on oversight. The rest is reinvested in strategic work.

100% Audit coverage

Every field change is now logged with timestamp, user, and approval status. First compliance audit passed with zero findings on data integrity.

3 days Migration time

3,000+ SKUs migrated from ERP and Salesforce via the ETL engine. Previous estimate for manual migration was 4-6 months.

0 Vendor dependency

Self-hosted, fully exportable, no SaaS licensing. The client owns the system, the data, and the roadmap.

4 Channels synced

B2B portal, e-commerce feed, print catalog, and structured export now all draw from a single source of truth.

The most important decision on this project was made before any code was written: choosing to build custom rather than configure off-the-shelf. That decision required a structured business case comparing 3-year TCO across options, not just a feature checklist. The custom build was cheaper at scale, more compliant by design, and ultimately more aligned with the client's risk profile than any SaaS alternative.

The "vanilla architecture" principle was also a product decision, not just a technical one. When you are building for a team that has no dedicated software operations function, the ongoing cost of complexity is real and ongoing. A system with fewer dependencies is easier to audit, easier to hand over, and easier to modify two years later when the original engineers are gone. Simplicity has a business value that rarely appears in a feature comparison.

The hardest conversation on this project was telling the client that one of their three top-priority features needed to move to phase two. The compliance approval workflow had scope that would have pushed the initial delivery by six weeks. Cutting it from v1 and shipping the core PIM on time - with a committed roadmap for the approval module - was the right call. The team had the data benefit within their original timeline, and the approval module shipped eight weeks later with better specification because we had real usage data to inform it by then.

Figma design - Simple prototype

The actual UI of the prototype is larger and only simple functions are recreated for the purpose of the case study.

← All work
Back to portfolio
View all case studies