MNR CEDEX Codes

Automating customer-specific container repair codes

Replacing manual CEDEX code lookups with automated, customer-specific code mapping, providing a solution for unprocessed invoices and cutting invoice rejections from 100% to about 10%.

Role

C#/SQL Developer (Full-Stack)

Timeline

December 2018 – June 2019

Team

Jeyam Silva & Michael Brutsch

~10%

invoice rejection rate, down from 100%

20% after Ver 1.0 in January, about 10% by June 2019

Timeline

Ver 1.0

8 weeks · Dec 2018–Jan 2019

Code dropdowns auto-filtered by invoice requirements, with selections saved to the database.


Ver 2.0

20 weeks · Feb–June 2019

Every step for this customer on one page: managing work orders, generating EDI invoices per location, and creating fusion files.

Support

June 2019 onward

Ongoing work to filter anomalies.




Who I worked with

End users (6)

3 AP specialists, 2 managers, and 1 director at our client.

External customers

Major shipping lines whose EDI rules and accepted repairs we had to match.

Development team

Me, a C#/SQL co-developer, and a senior developer.

Background

The situation

CEDEX (Container Equipment Data Exchange) codes are standardized alphanumeric values that shipping companies exchange to describe container components, damage types, and repair estimates worldwide. To invoice a customer for MNR (maintenance and repair) work, the finance team checked CEDEX charts for each repair, cross-referenced what that customer accepted (every customer differed), and typed a matching code onto each invoice file by hand. Wrong codes meant rejected invoices and further delays, leaving years of rejected work orders outstanding.

The goal

Ver 1.0

Recover years of rejected MNR work orders by quickly populating valid CEDEX criteria for each repair, with no manual input or user error. Success meant faster invoice processing, for the current week plus several backlogged weeks.

Ver 2.0

Remove the trip between the CEDEX page and the ABS home page by putting every step for this customer on the CEDEX page. Success meant lower cognitive load and better learnability.

Scope

In scope: database setup for the ETL process, SQL mapping tables and stored procedures, and the CEDEX page UI.

Out of scope: eliminating manual entry via Excel.

Constraints

• One customer with rigid EDI rules: accepted tariff string length, parts per repair, and which jobs could be bundled on an invoice
• Each shipping customer accepted a different set of repairs
• Some requirements only surfaced as invoices were accepted or rejected

Research & Process

With about six end users, research ran inside the agile build cycle. Observation, weekly sessions, and the customer’s own EDI feedback shaped both versions.

6

end users: 3 AP specialists, 2 managers, 1 director

3

users in the weekly usability sessions

8 wks

of weekly usability sessions

2

versions shaped by user feedback

01 Discover

Understand the real workflow

Observation and job shadowing

Watched how finance staff cross-referenced CEDEX charts and typed codes by hand.

Contextual inquiry

Separated what users said they needed from what the workflow needed.

Requirements gathering

Agile sessions feeding iterative design, documented in an SRS and SDD.

Stakeholder mapping

Mapped end users, leadership, and each shipping customer’s requirements.

02 Evaluate

Test with the people who use it

Weekly usability sessions

With 3 end users, every week for 8 weeks.

EDI message testing

Checked each message with the customer as new requirements were raised.









03 Iterate

Keep feedback flowing

Email feedback loop

Input from our client’s AP team and the customer’s AP department arrived by email.

Table-mapping updates

Used the customer’s success and rejection reports to update database mappings.

Agile delivery

Two versions, each shaped by what users told us.

04 Measure

Quantify the change

Error-rate analysis

Tracked invoice rejection rate, from 100% to 20% after Ver 1.0 and then to about 10%.







From Findings to Design Decisions

Each decision traces back to something users, or the customer, told or showed us.

What I learned

What we changed

Some requirements were learn-as-you-go, since our client’s repair descriptions differed slightly from the customer’s.

Used the customer’s success and rejection reports to keep updating the database table mappings.

Staff cross-referenced a customer pamphlet by hand and typed a code for every repair.

Code dropdowns auto-filtered to what each customer accepts, so nobody looks up or types a code.

The customer’s EDI rules were rigid: tariff string length, parts per repair, which jobs could bundle.

Built those rules into the code so the page enforces them, and tested EDI messages with the customer as new rules came up.

Users’ mental model of the flow didn’t match Ver 1.0, which sent them through ABS first. We had assumed matching ABS would feel familiar.

Ver 2.0 put every step for this customer on the CEDEX page, cutting cognitive load and favoring recognition over recall.

The designed system

Ver 1.0: the code dropdown filtered by the customer’s accepted repairs

Ver 2.0: the full workflow on one page

Results

100% to ~10%

invoice rejection rate, Dec 2018 to June 2019

$873K

of MNR work orders processed by June 2019

Beyond the numbers

Users no longer consult an external mapping sheet to search for, type, and track CEDEX codes for each work order. With Ver 2.0, the whole workflow for this customer sits on one page, so users were less confused and less likely to forget steps. Users adopted it, the customer accepted our EDI messages, and support requests fell.

Figures are measured. About 10% of the June total was flagged as rejections and set aside to iterate on.

Reflection

Look beyond what users say to find the system needs.

What the project taught me

Users may say they need one thing, but through contextual inquiry and observation, it’s the researcher’s job to use discernment and pinpoint what the system really needs. This project improved how I balance user needs with systems needs to serve both our clients and their operations.

What I’d do differently

Check the design against usability heuristics. Do more competitive analysis of comparable systems to encourage user efficiency. Run active surveys if possible, though the client didn’t ask for those metrics.

Looking back

Principles Ver 2.0 reflects:
• One-page workflow: recognition over recall
• Fewer steps: lower cognitive load
• Auto-filtered codes: error prevention