Have any questions Book a Discovery Call
How a master data service we ran for the client was automated once it was stable, cutting processing time by about 75% and catching duplicates before they became new codes.
| Client | Large, fast-growing industrial manufacturing group with operations across multiple regions and a hybrid central and regional procurement model. |
| Engagement | Managed procurement operations, including material master creation, delivered by EmpoweringCPO. Automation was introduced in the second stage after the process had stabilised. |
| Process Automated | Material master creation, covering request collection, ERP upload file preparation, and duplicate checks against the existing material master. |
| Volume | Approximately 300 requests and 2,100 line items per month, checked against an existing master containing well over 100,000 codes. |
| Delivery Model | The same dedicated EmpoweringCPO team, working remotely from India. The automation tool runs on the team’s workstations within the client’s environment. Approvals and final ERP uploads remain with the client. |
| Status | Live since approximately the end of the first year of the engagement. Monthly volumes are handled within regular working hours, with the existing SLA maintained unchanged. |
A fast-growing industrial manufacturing group asked EmpoweringCPO to run a defined set of procurement processes, including the creation of new material codes. In the first year, the team built what the process did not have: documented SOPs, a measurable SLA framework and reliable reporting.
After about a year, once the process was stable, EmpoweringCPO automated it. Requests are now collected and prepared as ERP-ready upload files through ten automated quality checks, and every new item is compared with the existing master before a code is created. Exact matches are accepted automatically, near-duplicates go to a reviewer, and the rest become new codes.
Processing time per 100 line items fell from about 18 hours to about 4.5, a reduction of about 75%. About one in three line items turned out to be a duplicate or near-duplicate of an existing code, so far fewer duplicate codes are now created and the quality of the master data has improved. The work now fits into regular hours, and the SLA commitment is unchanged.
Creating a new material code looks like data entry. In practice it is a chain of checks, repeated hundreds of times a month:
| Challenge | What it meant |
|---|---|
| Volume | About 300 requests and 2,100 line items a month, every one needing to be formatted, checked and prepared for upload. |
| Many Formats | Requests came from different users and departments, with different column names, units, abbreviations and levels of detail. |
| Duplicate Risk | Before creating a new code, someone had to search a master of well over 100,000 codes by hand. Manual searching was slow and could miss near-matches, allowing duplicate codes to creep in. |
| Capacity | The work was repetitive and took about two working days for every 100 line items, resulting in extra hours whenever volumes peaked. |
| Consistency | Short descriptions, units, make, model and part numbers varied from request to request, making the master harder to search and trust. |
The client first outsourced the process to EmpoweringCPO. We automated it later, not on day one, and that was deliberate. Automating an unstable process only makes inconsistent habits run faster. The engagement followed four stages:
| Step | What was in place | What it made possible |
|---|---|---|
| 1. Stabilise | Documented SOPs, an agreed SLA framework, standard reports and a quality audit routine for every process. | A process that is written down, measured and repeatable. |
| 2. Codify | The checks the team applied by hand were written as clear rules: what a clean description looks like, which fields are mandatory, and which differences make two items different. | Rules that can be tested and applied the same way every time. |
| 3. Automate | The rules were built into a tool that prepares the upload file and compares each new item with the existing master. | Routine checking done in minutes, with people kept for the judgement calls. |
| 4. Improve | Reviewers handle near-duplicates, and the search index is refreshed every day with new codes. | A service that stays accurate as the master grows. |
From request intake to a finished upload file, with people reviewing every unclear match.
Each line item passes through ten automated checks, grouped below. Fields that cannot be fixed automatically are highlighted for the team instead of being guessed.
| Group of checks | What is checked or fixed |
|---|---|
| Text Cleaning | Special characters, repeated symbols and spacing are cleaned so every description follows one standard. |
| Dimensions and Units | Inch marks and size notations are written out in one format, and units of measure are matched to the approved list. |
| Tags and Identifiers | Serial numbers and equipment tags are moved to their own fields and removed from the description. |
| Short Description | Written in capitals, with common abbreviations expanded, and kept inside the 40-character limit of the ERP. |
| Make, Model and Part Number | Missing values are picked up from the long description, and unclear or over-long entries are tidied. |
| Completeness | Anything still missing is highlighted in colour, so the team knows exactly what to chase. |
New codes are added to the search index every day, so the next request is compared with a master that is already up to date. The update takes seconds, rather than a full rebuild.
Measured per 100 line items, the process from downloading the request to a finished upload file took about two working days before automation. It now takes about 4.5 hours, including review.
Working hours per 100 line items, before and after automation.
About one in three line items was already in the master, or came very close to an existing code. Exact matches are accepted automatically. Near-duplicates are always checked by a person. Before automation, around 80 of every 100 line items ended up as new codes, because manual searches missed many items already in the master. Now about 66 do, so roughly 14 fewer duplicate codes are created per 100 line items, and the quality of the master data is higher.
| Measure (per 100 line items) | Before | After |
|---|---|---|
| Processing Time, from Download to Finished Upload File | About 18 hours | About 4.5 hours |
| Duplicates and Near-Duplicates Identified | Found by manual search | Found by the duplicate check, about 1 in 3 line items |
| Exact Matches Accepted Automatically (AUTO) | Manual, around 10 | About 26 |
| Near-Duplicates Sent for Review (REVIEW) | Checked manually, not counted separately | About 8 |
| No Match Found, So a New Code Is Created (NO MATCH) | Around 80, as many duplicates went undetected | About 66 |
| Extra Hours at Peak Periods | Regular | Significantly reduced |
| SLA Commitment | Met | Unchanged, met within regular hours |
| At current monthly volumes (about 2,100 line items) | Before | After |
|---|---|---|
| Hours of processing | About 380 | About 95 |
| Hours saved | About 280 a month |
The same workload is now completed within regular working hours, where it previously needed extra hours. Material descriptions are more consistent, and duplicate codes are caught before they are created.
We do not quote an accuracy percentage for automatic matches. The design is built around a simple principle: AI proposes, rules protect, and people decide anything that is not clear. Rules stop a match when part numbers, model series or ratings differ, and near-duplicates are never accepted without review.
| Value delivered | How |
|---|---|
| Faster turnaround | The same volume now fits into regular working hours, without overtime peaks. |
| Fewer duplicate codes | Every new item is compared with the full master before a code is created, so duplicates are stopped at the door. |
| More consistent master data | One set of rules gives every description, unit and identifier the same format. |
| Capacity for higher-value work | The team size did not change. The time freed is now used on near-duplicates and exceptions, and on finding other processes that can be automated. |
| Same service levels | The SLA commitment stays unchanged, and quality checks run on every request. |
The team size is unchanged, and the capacity freed is now used to identify other processes that can be automated in the same way. The rules and the review step continue to be refined. To see how this fits into our wider approach, read about our master data management and procurement outsourcing services, or contact us to talk about your own process.
Outsource your material master to EmpoweringCPO. We document and run the process to an agreed SLA, then automate it, with people reviewing anything unclear. Automation is normally part of the managed service. Scope, and any infrastructure or licence costs, depend on the complexity and on how you want it deployed.
Client identity withheld for confidentiality. Figures are rounded and adjusted, and are per 100 line items at the volumes described. ~ denotes approximate. Time includes download, formatting, duplicate checking, review and preparation of the upload file. Results will vary with the organisation’s data and processes.
EmpoweringCPO is a team of procurement professionals, founded in 2011 and based in India. We help organisations find procurement savings through procurement consulting and run their processes for them through procurement outsourcing, under agreed SLAs.
Get A Call Back