Share E-Book
Scan to open this page

Scan with your phone to open this page

Author: Martin Vance

Rating 3.0 AI initial rating

Today, every major corporate initiative, from AI and analytics to regulatory compliance, depends on data quality. As organizations invest in tools, platforms, and leadership roles, progress in data management often stalls. That's because most approaches assume ideal working conditions, while in reality, teams are often struggling with limited time, cultural resistance, and competing priorities. The Anti-Perfectionist Guide to Data Quality enables you to work within that messy reality. Data management expert Martin Vance breaks down the assumptions keeping your data efforts stuck and shows you how to move forward with the resources available. Each chapter pairs a "Perfectionist Trap" with an "Anti-Perfectionist Escape Route" to help you understand real-world strategies for building momentum, earning trust, and demonstrating impact.

AI Reading Assistant

Whole-book reading guide from stratified index samples; jump to passages in the text

AI guide
【One-Line Pitch】 A practical field guide for data practitioners who are tired of chasing flawless datasets and want to build a data culture that actually changes business decisions. Read it if you lead, staff, or depend on data quality work inside a real organization with limited time, skeptical executives, and competing priorities. 【Book Arc】 - **Opening (~0%–10%)**: Sets up the central provocation — the question "how do we get perfect data?" is the wrong question. Vance uses his Formula 1 experience with imperfect, high-volume telemetry to show that teams adapt and decide anyway, and introduces the book's structural pairing of a "Perfectionist Trap" with an "Anti-Perfectionist Escape Route." - **Early (~10%–30%)**: Dismantles the myth of perfect data from several angles: why executives don't care about "data quality" but do care about decision quality, how diminishing returns make the last 5% of cleaning the costliest and least valuable work, and what fit-for-purpose really means (purpose-graded data, thresholds tied to business objectives, not blanket rules). - **Middle (~30%–55%)**: Moves into prioritization and standards. Critical Data Elements (CDEs) as a way to avoid boiling the ocean, the human and behavioral obstacles (artificial harmony, fear of downstream impact, alert fatigue), and the case for governing "boring" reference data well — illustrated by the Mars Climate Orbiter loss as a shared-standard failure rather than a data-accuracy failure. - **Late (~55%–75%)**: The excerpts indicate a transition toward trust, ownership, and advocacy — chapter titles reference rebuilding trust and an "advocacy and ally model" for who owns data — though the sampled material does not cover these chapters in depth. - **Ending (~75%–100%)**: The practitioner's toolkit: finding a first win, getting results with existing resources, making progress visible, and preparing for AI. The excerpts do not cover these chapters' content, only their titles. 【Key Takeaways】 - **Retire the phrase "data quality" with business stakeholders** (Early): Replace it with "decision quality." Executives care about forecasting, pipeline, retention, and margin — not schema validation. Framing data work as decision quality changes the conversation and the budget. - **Fit-for-purpose beats perfection, and it is harder** (Early): Quality only exists relative to a defined use case. Purpose-graded data — campaign-grade, audit-grade, life-grade — demands explicit thresholds, negotiated trade-offs, and continuous measurement. Blanket "clean everything" is the lazy option. - **The last 5% costs the most and delivers the least** (Middle): The house-cleaning analogy makes diminishing returns visceral: one focused hour transforms a messy house from ~20% to ~90% clean; the remaining edge cases require manual, domain-expert effort for marginal gain. This is the single biggest reason data programs burn budget. - **Not all data is created equal — prioritize CDEs** (Middle): A small set of Critical Data Elements, chosen by the business, prevents alert fatigue and the erosion of trust that comes from monitoring everything. Start with three to five CDEs as a proof of concept. - **Prioritization is a human problem, not a taxonomy problem** (Middle): Artificial harmony and engineers' fear of breaking obscure downstream reports stall CDE work. Shift the conversation from classification to business impact, use three tiers, frame it as empowerment, and create explicit psychological safety to let some pipelines fail gracefully. - **Govern the boring data exceptionally well** (Middle): The Mars Climate Orbiter failure was not bad data but a missing shared standard between two internally consistent systems. Reference data agreements protect organizations more than governing exciting, cutting-edge data. - **Momentum and visible wins build the culture** (Late): The book's later chapters promise a launchpad approach — first wins, working with existing resources, making progress visible — though the excerpts do not detail the specific methods. 【Reading Tips】 - Read Chapter 1 closely; it carries the book's core arguments (decision quality, fit-for-purpose, diminishing returns) and the trap/escape-route pattern that repeats in every later chapter. - Skim the standards and regulatory material if you already know ISO 8000 — treat it as ammunition for stakeholder conversations rather than new theory. - When you reach the CDE discussion, pause and draft your own five-field exercise before continuing; the book is designed to be acted on, not just read. - Treat the later toolkit chapters as a checklist for your next 90 days rather than a linear read, since the excerpts suggest they are practice-oriented. - Keep the trap/escape-route pairing in mind as a diagnostic: when a chapter's trap sounds familiar, that is your signal to slow down. 【Coverage Limits】 This guide is based on stratified excerpts covering roughly the first half of the book in detail; the later toolkit chapters (first wins, working with existing resources, visibility, AI preparation) are referenced by title only, so their specific methods are not summarized here.
Excerpt 1
al sales department: 800-998-9938 or corporate@oreilly.com . Acquisitions Editor: Aaron Black Development Editor: Sara Hunter Production Editor: Destiny Bait...
View in text
Excerpt 2
ely different relationship with the concept of “good enough.” Executives Do Not Care About Data Quality The growth lever for an organization is rarely a flaw...
View in text
Excerpt 3
ality data as “portable data that meets stated requirements.”1 The word stated is doing critical work in that definition: quality is measured against explici...
View in text
Excerpt 4
ng: another textbook theory you’ve seen time and time again. And you’re right; it’s a widely used economic theory taught in first-year university courses, an...
View in text
Excerpt 5
t the organization confused purchasing with problem-solving. In Chapter 1 , I walked you through why perfect data is economically irrational, and why the las...
View in text
Excerpt 6
was so fundamentally broken it never launched to the public. The accusations are specific and damning: the data architecture Accenture designed could not sup...
View in text
Excerpt 7
half-billion-euro SAP HANA implementation they called eLWIS. Lidl kicked the project off in 2011 with the goal of replacing its in-house inventory system wit...
View in text
Excerpt 8
exactly the kind of intrusion they were about to experience. The tool was correctly installed, correctly tuned to Target’s network, and — here is the importa...
View in text
Tags
AI categories
DataBig DataTechnology
Publish Year: 2026
Language: English
File Format: EPUB
File Size: 7.6 MB