DataKitchen IO (@datakitchen_io) header
DataKitchen IO (@datakitchen_io) profile picture

👨‍🍳 Our #DataOps Platform simplifies complex data toolchains, environments & teams 👩‍🍳 Get Your 🆓 DataOps Certification ➡️ https://xitter.cf/t.co/Vdd3AAgTxI

Cambridge, MA
Joined January 2015
Photo by DataKitchen IO (@datakitchen_io)Photo by DataKitchen IO (@datakitchen_io)Photo by DataKitchen IO (@datakitchen_io)Photo by DataKitchen IO (@datakitchen_io)Photo by DataKitchen IO (@datakitchen_io)Photo by DataKitchen IO (@datakitchen_io)Photo by DataKitchen IO (@datakitchen_io)Photo by DataKitchen IO (@datakitchen_io)Photo by DataKitchen IO (@datakitchen_io)Photo by DataKitchen IO (@datakitchen_io)
Fabric takes no SQL logins, so a tool needs an Entra service principal, and three different admins hold the three permissions. Microsoft's steps, in order: hubs.ly/Q04D9W6T0
26
TestGen alerts on the data. DataOps Observability alerts on the pipeline. Point both at one Slack channel and one person sees the failed task and the bad rows it caused. Slack, Teams, email, and Jira, channel by channel: hubs.ly/Q04C2Wnb0
29
How to Do Data Contracts Webinar, tomorrow 12pm ET. Chris Bergh builds a working contract live from a messy data set in open source TestGen: 80% of the tests in 60 seconds. Free, one hour: hubs.ly/Q04FlXwF0
23
AI 101 for data quality people: you already know most of what AI needs from you. DataKitchen's Chip Bloche and Danette McGilvray teach Quality Data and Trusted AI at DGIQ + AIGov, Mon Nov 16: hubs.ly/Q04D61vw0
18
Tomorrow, 12pm ET: Chris Bergh builds a data contract live from a messy data set in open source TestGen. 80% of the tests in 60 seconds, 90% in five minutes with an LLM. One hour, free: hubs.ly/Q04DMLDs0
1
67
Purview data quality or TestGen in Microsoft Fabric? Usually both, for different jobs: Purview governs the estate, TestGen tests the tables: hubs.ly/Q04D9VTQ0
52
Fabric Warehouse rejects SQL logins, and every test query uses Fabric capacity. How to connect, what to test, and how to run it after each load: hubs.ly/Q04D9VP-0
34
"Consistency is at 70%" has never made anyone fix a row of data. "These 412 rows have a bad customer ID, here they are" does. The six types of data quality dashboards, and where each one stops working: hubs.ly/Q04DML6p0
1
43
What data governance does a three-person SQL Server team actually need? Five things, no program: catalog, Critical Data Elements, tests, a quality score, owners: hubs.ly/Q04D9V320
36
How many hours a week does a data quality tool take to run? Measure it in your trial. Alert noise drives it more than table count, and hand-writing tests costs more: 2,500 tests is 1,250 hours. hubs.ly/Q04C2WQC0
28
Data contracts shipped in open source TestGen. Generate one from tables you've profiled or import an ODCS file, freeze it in numbered versions, compare any two, and do the round trip from your AI assistant: hubs.ly/Q04FddpC0
26
Ask your data observability vendor to define a credit in writing. Monte Carlo's Scale rate on 1,000 tables is about $179,000 a year. TestGen Enterprise for the same team: $15,600. The receipts: hubs.ly/Q04DXzhR0
33
Purview Data Quality: $15 to $240 per processing unit plus $0.0165 per governed asset per day (East US, Nov 2026). Open-source TestGen: $0 for one connection. The math: hubs.ly/Q04D9-Nh0
37
Google's Knowledge Catalog checks the BigQuery rules you write. Somebody still writes them, up to 1,000 per scan. Five tools compared, including the one that writes the tests for you: hubs.ly/Q04C2W8b0
21
114 organizations now use the ODCS data contract standard, up from nine in 14 months. We put about 30 contract tools through a round-trip test and a run-the-tests test. Nine passed: hubs.ly/Q04DMHDZ0
16
Can you ask Claude or Copilot what's wrong with your SQL Server data? Yes, if it can read real data profiling results and test history instead of guessing: hubs.ly/Q04D9ScF0
1
43
80% correct data, queried by an 80% accurate model, gets you about 64%. The errors multiply. A six-row check for whether your commercial pharma data is ready for AI: hubs.ly/Q04C2W900
1
28
One blended data quality score can tell brand share and incentive comp they're both fine. Give each consumer its own scorecard, pass lines and owner: hubs.ly/Q04DXl7g0
1
16
One row moved in a 40-row market basket file, and your lead product's market share dropped four points. The IQVIA feed loaded clean. Why small pharma files do the most damage, and four practices that keep them honest: hubs.ly/Q04DMH500
1
32
Yes, you can check on-prem SQL Server data quality without sending data to the cloud. What a tool has to do to qualify, and every connection TestGen makes: hubs.ly/Q04D9Rwc0
27