South Walker Creek: Restoring Trust in the Engineering DMS
When South Walker Creek was purchased, more than 96,000 engineering artefacts were unsearchable. Every drawing had to be opened by hand. Here is how the live drawing set was rebuilt, and trust restored, without compromising governance.
Search no longer worked
As part of a digital transformation initiative, South Walker Creek migrated its engineering drawings into a new Document Management System. That is rarely a clean exercise. Consolidating engineering data spread across spreadsheets and existing document systems, then verifying it all into a new DMS, is laborious, painstaking work, and detail inevitably gets lost along the way.
Titles came across missing or generic, revisions were unclear, and thousands of non-drawing artefacts ended up mixed into the live engineering environment. With more than 96,000 artefacts in the system, engineers and coordinators could no longer trust search results. Every drawing had to be opened and checked by hand before use, slowing work down and raising the risk of referencing a superseded or incorrect drawing.
Adopting a new system is hard. Managing the change is harder.
From 96,000 artefacts to a clean live set
Around 96,000 artefacts were brought into DrawingHub and re-processed from scratch. Two passes of clean-up then removed everything that didn’t belong in a live engineering environment, leaving a current set of over 51,000 drawings. More than a third of what was processed was discarded as a duplicate, a superseded revision, or simply not a drawing.
Any file that is an exact copy of one already brought in is caught and set aside before it reaches the live data. This removed 21,692 duplicate files. Nothing is lost: every file is accounted for as either a duplicate set aside or a unique drawing taken forward.
Of the 74,432 drawing records created, a further 22,700 superseded or out-of-date sheets were retired from the live set, so current and superseded drawings are no longer mixed together.
Read from the drawing, not the migration
Rather than rely on the generic fields carried over in the migration, metadata was extracted directly from the drawings themselves, read off each drawing’s title block. The title block was located on ~93% of drawings, and the key fields read straight from it: drawing title, drawing number, and revision, plus the full revision history where the drawing records it.
Roughly 7 in 10, end to end
Most of the metadata was AI-discovered, found and populated automatically, with people stepping in to verify and correct only where it mattered. Counting a drawing as fully AI-discovered when every field was found by the AI, and human-assisted when a person entered or corrected at least one field:
Find the right drawing, first time
Engineers and coordinators on site can now search and actually find the right drawing the first time, without opening files to verify them. That removes a daily friction point, speeds up work, and lowers the risk of acting on a wrong or outdated drawing.
It also strengthens revision confidence, which was at the heart of the original problem. With each drawing’s revision now captured, the data shows where the same drawing number exists at more than one revision. The information needed to tell a current drawing from a superseded one now lives in the metadata, rather than being locked inside each file.
After migrating to our Engineering Drawing Management System, our engineers and coordinators couldn’t find the drawings they were searching for. DrawingHub reviewed our current state, re-processed all our drawings, removed duplicates and populated the metadata fields… We are now at close to 100% metadata extracted from our drawings. This rebuilt confidence in the system without compromising governance. We’ve seen an uptick in people finding drawings that aren’t right and submitting red pen markups quickly. Our DMS is now a system our technical personnel on site can rely on again.
Good metadata is the start, not the end
Inferring new metadata
Once the core fields are accurate and trustworthy, they become a foundation for working out information never written on the drawing itself. Functional Location (FLOC), tying each drawing to the physical asset it belongs to, is one example the team is exploring. The cleaner the base metadata, the more can be inferred on top of it.
Improving findability
Accurate metadata is only the first step in helping people find the right drawing. Finding new ways to enhance a customer’s existing DMS, and to get more value from the metadata already in their system, is one of the newest frontiers we are exploring.
The work in figures
−21,692 exact-duplicate files removed before ingestion
−22,700 culled — duplicate / superseded / non-drawing> 1 in 3
Let’s make your drawings findable
“DrawingHub exists to make every engineering drawing findable the moment it’s needed, so frontline workers never take a risk they didn’t have to.”