CPO Path #32

CPO Path #32 CW 44 2026

Hi {{first_name}}

Part 3 of 6 in The Redesign. Last week’s value bridge connected released hours to an outcome finance can recognise. Before any of that can happen, one question tends to stop the pilot: is the data good enough?

Enjoy and have a great week ahead

Pascal

An AI initiative is ready for testing. Procurement sees the opportunity; IT points to incomplete master data. Implementation is deferred until the data is fixed.

ProcureAbility’s 2026 benchmark of 160 senior procurement leaders found that 50% identified poor data quality as the leading barrier to scaling AI, while only 11% reported scaling AI across multiple procurement processes (ProcureAbility, 2026).

A useful leadership question is whether the entire database needs remediation — or only the information on which a specific decision depends.

The same benchmark identifies data quality as the leading analytics gap for 51% of respondents and the principal supplier-management challenge for 44% (ProcureAbility, 2026).

These findings establish data quality as a significant concern. They do not establish that comprehensive master-data cleansing is a prerequisite for every AI application.

The distinction matters for procurement directors whose information dependencies extend across engineering, finance, IT and supplier management.

A broad remediation programme may improve the underlying infrastructure. Yet its scope, cost and timeline can exceed what a controlled commercial use case requires.

The opportunity lies in a narrower decision: which information needs to be reliable, to what standard, and under whose accountability?

Data quality is not a binary condition.

The UK Government Data Quality Framework, drawing on DAMA UK’s dimensions, distinguishes completeness, uniqueness, consistency, timeliness, validity and accuracy. It defines quality as fitness for purpose and recognises that dimensions can trade off against one another depending on the intended use (Government Data Quality Hub, 2020).

Thomas Redman’s work in Harvard Business Review offers a complementary perspective: assessing actual records through a small structured sample can reveal whether data-quality concerns are material before wider remediation begins (Redman, 2016).

Consider an AI-assisted quotation comparison for electronic components.

A supplier record might lack a sustainability classification or contain outdated contact details. Those defects may matter elsewhere, but neither necessarily prevents a controlled comparison of technically equivalent quotations.

An incorrect price unit, engineering revision or currency, however, could materially distort the purchasing decision.

The distinction is not between clean and dirty data. It is between decision-critical defects and defects outside the current decision.

That raises an organisational question. Engineering controls technical equivalence. Procurement controls supplier quotations. Finance establishes monetary consistency. IT governs data access and integration.

A procurement director cannot guarantee all four alone. The director can, however, define the decision, obtain agreement on its data requirements and assign responsibility for exceptions.

This is where a data-quality discussion becomes an exercise in cross-functional leadership.

They start with one bounded use case — here, AI-assisted comparison of electronic-component quotations.

They select 100 recent RFQ lines and measure the share of records passing each field’s quality check, using the original quotation and the approved technical specification as references.

The Data Minimum — Illustrative RFQ Triage

*Illustrative results, not industry benchmarks. Each rate is the share of sampled RFQ lines passing the stated check. Effort assessments are illustrative.

The acceptance threshold is 100% validation of these seven checks for every quotation admitted into the pilot comparison — not 100% completeness across the underlying database.

Records failing a check are flagged for correction or excluded. AI may prepare the analysis; a designated buyer retains responsibility for commercial interpretation and approval.

Equally important is the stop list: historical supplier descriptions, unused classification codes, legacy contacts and unrelated attributes remain outside this remediation scope. The exclusions do not waive applicable compliance, supplier-qualification or security requirements.

The director can now put a specific proposal to finance, engineering and IT:

The pilot proceeds when accepted records meet the checks and the required controls are in place. Otherwise, the affected comparisons remain blocked until corrected.

This creates a measurable boundary between sufficient readiness and unresolved risk.

Data readiness becomes an executive decision when procurement can define what must be trusted, for which purpose and under whose accountability.

Which single field would stop your pilot today? Reply with one line — I read every answer.

I read every single message, and past posts already shaped this space. Reach out to me and tell me what you’re up to and the path you’ve chosen or post online your thoughts about this episode. I read every email and linked post.

Did this resonate with you?

👍 aaaaaaaaa👎

Automate your emails into scheduled tasks without AI

Conduct better interviews with predictable results

Mountain air, city lights or somewhere new? There's still time to make your fall plans take flight with Alaska Airlines. With hundreds of destinations, you can celebrate autumn to the fullest. See the fall foliage in Boston, sip cider in Seattle, or swap your flannel for a swimsuit and head to Hawai‘i. 

With nonstop routes to over 100 cities across North America and beyond, getting to your fall destination is easier than ever. Whether you're chasing changing leaves or year-round sunshine, Alaska’s growing network means more ways to reach where autumn is calling. 

Fares at great prices, so it's easier than ever to say ‘yes’ to a getaway before the season's gone.

Book Now

_{^{Subject to availability. Terms and conditions apply.}}

Government Data Quality Hub. (2020, December 3). The government data quality framework. GOV.UK. https://www.gov.uk/government/publications/the-government-data-quality-framework/the-government-data-quality-framework

ProcureAbility. (2026, September 9). ProcureAbility CPO benchmark study finds procurement expectations outpacing organizational readiness [Press release]. PR Newswire. https://www.prnewswire.com/news-releases/procureability-cpo-benchmark-study-finds-procurement-expectations-outpacing-organizational-readiness-302873422.html

Redman, T. C. (2016, July 28). Assess whether you have a data quality problem. Harvard Business Review. https://hbr.org/2016/07/assess-whether-you-have-a-data-quality-problem

Thank you for your time and trust,

Pascal Hecker | Editor-In-Chief, CPO Path.