Le podcast de stratégie data

ETS data breach: one stolen SFTP login exposed student records

27 oct. 2026 · 9 min · transcription

How did one compromised login give an intruder nearly four months inside the SFTP server where California swapped student test data with Educational Testing Service (ETS)? The files held names, test results, disability status, Section 504 status and records on students who left as far back as 2006. The hosts split on blame: ETS for the dwell time, or the state for pooling every child's record at one vendor behind one login. You walk away able to audit your own vendor file drops: name an owner, check whether MFA is on, enforce data minimisation with a scheduled deletion job, and treat contract clauses as no substitute for controls that block a bulk download at midnight.

NAV error threshold: back-solving reconciliation tolerances

20 oct. 2026 · 11 min · transcription

Who pays when a fund's NAV is wrong and the manager never saw the error? Under CSSF Circular 24/856 the investment fund manager corrects and compensates, even when the administrator calculated the price. The episode argues that managers should own one position and pricing record, backed by a written pricing policy, rather than inheriting whatever the administrator used. Listeners learn to back-solve reconciliation tolerances from the 0.5% NAV materiality threshold for fixed income UCITS funds: a tenth of it per break for same-day ownership, a quarter of it in open breaks as a NAV gate. Also covered: the aggregation clause, EU T+1, Invesco's identifier consolidation, circular matching, Bloomberg pricing drift and the SEC's Madison Capital Funding settlement.

Asos hack: why marketing tools hold your Snowflake keys

13 oct. 2026 · 10 min · transcription

How did hackers send a ransom note through the real Asos app? Reporting points to a conned employee and, per researcher Kevin Beaumont, Simon AI, a composable customer data platform owned by Monetate that reads straight from Snowflake with MFA optional and off by default. That route is unconfirmed. The episode argues the real failure is ownership: marketing bought it, data engineering wired it to the warehouse, and nobody owned the login. You leave with a concrete audit: list every third-party tool holding warehouse credentials, check whether MFA is enforced or only available, check whether it can message customers, and fix the risky ones with mandatory MFA or single sign-on before Friday.

dbt State and agent-ready data: why declared meaning comes first

13 oct. 2026 · 11 min · transcription

Agents dominated dbt Summit, but does smart orchestration or a context layer make data agent-ready on its own? The episode argues no. Fanatics Betting and Gaming saw about 0.2% model reuse from dbt State until it declared source freshness and model-level SLAs, which raised reuse to roughly 15%. The same pattern runs through Fivetran's Context Layer, AI audit trails and token costs. Listeners learn how dbt State is billed per daily active target table, including tests, why the Context Layer's beta status needs checking in writing, how git history serves as an audit trail, and how dbt models served through the MCP server cut Gong transcript tokens by 20x or more.

ARTEX bank breaches: why side-door data access is a CDO issue

9 oct. 2026 · 10 min · transcription

How did an attacker using ARTEX, an open-source agentic AI penetration-testing tool, take data on more than 67,000 customers from seven South Korean firms, including Shinhan, KB Kookmin and Hana, without touching core banking? The episode argues the failure was data access design. Broker portals and employee apps returned income, loan limits and resident registration numbers to callers who did not need them, despite nearly 124 billion won in security spending. You come away knowing what the Financial Services Commission ordered, how CrowdStrike linked the attack to Claude Code and DeepSeek, and why basic cyber hygiene matters more than an AI defense product. You also get a concrete plan to inventory every externally reachable system that returns customer data and remove unneeded fields.

UK GDPR Article 21: 44,000 objections hit NHS Palantir

6 oct. 2026 · 9 min · transcription

What happens when a data platform is legally defensible, operationally embedded and publicly distrusted at the same time? More than 44,000 people have filed Article 21 objections under UK GDPR against NHS England's Palantir-powered Federated Data Platform, live in 142 trusts, weeks before a February 2027 contract break clause comes into reach. The position here: objection handling is a throughput problem, not a policy document. You come away able to pressure-test your own platform. Check the published benefits claims, 117,000 extra operations, a 14.3% cut in discharge delays, against Foxglove FOI data showing 13 of 41 trusts doing fewer operations, and name the owner and service level for mass objections before someone else does.

dbt State: lag_tolerance is now a budget decision

6 oct. 2026 · 9 min · transcription

dbt State went generally available in September 2026, so rebuilding every model on a cron schedule is now a choice rather than a default. The position here: the tool cannot tell you how stale each model is allowed to be, and the lag_tolerance config most teams set once and globally is a business decision about decision cadence, not an engineering setting. You leave with a method: meter wasted rebuilds using BigQuery INFORMATION_SCHEMA.JOBS or Snowflake credit consumption, tie tolerance to the decision each model feeds, keep source freshness checks running, read the Explain tab in audit mode, and cap agent queries with resource monitors and maximumBytesBilled. Includes the RxBenefits 59% result, dbt's 15 to 30% range, FinOps Foundation and Flexera 2026 findings.

SB 947: proving who reviewed an AI firing call

2 oct. 2026 · 10 min · transcription

California's No Robo Bosses Act bars employers from disciplining or firing a worker on an automated system's output alone, and from 1 July 2027 a human must independently corroborate that output in writing. The position here is that this reads as an HR matter but lands on data teams, because the statute covers statistical modeling and data analytics, which means attrition models and attendance scores, not only frontier AI. You come away knowing what SB 947 and SB 951 demand, how Cal-WARN notices must now name AI-driven displacement, what the Chamber of Commerce and Senator McNerney argued, why the SB 7 inventory clause was dropped, and how to audit every system scoring a named employee.

Churn models in telecom: why uplift beats accuracy

1 oct. 2026 · 5 min · transcription

Why do churn models produce accurate predictions that never move retention numbers? The position here is that most telcos predict churn the week it happens, when the customer has already price-shopped and gone. Telefonica's work came from years of accumulating unglamorous signals such as degraded data speed at a single cell tower, payment timing drift and the gap between plan and actual usage, combined with contract stage and competitor pricing by postcode. You walk away able to separate accuracy from uplift, size the movable middle, defend a holdout control group to a revenue VP, and wire a score into a pre-approved retention offer the agent can act on. References include MIT Sloan Management Review on unactionable analytics and dbt Labs on data cleaning time.

AI slop detectors: why filtering training data backfires

30 sept. 2026 · 5 min · transcription

Should you strip suspected AI-generated text out of a training set? The position here is no, not blind. Slop detectors are classifiers that flag one to five percent of clean human writing as fake, and they misfire on dense expert prose and standardized legal boilerplate, so teams in financial services and legal have deleted the backbone of their corpus and watched accuracy fall. A recent experiment found filtering degraded performance more than the contamination removed. You leave with a Monday plan: quarantine and hand-sample flagged items rather than deleting, tag provenance at ingest instead of at training time, and read the unknown-provenance estimates from dbt Labs and MIT Sloan Management Review with the vendor incentive in mind.

Cross-agency data: why 70% of failures are semantic

29 sept. 2026 · 5 min · transcription

When two government agencies define "household" differently, one by mailing address and one by shared cooking facilities, no integration platform reconciles the numbers. This episode argues that roughly 70 percent of cross-agency integration failures trace to the semantic layer rather than the plumbing, and that the median time to reach a trusted cross-agency data agreement is 18 months. You get a worked example from benefits coordination, where food assistance uses an economic household and Medicaid often uses the tax household, plus the trap of mapping household_id to household_id. You leave with a sequencing rule for governance spend, a named-owner test, and a three-entity definition exercise. References include MIT Sloan Management Review and dbt Labs.

Amazon's data flywheel: twelve years of engineering

27 sept. 2026 · 5 min · transcription

Does more data automatically compound into advantage? This episode argues it does not. Drawing on an MIT Sloan Management Review breakdown tracing Amazon's flywheel to its origins, the discussion shows the compounding came from roughly twelve years of feedback loop design, architecture and organizational choices, including teams owning data as a product with an internal consumer they answered to. You come away with a test for your own setup: time to feedback, meaning how long from a customer action until that signal changes what the next customer sees. You also get a check on modern data stack claims from tooling vendors such as dbt Labs, and a working definition of data monetization measured on the P&L.

Boeing component traceability: pay for the birth record

26 sept. 2026 · 5 min · transcription

When a 737 MAX fastener is installed without a birth record, who owns the defect? The answer is the assembler, not the tier-3 shop that made the part. This episode takes apart Boeing's multi-year traceability program, the digital genealogy record that follows one component from raw metal to installed part, and why three-quarters of the cost sits in supplier process rather than software. DBT Labs' claimed 30% reconciliation tax is weighed against MIT Sloan Management Review's manufacturing data work. You leave knowing how to write supplier clauses that pay for clean records and penalize broken ones, how to scope traceability to your five most recall-critical parts, and why orphaned part numbers break chains of custody.

Semantic layers: one revenue number for the board

25 sept. 2026 · 5 min · transcription

What happens when finance, sales and product each compute revenue their own way? The episode opens on a board meeting where three presenters defend three different quarterly numbers, and argues that the cost is not bad data but frozen decisions while teams re-litigate whose spreadsheet counts. Finance recognizes revenue on fulfillment, sales on signature, product on ship date, and each team's bonus depends on its version. You come away with a build sequence for a semantic layer: inventory the 12 to 15 metrics that appear in board decks and budget fights, lock each to one written formula with finance and sales signing off together, then choose tooling. The discussion weighs a dbt Labs vendor claim against MIT Sloan Management Review's work on trust in shared numbers, and covers churn, customer acquisition cost, and the CDO's refereeing role.

Identity resolution in pharma: one HCP, six records

24 sept. 2026 · 5 min · transcription

Why does a single cardiologist show up as six separate entities across a pharma company's CRM, purchased claims data and analytics warehouse, and what does that cost? The position here is blunt: identity resolution is not a tool purchase, it is plumbing. Bad matching splits one high prescriber into six low ones, which skews sample allocation, wastes temperature controlled product, and can hide adverse event patterns in pharmacovigilance reporting that regulators will not excuse. You get a working approach: make the federal NPI registry the source of truth, run nightly matches, write down survivorship rules, and name one accountable owner with budget for the prescriber master record. References include DBT Labs and MIT Sloan Management Review, plus a master data example at Novartis.

AI agents querying your warehouse: measure before rebuilding

23 sept. 2026 · 5 min · transcription

Fivetran and dbt Labs used dbt Summit to argue that AI agents, not human analysts, are now the main consumer of enterprise data. The position here is that the architectural shift is real but the urgency is a sales pitch, and that re-platforming off a keynote is the expensive mistake. You come away able to test dbt Labs' claim that 30% of warehouse queries are machine-generated against your own logs, spot the failure mode where agents pass nonsense results downstream with no scar tissue, make the case for a written semantic layer over new tooling, and set a query budget before an agent touches production. References include O'Reilly Radar and MIT Sloan Management Review.

AML pipelines: three design calls examiners test

22 sept. 2026 · 5 min · transcription

Why do fintech AML pipelines fail regulatory examination when the models themselves are sound? The position here is that examiners do not score area under the curve, they ask reproducibility questions, and most data architectures cannot rebuild a March alert with March data. Three design decisions decide the outcome: immutable timestamped feature storage, documentation that records who approved a feature and which regulation it maps to, and explanation captured at inference time rather than manufactured later. You walk away knowing why dbt column level lineage maps transformations but not business intent, why SHAP values may be plausible rather than faithful, and how to run a one hour reproduction test on your last high value alert.

DuckDB and text-to-SQL: redesign the analytics team

21 sept. 2026 · 5 min · transcription

If a manager can type "show me last quarter's churn by region" and get an answer in ten seconds, what is the analytics team for? The position here is that DuckDB plus language models that write SQL is a workforce redesign, not a tooling upgrade. The ad hoc request queue goes first, dashboard building is half gone, and the analytics engineer who owns the semantic layer becomes the centre of the team. You get a concrete redesign of a team of eight, the reason a model will compute churn three contradictory ways, the caveat on dbt Labs and O'Reilly survey numbers, and a Monday test: count how many live definitions exist for your five most quoted metrics.

Grey market leaks: Richemont's serial number fix

20 sept. 2026 · 5 min · transcription

Genuine Cartier and IWC pieces were surfacing in unauthorised Asian markets at 20 to 35 percent discounts, and the villain was Richemont's own distributors. The position taken here is blunt: anti-counterfeiting gets the attention, but the grey market is the larger financial wound, and distribution integrity is a data governance problem wearing a legal costume. You get the sequence Richemont followed to build a product-level data layer where each serial number is a live record updated at every handoff, the 40-odd maison fight over one shared definition of a serial event, and why enforcement worked as deterrence rather than analytics. Referenced sources include dbt Labs on 18-month data modeling rebuilds, O'Reilly Radar and MIT Sloan Management Review on ownership failures.

Data flywheel: why loop latency beats data volume

19 sept. 2026 · 5 min · transcription

Board decks show a circular arrow labelled data flywheel, but few teams can say what spins it. This episode separates compounding from speed: a faster pipeline processes the same low value data quicker, while a real flywheel makes each new record more valuable because of the data already held, as Google search ranking does with click data. It examines the dbt Labs claim that pipeline automation cuts time to insight by around 60 percent, plus findings from MIT Sloan Management Review on flattening accuracy gains and O'Reilly Radar on warehouses that never feed decisions. You leave with three failure modes to test for, data saturation, data decay and a loop that never reaches the product, and a single metric to measure this week: the days between data arriving and a decision changing.

MES and ERP integration: the real Siemens Amberg win

18 sept. 2026 · 5 min · transcription

Siemens Amberg is usually told as a robotics story, with 99.99885% quality and a heavily automated line. The position here is different: the recent gains came from making the Manufacturing Execution System and ERP agree on reality, not from more sensors. The route was treating manufacturing data as a product with owners, contracts and quality checks, rather than chasing one monolithic platform. You get the specific failure mode to look for, where MES counts a unit complete when it leaves the line and ERP counts it when booked to inventory, and what that gap does to supply forecasts. References include MIT Sloan Management Review, O'Reilly Radar and dbt Labs, plus a first project any CDO can start with a meeting.

ML in production: why models die after the notebook

17 sept. 2026 · 5 min · transcription

Why do most machine learning projects never leave the lab? The position here is blunt: the model is about ten percent of the work and the other ninety percent is pipelines, monitoring and ownership. The episode walks through the three failure points at the notebook to production handoff, data drift, unowned pipelines after launch, and systems never designed for retraining, and why a drifting model costs money quietly instead of paging you at midnight. You come away able to name the roles that fix this, the ML engineer and the data engineer, read vendor claims from DBT Labs against O'Reilly Radar and MIT Sloan Management Review, apply data as a product, and ask the accountability question before funding an AI project.

Data literacy: measure decision quality, not training

16 sept. 2026 · 5 min · transcription

Why do data literacy programs teach medians, filters and chart types and still leave the business running on gut feeling? The position here is blunt: knowledge and behavior are different things, and almost nobody measures the second one. The fix is to build training around real recurring decisions like pricing, hiring and inventory, and to require anyone proposing a decision to state in advance what evidence would change their mind. You come away with a way to score decision quality across ten recurring decisions before and after a program, a read on Booking.com's experimentation default, and evidence from MIT Sloan Management Review, O'Reilly Radar and DBT Labs on why culture and trust, not skills, are the bottleneck.

GDPR retention and minimization: the real compliance gap

15 sept. 2026 · 5 min · transcription

Is a consent banner and a privacy notice enough to call a GDPR program finished? The position here is no. Consent management got budget because customers see it, while retention and data minimization stayed invisible and unfunded. Modern data tooling in 2026, including lineage work pushed by vendors like dbt Labs, now makes every copy of a customer email visible, and research from O'Reilly and MIT Sloan Management Review points to personal data being kept with no defined deletion schedule. You walk away able to treat unused personal data as a liability, defend against data subject access requests, hash or aggregate identifying fields at ingestion, and set a retention schedule on your largest personal data table.

Alternative credit scoring: how Tala lends without bureaus

14 sept. 2026 · 5 min · transcription

Can you underwrite a loan for someone who has no credit bureau file at all? Tala says yes, and the episode explains how: permissioned smartphone data such as contact counts, airtime top-up timing and money movement frequency, used as behavioral proxies for the habits a bureau score normally captures. The position taken here is that Tala's edge was never a cleverer algorithm. It was owning a data source nobody else collected, a point MIT Sloan Management Review has argued for years. You come away knowing how model drift erodes a live lending model, why retraining and freshness tests matter more than model choice, and the limit of behavioral data once borrowers start performing for the algorithm. It closes with a one-week audit of your three most important data sources.

Luxury waitlist allocation: building a defensible scarcity model

13 sept. 2026 · 5 min · transcription

Who decides that a Hermes Birkin buyer waits six years? The position here is blunt: scarcity is an asset a data team manages, and every "nothing available" is an allocation decision someone modeled. The episode separates manufactured scarcity from the allocation logic that ranks clients by lifetime spend, purchase breadth and loyalty, and looks at the US class action arguing that tying Birkin access to other purchases is illegal bundling. You come away with a three-layer design, scarcity forecast, eligibility stripped of protected proxies, and allocation ranking, plus why transformation tools like dbt give you lineage as a legal shield. Also covered: the Patek Philippe Nautilus 5711 discontinuation and reading waitlists as intent data for product planning.

Mistral's $3.5B raise: why open weights favor portability

13 sept. 2026 · 5 min · transcription

Mistral's September 2026 raise of $3.5 billion made open-weight models a credible enterprise option, and most data leaders read that as a signal to stop buying and start building. The position here is the opposite: capability parity raises the ceiling, not the floor, and the hard part was never the model. It is the data plumbing and the retrieval layer, a point O'Reilly's 2026 adoption research supports. You get the three conditions that justify self-hosting, the token volume crossover point DBT Labs published (with the caveat that they sell data tooling), and the MIT Sloan Management Review argument that open weights are worth having for optionality. You leave with a lock-in audit to run on your top three AI use cases.

Decision intelligence: engineer the decision, not the dashboard

11 sept. 2026 · 5 min · transcription

Why do organisations with hundreds of dashboards still decide badly? The answer here is that the numbers arrive nowhere near the moment of choice. A support agent deciding on a refund needs lifetime value, refund history and churn risk inside the ticket, not in a tool they have no login for. Embedded analytics is the plumbing; decision intelligence is treating the recurring decision as the thing you design around, then working backwards. You walk away with three tests for picking a decision worth engineering (frequent, consequential, currently made blind), a rule for what to leave alone, and a warning on reading DBT Labs vendor figures, with supporting points from O'Reilly Radar and MIT Sloan Management Review.

GDPR versus GxP audit trails: the deletion trap

10 sept. 2026 · 5 min · transcription

Pharma data sits under three rulebooks at once: GxP quality regulation, the FDA's 21 CFR Part 11 electronic records rule, and GDPR. The first two say keep everything and prove everything. The third says erase on request. This episode takes the position that the damage happens where they conflict, when a right to be forgotten workflow reaches into a Part 11 audit trail and makes it look modified to an inspector. You come away able to audit your own deletion path, apply pseudonymization with a separately held key, use the GDPR research exemption by design, and argue why the chief data officer should own all three regimes. References include MIT Sloan Management Review and dbt Labs.

Open data portals: publish fewer datasets as products

9 sept. 2026 · 5 min · transcription

Government portals hold thousands of datasets that get downloaded once and never opened again. The position here is that dataset count is a vanity metric and repeat use is the real measure of a portal's health, citing Data.gov's 300,000 plus datasets and their dead long tail. The answer is to treat a dataset like a product with an owner, a release schedule and documentation, and to retire the rest. You will learn how to separate the needs of journalists, advocates and oversight bodies, why Transport for London's live feed produced CityMapper and tens of millions of pounds in value, how to read public records requests as a demand signal, and why a plain language data dictionary decides whether a dataset gets cited.

Fanatics data platform ROI: making the board care

8 sept. 2026 · 5 min · transcription

How does a chief data officer get a board to fund a data platform instead of tolerating it? The answer here is attribution. Fanatics stopped presenting infrastructure cost and presented measurable business output, treating the platform as a profit and loss line. The episode walks through the use cases they attached numbers to, including pricing, inventory markdowns and fraud detection during big jersey drops, and flags that the widely cited dbt Labs figures come from a vendor and need cross-checking against sources like MIT Sloan Management Review. You finish able to pick one money decision, price the delta your data creates, and present it as a conservative, likely and optimistic range rather than a single hero number a CFO can dismantle.

Load forecasting: why DER signals break utility models

7 sept. 2026 · 5 min · transcription

Utility teams still treat load forecasting as a regression against temperature and industrial schedules. This episode argues that world is gone. Rooftop solar can flip the sign of the weather relationship, a passing cloud can return 40 megawatts in 90 seconds, and a midnight EV rate discount can synchronize 10,000 chargers into a spike that never existed before. Bolting DER signals onto the old model produces a more complicated wrong answer. You come away knowing where the money leaks, through NERC reliability reports and rate cases where a 2% error becomes a nine-figure conversation, why endogeneity breaks price-sensitive models, what probabilistic forecasting with confidence bands gives operators, and why dbt Labs style version control keeps models auditable without telling you they are wrong.

ELT stack: where dbt, ingestion and orchestration break

6 sept. 2026 · 5 min · transcription

How do the three layers of a modern ELT stack actually connect, and why do pipelines still produce wrong numbers when every tool in them is good? The position here is that ingestion, dbt and orchestration are each mature, but the joints between them are untested, and that is where bad data comes from. You get a plain definition of extract, load, transform and why cheap storage in Snowflake, BigQuery and Databricks flipped ETL around. You also get the cost picture on Fivetran monthly active rows versus open source Airbyte, what dbt adds through version control, tests and documentation, what Airflow and Dagster decide about run order, and a one hour test to catch upstream schema changes.

Data ROI to the board: why attribution models backfire

5 sept. 2026 · 5 min · transcription

Why do chief data officers present cost savings and revenue attribution to the board and still lose budget? The position here is that the problem is the mental model, not the numbers. Attribution percentages like "our model drove 12% of pipeline" invite the board to argue about credit, and manufactured precision reads as hiding rather than rigour. Cost savings is the language of a department fighting for survival. You walk away able to reframe the conversation around decisions improved and bets de-risked, using a retailer example where a demand model caught a forecast inflated by 30% before a nine-figure inventory commitment, and able to read Snowflake and Databricks ROI calculators with proper scepticism.

Data contracts: how JPMorgan Chase named data owners

4 sept. 2026 · 5 min · transcription

Who owns a dataset, and what happens when it breaks? JPMorgan Chase could not answer that for years across hundreds of business lines, so it replaced governance policy documents with data contracts covering more than 50 domains, each with a named person who signs off on format, quality, freshness and escalation. The position here: accountability without a name attached is theater. You get the design that avoids organizational paralysis, contracts strict at the core and light at the edges, where consumers subscribe to a published interface instead of renegotiating each request. You also get the failure mode being avoided, silent schema changes, the lineage case regulators care about, a caution on the vendor-quoted 60% cleanup figure from Monte Carlo and Anomalo, and a one-page contract to write this week.

Data literacy programs: change decisions, not courses

3 sept. 2026 · 5 min · transcription

Does data literacy training change how a business decides anything? The position here is blunt: around 80 percent of programs teach people to pass a quiz, and Gartner surveys keep finding chief data officers who cannot name a business metric their literacy effort moved. Vocabulary like statistical significance is not the same as a category manager killing a promotion because the lift sits inside the margin of error. You come away with a working method: pick the recurring decisions, baseline them, and put the number in front of the decision maker at the moment of choice. Examples include a European retailer's delisting rule worth low single-digit millions a year, and a bank that halved model overrides by requiring written justification.

Zero-trust data access: what CDOs need to check first

2 sept. 2026 · 5 min · transcription

Zero-trust appears on every vendor slide, but most explanations stop at the network perimeter. This episode asks what the model means at the data layer, where a Chief Data Officer actually works, and argues that zero-trust is a posture you maintain rather than a product you buy. The position is blunt: classification comes before technology, and access rules written without honest data tiering create friction people route around. Listeners come away able to separate network zero-trust from data zero-trust, apply the three checks at the moment of access (identity, context, data sensitivity), tier data as public, internal, confidential or regulated, and run a five-dataset access audit. Includes Microsoft Purview figures on unlocated sensitive data and an insurer rollout that failed.

Master data management: how Salesforce made MDM stick

1 sept. 2026 · 5 min · transcription

Can a company that sells data quality run on fragmented records of its own? Salesforce did for years, inheriting duplicate customer and product records from Tableau, Slack, MuleSoft and Demandware until the same account appeared five times across five systems and quietly inflated its own pipeline numbers. The position here is that master data management fails the moment it is funded like a project with a ribbon cutting. You come away with the three moves Salesforce used: name one accountable owner per data domain, sequence the revenue critical records first, and embed duplicate checks at the point of entry. Also covered: Gartner's $13 million poor data quality estimate, the 10 to 1 prevention versus remediation ratio, and why large language models amplify bad master data rather than fix it.

Feature stores and the ML data supply chain: what CDOs actually need to understand

30 août 2026 · 5 min · transcription

Feature stores sit at the intersection of data engineering and machine learning operations, yet most organizations treat them as a tooling decision rather than a strategic one. This article explains how they work, why the architectural choice matters at the CDO level, and where the tradeoff between standardization and flexibility bites hardest.

Pricing and packaging a data product for external revenue

29 août 2026 · 4 min · transcription

Most organisations that decide to monetise their data externally know what data they have, but stumble badly on how to price and package it. This article breaks down the mechanics of data product pricing: what actually drives willingness to pay, how to structure tiers, and where the common traps are.

Data observability: catching bad data before it reaches decisions

28 août 2026 · 5 min · transcription

Bad data doesn't announce itself. This playbook shows CDOs how to build detection mechanisms that intercept data quality failures before they corrupt reports, models, and the decisions that follow.

Hub-and-spoke data teams: the model that sounds right and works badly

27 août 2026 · 5 min · transcription

Hub-and-spoke has become the default answer when CDOs are asked how to balance central governance with business-unit agility. The reality in most organisations is slower decisions, diluted accountability, and data professionals caught between two bosses with conflicting priorities.

The data flywheel field guide: who built compounding advantage and what they actually did

25 août 2026 · 5 min · transcription

The data flywheel is one of the most cited concepts in data strategy, and one of the least examined in practice. This field guide cuts through the abstraction and names the companies and moments that show what compounding data advantage actually looks like when it works.

Privacy-enhancing technologies in practice: a CDO playbook

24 août 2026 · 5 min · transcription

Privacy-enhancing technologies have moved from research papers to production deployments, and CDOs who treat them as theoretical still carry unnecessary legal and competitive risk. This playbook walks through how to select, sequence, and embed PETs into your data architecture without stalling your analytics programme.

How Walmart proved data ROI to its board: lessons from a $1 billion bet on supply chain intelligence

23 août 2026 · 5 min · transcription

Walmart's decision to invest heavily in data infrastructure and analytics for its supply chain gave its board a concrete, measurable case for data spending. The mechanics of how that case was built, and what CDOs at other organisations can borrow from it, are more instructive than the headline numbers.

Natural-language BI and the analyst's new role: why the "democratisation" story is only half true

22 août 2026 · 5 min · transcription

Natural-language query tools promise to put business intelligence in everyone's hands, removing the analyst bottleneck. The reality is more complicated, and CDOs who act on the simple version of this story will make costly structural mistakes.

How Walmart's CDO built credibility and structure in the first 90 days

21 août 2026 · 5 min · transcription

When Walmart reorganized its data function in the early 2020s, the incoming data leadership faced a familiar problem: scattered ownership, competing priorities, and a business that wasn't sure what to expect from a CDO. The choices made in those first three months set the terms for everything that followed.

Clean rooms in practice: a CDO's playbook for data collaboration without the risk

20 août 2026 · 5 min · transcription

Data clean rooms promise the ability to share audience insights across company boundaries without exposing raw data. Here is a concrete sequence for CDOs who want to move from pilot anxiety to production-grade collaboration.

Internal data products: a field guide to the teams and companies worth studying

19 août 2026 · 5 min · transcription

Platform thinking for internal data is no longer a theoretical aspiration, a small group of companies have built the real thing and their choices reveal what actually works. This field guide identifies the most instructive players, ranked by documented influence on how the industry thinks and builds.

Proving data ROI to the board: the contribution margin method explained

18 août 2026 · 5 min · transcription

Most CDOs struggle to quantify the value of data investments in terms a CFO will accept. This article explains one specific method, contribution margin attribution, that makes the case in the language boards actually use.

Real-time streaming data: a CDO playbook for getting it right

17 août 2026 · 4 min · transcription

Most organizations collect streaming data but few actually act on it fast enough to matter. This playbook gives CDOs a concrete sequence for building real-time data capability that delivers operational value, not just architectural complexity.

The data flywheel: how compounding data advantage actually works

16 août 2026 · 5 min · transcription

The data flywheel is one of the most cited concepts in AI strategy and one of the least understood in practice. This article breaks down the actual mechanics so that CDOs can assess whether their organization is genuinely building one or just accumulating data.

How JPMorgan Chase built data contracts across 50+ domains

15 août 2026 · 5 min · transcription

JPMorgan Chase's data mesh initiative forced the bank to confront a problem most large organizations prefer to defer: who actually owns a data product, and what obligations come with that ownership? Their approach to data contracts offers a detailed, replicable model for CDOs managing complex, federated data environments.

Privacy-enhancing technologies in practice: the hype is ahead of the implementation

14 août 2026 · 5 min · transcription

Privacy-enhancing technologies have generated serious boardroom attention, and the underlying science is real. But the gap between pilot programs and production-grade deployment is wider than most CDOs are being told.

Proving data ROI to the board: the business value bridge method

13 août 2026 · 4 min · transcription

Most CDOs can measure data programme activity. Far fewer can connect it to the numbers a CFO or board chair actually cares about. The business value bridge changes that.

Operationalizing the EU AI Act for data teams: a practical playbook

12 août 2026 · 5 min

The EU AI Act's phased enforcement schedule is already creating compliance obligations for data teams, with high-risk system requirements fully applicable from August 2026. This playbook walks CDOs through the concrete steps to build an operational response, not just a policy document.

How Airbnb rebuilt metric consistency with a semantic layer

11 août 2026 · 5 min

Airbnb's analytics teams were producing conflicting numbers for the same business questions, undermining trust in data across the company. Their response, building Minerva, a centralised semantic layer, offers a precise and transferable blueprint for CDOs dealing with the same problem.

Data literacy programs that change behavior: a CDO's execution playbook

10 août 2026 · 5 min

Most data literacy programs produce certificates, not decisions. This playbook shows CDOs how to design and run programs that visibly shift how people work with data, from the shop floor to the executive committee.

Proving data ROI to the board: a CDO's execution playbook

9 août 2026 · 4 min

Most boards will fund a data initiative once. Getting continued investment requires a financial narrative built on concrete numbers, not technology enthusiasm. This playbook shows CDOs exactly how to build and deliver that case.

Feature stores and the ML data supply chain: what CDOs actually need to understand

8 août 2026 · 5 min

Feature stores solve a problem that most organisations discover too late: the painful gap between raw data and production-ready ML inputs. Understanding the mechanics and honest tradeoffs is essential before committing to one.

How Cloudflare rebuilt its data stack around dbt, Fivetran, and Airflow

7 août 2026 · 5 min

Cloudflare's rapid growth exposed the limits of hand-coded SQL pipelines and fragmented ingestion scripts that no engineer wanted to touch. This case study traces how the company restructured its analytical data layer using a modern ELT approach, and what that shift actually required in practice.

Model monitoring, drift detection, and retraining triggers: a CDO playbook

5 août 2026 · 5 min

Production ML models degrade silently, and most organizations only notice when business outcomes have already suffered. This playbook gives CDOs a concrete sequence for detecting drift early, deciding when to retrain, and building the governance structure that makes both systematic.

Privacy-enhancing technologies in practice: a CDO playbook

4 août 2026 · 5 min

Privacy-enhancing technologies have moved from cryptography research papers into production pipelines at major financial institutions and healthcare networks. This playbook gives CDOs a concrete sequence to deploy PETs without stalling analytics programmes or exposing the organisation to regulatory backlash.

Clean rooms in practice: a CDO playbook for data collaboration that actually works

3 août 2026 · 5 min

Data clean rooms offer a principled path to collaborative analytics without exposing raw customer data, but most implementations stall on governance gaps and misaligned incentives. This playbook gives CDOs a concrete sequence to stand up a clean room partnership, avoid the common failures, and extract value quickly.

Feature stores and the ML data supply chain: what CDOs actually need to understand

2 août 2026 · 5 min

Feature stores solve one of the most expensive and least visible problems in enterprise ML: the repeated, inconsistent transformation of raw data into model-ready inputs. Understanding how they work, and when they are worth the investment, is now core CDO territory.

Building a semantic layer for consistent metrics across business units

1 août 2026 · 5 min

When Finance reports one revenue number and Sales reports another, the problem is rarely the data itself. This playbook shows CDOs how to build a semantic layer that enforces metric consistency across every business unit, without requiring a full data warehouse overhaul.

How Uber built its ML data supply chain: lessons from the Michelangelo feature store

31 juil. 2026 · 5 min

Uber's Michelangelo platform forced the company to confront a problem most ML teams hit eventually: the same features being rebuilt repeatedly by different teams, with no shared infrastructure underneath. The decisions Uber made in 2017 and 2018 still define how serious organisations think about feature stores today.

The CFO who almost killed the data team (and what saved it)

30 juil. 2026 · 4 min

In 2019, a major UK retailer's data function came within one budget cycle of being dismantled entirely because it couldn't show a single number the CFO believed. What rescued it is a story every CDO should know by heart.

Feature stores and the ML data supply chain: a CDO's execution playbook

29 juil. 2026 · 5 min

Most ML projects stall not because of model quality but because data preparation is reinvented from scratch every time. This playbook gives CDOs a concrete sequence for building a feature store and treating ML data as a managed supply chain.

The modern ELT stack explained: dbt, ingestion, and orchestration working together

28 juil. 2026 · 5 min

The shift from ETL to ELT reshaped how data teams build pipelines, but the real complexity lies in understanding how the three layers, ingestion, transformation, and orchestration, actually interact. This article breaks down the mechanics of the modern stack with concrete examples, and explains where the genuine tradeoffs sit for leaders making architecture decisions.

GDPR beyond consent: a CDO playbook for retention and minimization

27 juil. 2026 · 5 min

Most organizations fixed their consent banners years ago and assumed that was the hard work done. Retention schedules and data minimization remain the two most frequently cited GDPR violations in supervisory authority enforcement, and closing that gap requires a deliberate operational program, not just a policy document.

How JPMorgan Chase built data contracts across 50+ domains

26 juil. 2026 · 5 min

JPMorgan Chase spent years wrestling with data inconsistencies across hundreds of business lines before committing to a structured data contract framework. The mechanics they chose, and the organizational friction they encountered, offer a practical blueprint for CDOs facing the same ownership vacuum.

Data clean rooms explained: what they actually do and when they're worth the effort

25 juil. 2026 · 4 min

Data clean rooms allow organisations to collaborate on sensitive datasets without either party exposing the raw data. For CDOs weighing privacy-preserving analytics against operational complexity, understanding the mechanics matters before signing any partnership agreement.

Cutting cloud data costs without breaking analytics

24 juil. 2026 · 5 min

Cloud bills for data infrastructure have become one of the fastest-growing line items in enterprise IT budgets, and most organizations are overpaying without realizing it. This playbook gives CDOs a concrete sequence of moves to reduce spend significantly while keeping analytical capability intact.

Feature stores: the missing infrastructure layer in your ML data supply chain

23 juil. 2026 · 5 min

Most ML failures are not model failures. They are data supply chain failures, and feature stores are the architectural response that serious ML organizations have adopted to fix them.

Why most data culture initiatives fail before they start

22 juil. 2026 · 4 min

Most organizations have data strategies on paper and data silos in practice. The gap between the two is rarely a technology problem.

When the AI model is wrong: what CDOs must own in 2026

21 juil. 2026 · 4 min

Most AI failures in production are not model failures. They are governance failures, and CDOs who treat the two as interchangeable are building on unstable ground.

When privacy becomes a board-level liability: what CDOs need to own in 2026

20 juil. 2026 · 5 min

Data privacy is no longer a compliance checkbox managed by legal teams. CDOs who treat it as an operational afterthought are accumulating risk that will eventually surface at the worst possible moment.

From data asset to data product: what CDOs get wrong about monetization

19 juil. 2026 · 5 min

Most organizations sitting on valuable data fail to monetize it not because the data is poor, but because they treat productization as a technical problem rather than a commercial one. This article examines what separates data product leaders from laggards, and what CDOs need to change operationally to close the gap.

Data governance in 2026: why compliance alone is failing CDOs

18 juil. 2026 · 5 min

Most organizations have data governance frameworks on paper. The ones that actually work have something different, and it has less to do with regulation than with how governance is wired into daily decision-making.

Self-service analytics is failing most organizations, and CDOs are partly to blame

17 juil. 2026 · 5 min

Most self-service analytics programs deliver far less than promised, with adoption stalling and shadow IT filling the gaps. The problem is rarely the technology.

Data mesh vs. data lakehouse: what CDOs actually need to decide in 2026

16 juil. 2026 · 4 min

The debate between data mesh and data lakehouse architectures has moved past theory and into boardroom budget conversations. Here is what the choice actually involves, and why framing it as an either/or question is the first mistake most CDOs make.

Why data culture fails before strategy does

15 juil. 2026 · 4 min

Most data initiatives stall not because of missing technology or budget, but because the organization never genuinely changed how it thinks about data. For CDOs, this is the defining operational challenge of 2026, and it demands a different playbook than the one most executives were handed.

When AI strategy becomes infrastructure: what CDOs must own in 2026

14 juil. 2026 · 4 min

Most organizations have moved past the question of whether to invest in AI. The real pressure on CDOs now is deciding which decisions to own, which to delegate, and how to build the underlying data infrastructure that makes AI something other than a series of expensive experiments.

When privacy becomes a liability: what CDOs must own in 2026

13 juil. 2026 · 4 min

Data privacy has moved from compliance checkbox to board-level risk, and the CDO is increasingly the executive expected to own it. Here is what that shift looks like in practice and what it demands of your operating model.

Data products that generate revenue: what separates strategy from wishful thinking

12 juil. 2026 · 4 min

Most organizations claiming to "monetize their data" are selling reports, not products. This article examines what genuine data product commercialization requires, where CDOs consistently misjudge the path, and what separates the organizations actually generating revenue from those still in pilot mode.

Data governance under pressure: what CDOs must get right in 2026

11 juil. 2026 · 4 min

Regulatory pressure, AI proliferation, and cross-border data flows have turned data governance from a compliance checkbox into a board-level strategic concern. CDOs who treat it as infrastructure rather than policy will be the ones still standing when the audits arrive.

Self-service analytics at scale: what CDOs keep getting wrong

10 juil. 2026 · 4 min

Most organizations have deployed self-service BI tools. Far fewer have made them work. Here is what separates the ones that do.

Data mesh vs. data lakehouse: what CDOs actually need to decide in 2026

9 juil. 2026 · 4 min

The debate between data mesh and lakehouse architectures has moved beyond whitepaper theory into real organizational consequences. CDOs who treat this as a purely technical choice are already behind.

Why most data culture initiatives fail before they start

8 juil. 2026 · 4 min

Most organizations invest in data tools and governance frameworks while leaving the harder problem untouched: the human behavior that determines whether any of it works. This article examines why data culture efforts stall, and what CDOs can do differently.

Privacy debt: the hidden liability CDOs can no longer defer

6 juil. 2026 · 4 min

Most organizations have spent years accumulating privacy debt, patching compliance gaps rather than building coherent data governance. For CDOs, 2026 is the year that debt comes due, and the bill looks different than many expected.

Why data culture fails before the technology does

1 juil. 2026 · 4 min

Most data transformation programs collapse not because of bad architecture or wrong tool choices, but because the organization never genuinely changed how it thinks about data. For CDOs, understanding this distinction is the difference between building a legacy and managing an expensive disappointment.

Data products: how CDOs are turning internal assets into revenue engines

28 juin 2026 · 5 min

Most organizations are sitting on data assets worth millions, and doing almost nothing with them. Here is how forward-thinking CDOs are reframing data as a product and building sustainable monetization models.

The data mesh reckoning: why most enterprise architecture decisions made in 2022 are failing in 2026

25 juin 2026 · 4 min

Thousands of enterprises committed to data mesh, lakehouse, or hybrid architectures between 2020 and 2023, many are now quietly rebuilding. Here is what separates the architectures that scale from the ones that become expensive technical debt.

Why your data culture initiative is failing before it starts

24 juin 2026 · 4 min

Most organizations invest in data tools and governance frameworks, then wonder why adoption stalls and insights gather dust. The problem is rarely technical, it's cultural, and fixing it requires CDOs to operate more like organizational psychologists than technology executives.

Why most AI strategies fail before they start: the CDO's structural blind spot

23 juin 2026 · 5 min

Most organizations invest heavily in AI tooling while systematically underinvesting in the data foundations that make those tools work. For CDOs, closing this gap is not a technical problem, it is a governance and organizational design challenge that demands a fundamentally different approach.

When your data becomes the breach: how CDOs must rethink privacy as a strategic asset in 2026

22 juin 2026 · 4 min

Data breaches are no longer just IT incidents, they are existential threats that land squarely on the CDO's desk. Here is how the most effective data leaders are turning privacy from a compliance checkbox into a genuine competitive differentiator.

From data asset to data product: the CDO's most urgent strategic shift

21 juin 2026 · 4 min

Most organizations are sitting on data goldmines they've never learned to extract value from. The shift from managing data as an internal asset to engineering it as a monetizable product is redefining what CDO leadership actually means.

Data governance is not a compliance exercise, it's your most underutilized competitive weapon

20 juin 2026 · 4 min

Most organizations treat data governance as a defensive posture, a checkbox for regulators and auditors. The CDOs who are pulling ahead understand it as an offensive capability that accelerates decision-making, unlocks AI readiness, and builds institutional trust at scale.

Why your data culture initiative is failing, and what high-performing CDOs do differently

19 juin 2026 · 4 min

Most organizations have declared data culture a strategic priority, yet fewer than 30% of data initiatives deliver measurable business value. The gap between intention and execution reveals a fundamental misunderstanding of what building a data culture actually requires.

Beyond dashboards: why most BI programs fail to deliver strategic value, and what CDOs must do differently

18 juin 2026 · 4 min

Organizations spend millions on business intelligence infrastructure, yet fewer than 30% of analytics initiatives measurably influence executive decision-making. The gap between data availability and data-driven culture is not a technology problem, it's a leadership problem that sits squarely on the CDO's desk.

From cost center to revenue engine: how leading CDOs are building data products that actually sell

17 juin 2026 · 4 min

Most organizations sit on data assets worth millions yet generate zero external revenue from them. The CDOs who are changing that equation aren't just thinking about governance, they're thinking like product managers and venture capitalists simultaneously.

Why most AI strategies fail before they start: the data foundation problem CDOs can't ignore

16 juin 2026 · 4 min

Organizations are pouring billions into AI and machine learning initiatives, yet Gartner estimates that 85% of AI projects never make it to production. The root cause is rarely the algorithm, it's the data strategy underneath it.

Why your data architecture is lying to you, and what modern CDOs are doing about it

15 juin 2026 · 4 min

Most enterprises believe they have a data architecture. What they actually have is a collection of historical accidents held together by good intentions and expensive middleware. Here's how the CDOs redefining competitive advantage are building differently.

Data governance is not a compliance exercise, it's a competitive weapon

14 juin 2026 · 4 min

Most organizations treat data governance as a checkbox activity driven by legal pressure. The CDOs who are winning in 2026 have reframed it entirely, as the operational backbone of enterprise intelligence and a direct driver of shareholder value.

Privacy is not a compliance checkbox: how CDOs can turn data protection into competitive advantage

13 juin 2026 · 4 min

Most organizations treat privacy as a legal burden, a cost center managed by lawyers and auditors. The CDOs who are winning in 2026 have figured out something different: privacy architecture is a strategic asset that drives customer trust, accelerates data monetization, and reduces existential risk.

Why your data strategy will fail without a culture strategy first

12 juin 2026 · 4 min

Most CDOs can architect a data platform in their sleep, but fewer than 30% of data-driven transformation initiatives actually deliver measurable business value. The missing variable is almost never technology; it's the human system surrounding it.

From dashboards to decisions: why most BI programs fail to deliver business value

11 juin 2026 · 4 min

Organizations collectively spend billions on analytics infrastructure, yet fewer than 30% of business decisions are actually informed by data. The gap between BI investment and business impact is not a technology problem, it's a strategy problem that falls squarely on the CDO's desk.

From cost center to revenue engine: how leading CDOs are building data products that actually sell

10 juin 2026 · 4 min

Most organizations sit on data assets worth millions, and do nothing with them. Here's how the most commercially aggressive CDOs are turning internal data into structured products that generate real, measurable revenue.