Asset data collection from the nameplate up — every value read from the photo, checked against your asset register and approved by a person before your EAM loads it
See one case, screen by screen ↓Built forField techniciansMaster data reviewersAsset data ownersMaintenance planners
Field crews copy each nameplate onto paper and someone retypes it into the asset register, so ratings are mistyped or guessed.
Every plate, copied by hand
Built-in controls
Parts orders and capital plans can rely on ratings traced back to the plate.
Asset data collection is gathering the facts about each piece of equipment (manufacturer, model, serial number and ratings) so the asset register in your EAM is complete and correct. The most accurate source is usually the equipment nameplate. Asset Data Collection turns nameplate photos into validated, approved asset records, each value tied to the place on the plate it came from.
How to choose AI solutions for utilities →
Updated
Asset data collection still starts with a person in a pump station reading a nameplate, writing it down, and someone else typing it into the EAM. Values are mistyped. A plate rated 230/460 volts is guessed at one of the two. Characters worn away by rust are filled in from memory. The same drive is created twice because nobody checked the register, and controlled lists fill up with free text.
By the time the record reaches the asset register, nobody can show where a value came from. Then the spreadsheet has to be remapped, column by column, into the layout of the EAM's asset loader template before it can be loaded. A mistyped rating sends a crew out with the wrong part, and capital plans rest on age, size and ratings that are missing or wrong.
Estimated split for one walk-down of about 50 assets, in working hours, by hand and with the solution. Our estimate, not a measured result.
A water utility, seven weeks after go-live. A utility technician walks a booster pump station with a managed iPad, and a Master Data reviewer has eight open records in the queue. Here is what happens to one pump motor, screen by screen, in the working solution.
A worked example on sample data. Names, IDs and figures on the screens and in this section come from the demo.
The technician scans the asset tag and photographs the 75 HP motor. The capture coach flags glare on the first photo, so the technician takes a second angle and an overall shot. There is no signal in the station, so the photos wait on the iPad and upload once it returns. On the next pump motor, the same check asks for a tilt of about 10° or shade over the plate.
Safety first: no one opens a guard or climbs for a photo.
The planner drops 22 photos from a raw water pump station. The agents fingerprint each photo, group them into 7 assets by tag and time, flag glare on one photo and low light on another, and skip one photo as identical to an earlier upload. Nobody sorts photos by hand.
1 photo skipped as identical to an earlier upload
The plate reader reads 3 identity values and 10 characteristic pairs, plus 6 more plate values that are kept but not exported. The serial reads at 98%. When the reviewer clicks the serial number, the exact region of the photo it was read from is outlined and zoomed. Approve stays locked: 3 items to resolve.
Click any value to see the part of the plate it was read from.
The service factor sits under glare from the work light in photo 1, where it read at 62%. Photo 2, taken at a second angle, reads 1.15 at 93%. Both readings are kept, and the reviewer confirms 1.15 against photo 2.
Both readings are kept; the reviewer confirms the value against the photo it came from.
The plate lists 230/460 volts and the EAM characteristic takes one value, so the normalizer leaves it blank under the dual-rating rule. The reviewer picks 460 V, the voltage the motor is connected for, and full-load current follows that choice from 172/86.0 to 86.0 A under the next rule in the set. Both values carry the reviewer's initials as human-verified.
A dual rating is not averaged or guessed: the reviewer picks one.
The motor's serial has no match in the latest monthly register extract. The pump's drive does: its serial is already EAM equipment, with the same horsepower, voltage, current, phase and enclosure. The submission and the existing equipment sit side by side, and the reviewer marks it as the existing asset, so it is not exported and its photo is kept as evidence.
The duplicate checker does not merge or reject; a reviewer decides.
On an aeration blower at the water reclamation facility, four characters of the serial are rusted through. Brake horsepower is washed out by glare from the skylight. Both are left blank and flagged, the duplicate check falls back to model and tag, and the reviewer requests a recapture, which lands on the planner's capture list with the guidance.
Four characters are rusted through: left blank and flagged, with none guessed.
The approval shows exactly what the reviewer signs: 3 identity values, 10 characteristic pairs, 3 values the reviewer verified and 19 customer-owned columns left blank. The reviewer's name, the time and version 1 are recorded. Any later change to the record needs a new approval.
A reviewer approves each record; confidence does not replace sign-off.
The preview shows all 62 columns, A to BJ, of the SAP-style example template: manufacturer, model and serial in N, O and P, ten characteristic pairs in AB to AU, and the customer-owned columns blank. Each cell opens to its source, rule and approver. Column P for one transformer holds its serial, read from photo 1, within 18 characters, with the reviewer's name and the time of approval.
Empty characteristic slots stay empty, not filled to reach ten pairs.
Only approved records go in. Nine validation checks pass, from 62 columns in template order to codes from the Characteristics sheet and nothing already in the EAM. The Excel file is written with its validation report: 8 records, 0 errors. The EAM data owner loads it into the EAM and records 8 accepted and 0 rejected, so the load reconciles.
The EAM data owner loads the file without reworking a column.
The same pump motor, seen by the five people who photograph it, upload it, approve it, administer it and load it: their week before, and their week now.
From the photo coach to the loader file, seven agents handle each nameplate; a Master Data reviewer approves each record.
Checks each photo as it is taken (focus, exposure, glare, how much of the frame the plate fills and the viewing angle) and gives one plain instruction to fix a failed check: tilt, shade, wipe, step closer or take a second angle.
Gives each submission its ID, stores the original photo untouched with its fingerprint, groups desk uploads by asset tag and capture time, and resumes an interrupted upload from where it stopped.
Reads the manufacturer, model, serial and each characteristic with its unit, returning the text exactly as stamped, the photo and region it came from and a confidence. When two photos disagree, it returns both readings for review.
Maps each value to your EAM's characteristic code and unit under the approved rule set, and checks asset types, the process hierarchy, equipment abbreviations, lengths and required fields. Raw, normalized and reviewer values are kept apart.
Compares each record with the monthly asset register extract from your EAM and with the other submissions, by serial, model and tag, and by photo fingerprint.
Writes the Excel file in the exact layout of your asset loader template from approved records, and the validation report that goes with it.
Scores extraction field by field on the acceptance images and on reviewer corrections, and writes the monthly accuracy, volume and completeness report.
Approves; owner rules lists; data owner loads.
Reviewers ask about the queue, a nameplate, a value or an export in plain words. Each answer points to the record and the photo region it comes from. These answers use the sample data.
What needs me today?
Eight open records, none overdue. Two are ready to approve: a pump and an aeration blower. Four need a decision: the pump motor has a dual voltage to pick and one value to confirm; a second aeration blower has two unreadable values; the pump drive matches existing EAM equipment; a blower motor has an enclosure value that is not on the list.
Why is service factor flagged on the pump motor?
Glare from the work light covers the SF cell in photo 1, where it read at 62%. Photo 2, taken at a second angle, reads 1.15 at 93%. Both readings are kept, and you confirm the value against photo 2 before approval.
Which records are blocked from export?
Anything not yet approved is blocked: 8 open records. The 7 approved records pass each check: 62 columns in template order, identity in N, O and P, codes from the Characteristics sheet, and the customer-owned columns blank.
How accurate are we against our go-live target?
Against the target set at go-live (95% in this example), on the acceptance images: identity fields 97.1% over 888 readable fields, characteristic values with units 95.8% over 2,210, and 137 of 141 unreadable fields correctly left blank. Frame size and NEMA rating sit below 95% on their own and are on the error list.
From the working solution, on its sample data. Pick one to see it large.
The target system, the loader template, the codes and lists, the accuracy target and where your data lives are settings your administrators and Master Data owner control.
| Setting | Default | Choose from |
|---|---|---|
| Target system | SAP-style equipment records in this example | Your EAM or ERP: SAP, Maximo and similar · no connection, the file is the hand-off |
| Asset loader template | AL-1.0 · Data sheet, 62 columns | Your own template · a change is assessed, configured, regression-tested and released through change control |
| Characteristic profile | Up to 10 code and value pairs per asset type | Core codes on every motor and drive record · the others come with each asset type's profile |
| Normalization rule set | v1.4 · 9 rules, approved by the Master Data owner | A proposed change goes to the Master Data owner and is run on the acceptance images and the last three exports before release |
| Accuracy target | 95% in this example | Set at go-live · a pipeline change that scores below it is not released |
| Review time | Within 2 business days of capture | Set by your administrator |
| Usage alerts | At 70%, 85% and 95% of the image allowance | At the approved not-to-exceed limit new extraction pauses, every submission is kept and nothing is billed above the limit without written approval |
| Train any model on your data | Off, locked | Off for us and every provider · changing it needs your written consent |
Each figure carries its label: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data (the accuracy figures are its sample figures) · “target” = the design goal, measured in the live solution; the accuracy target is set by each utility at go-live, 95% in this example · “estimated” = our split of the working hours for one walk-down, not a measured result · “typical” = published figures (US EPA, 7th Drinking Water Infrastructure Needs Survey and Assessment, 2023: $625 billion over 20 years; Gartner, citing its 2020 data quality survey: $12.9 million a year on average). Context: EPRI notes that utility asset databases often hold incorrect or missing asset descriptions, and that nameplates are the most accurate source. SAP and Maximo are named as examples of target systems only. Names on the screens are sample data.
Asset data collection is gathering the facts about each piece of equipment (manufacturer, model, serial number and ratings) so the asset register in your EAM is complete and correct. The most accurate source is usually the equipment nameplate. Asset Data Collection turns nameplate photos into validated, approved asset records, each value tied to the place on the plate it came from.
Asset nameplate data collection starts in the field. Field crews photograph equipment nameplates on a phone or tablet, or upload walk-down photos from a desk. Seven agents check the photo, read every value with a confidence and its region, map it to your EAM codes and units, and check the register for duplicates. A Master Data reviewer approves each record, and only approved records go into the asset loader file.
It produces the file your team loads. Approved records are written into an Excel file in the exact layout of your asset loader template for your EAM or ERP, such as SAP or Maximo. The example is an SAP-style Data sheet with 62 columns. There is no connection to the EAM, and nothing is written to a system of record.
Unreadable characters are left blank and flagged, not completed, and the reviewer can request a recapture with guidance for the field user. When a plate lists two ratings, such as 230/460 volts, the value waits for the reviewer to pick one, and related values like full-load current follow that choice.
Each value is checked against your codes, units, controlled lists, lengths and required fields, and against the monthly register extract and other submissions for duplicates. Before a file is built, nine checks run, including 62 columns in template order, customer-owned columns blank and nothing already in the EAM. Each file carries its validation report.
You set the accuracy target at go-live (95% in this example), and it is measured field by field on your own acceptance images and each month. On the sample acceptance set, identity fields read at 97.1% over 888 readable fields and characteristic values with units at 95.8%, and 137 of 141 unreadable fields were correctly left blank. A pipeline change that scores below the target is not released.
No. A named reviewer approves each record, the Master Data owner decides each list and rule change, and the EAM data owner decides each load. Agents do not create a code, choose between dual ratings, merge a duplicate or fill a customer-owned column.
Your asset loader template, characteristics sheet, controlled lists and a register extract. The Agentic Solution Engine builds and deploys the solution from those requirements and documents, and it goes live once each quality gate has passed, including the accuracy target you set on your acceptance images. We will walk you through it on your own nameplates first.
Defects caught while they grow: every structure compared with last year's flight.
1-min demoAsset Health IndexThe same capital budget is built to remove 2.5× the risk, every score traced to its evidence.
1-min demoMaintenance Knowledge SearchDecades of fixes, searchable before the expert retires.
1-min demoWe'll run Asset Data Collection on a sample of your own nameplate photos and your asset loader template.