Data Engineering

Role summary

Master data is the constraint on the FY27 CRM portfolio. Several funded projects assume clean customer and order data as a starting condition, and today that data is corrected largely by hand.

This is a build role. The holder designs and builds the matching, merging, validation and assignment logic in Salesforce, proves it against real records, and documents it. Data remediation is part of the job, but as the means of validating and tuning what is being built — not as a standing queue to be absorbed.

The measure of success is the manual share falling. A candidate whose answer to volume is more hours is the wrong candidate for this role.

1 · Automation, rules and validation

• Build the matching, merging and deduplication logic in Salesforce so that rule-based decisions stop being made by a person.
• Build and maintain validation rules that prevent the defect classes that are currently corrected after the fact.
• Build the OPSI assignment logic and the checks that stop an account being created without one.
• Build hierarchy resolution for parent–child, reseller, distributor and end-customer relationships that do not resolve on first pass.
• Work the master data items in the AI automation funnel — missing-data detection, classification of incomplete orders, account-matching assistance.
• Test what is built and evidence it: match precision on an audited sample, manual touches per order before and after, defect rate before and after.

2 · Proving it on real records

• Run each rule set against live account and order populations to measure precision and recall before it is released, and tune it on the result.
• Remediate the records surfaced by that testing where the remediation is what proves the rule works. This is test execution against production data, scoped to the increment being built.
• Establish the before-and-after baseline for every rule released: match precision on an audited sample, manual touches per order, defect rate.
• Sign off the exit criteria for each increment with the Product Owner Master Data before it goes live.

3 · Data operations while the build progresses

• Work the residual account remastering and indirect order queues that the automation does not yet cover. More than 30 per cent of indirect orders arrive incomplete today.
• Apply OPSI assignment where no rule yet handles the case. OPSI gaps are a known source of downstream failure — in 2024 more than 500 accounts were created without one.
• Hold the match yield while the build progresses. The FY26 baseline for order data matching is 80 per cent, up from 8 per cent at the start of the remediation.
• Feed the Distributor Order Scorecard so order quality can be measured by distributor rather than estimated.
• Where a defect pattern repeats, it is converted into a validation rule rather than corrected twice. This part of the work is expected to shrink across FY27 as the automation lands.

4 · Functionality specification and design documentation

• Write the functionality specification for each increment before it is built: the matching and merge criteria, OPSI assignment conditions, hierarchy resolution logic, validation rules and the expected behaviour of each.
• Produce the design documentation that accompanies the build — data model impact, rule dependencies, and the decision logic in a form another engineer can pick up.
• Record the test approach and the acceptance evidence for each specification, so what was specified can be shown to have been delivered.
• Keep the specification current as the data model changes. A specification that lags the rule set is worse than none.
• Maintain the operating procedures for the residual queues so that work is transferable and does not sit with one person.

5 · Supporting the FY27 portfolio

• Territory and account data for Jetstream, the US territory re-alignment that goes live 1 November 2026. Master data capacity is the stated constraint on it.
• Customer and entitlement data quality as the input to Customer 360 and the WXP entitlement work, both of which assume clean data as a starting condition.
• BPID and trading-structure setup where it blocks order flow.

6 · How the work is run

• Work inside the Jira flow with the rest of the value stream: estimated, tracked and visible. No side queues.
• Log time against the project and the work item, so throughput and automation impact can be measured.
• Report throughput, match yield and manual effort removed on an agreed cadence.

Required experience and skills

• 5–8 years relevant experience in master data, data operations or CRM data quality in a large enterprise.
• Hands-on Salesforce build: flows, validation rules, duplicate and matching rules, data loader and bulk operations, reports. This role configures and builds, it does not only operate.
• Scripting — Python or equivalent — and experience applying automation or ML to classification or matching problems. This is what distinguishes a strong candidate.
• Demonstrable experience cleaning customer or account data at volume: able to describe a specific backlog, what was done to it, and what changed in the numbers.
• Comfortable working to a measure. The role is assessed on throughput, match yield and manual effort removed.
• Works across distributed teams in EMEA, AMS and APJ with overlapping hours.
• Clear written English. The rules this role writes are used by other people.

Nice to have

• Fortune 100 experience in CRM, order management or channel data.
• Familiarity with OPSI, BPID, HPD and OSS, or equivalent enterprise customer and order master structures.
• Salesforce Administrator, Platform App Builder or Data Architect certification.
• Exposure to Jira, TestRail and an agile delivery cadence.