Data & Analytics
Decim
Find out why your data pipeline broke.
- Category
- Data & Analytics
- Stage
- Beta
- For
- Data engineers and engineering teams responsible for data pipelines and on-call incident investigations.
- Platforms
- Web
- Pricing
- Free during the design-partner pilot; Team and Enterprise priced per environment (https://decim.dev/pricing).
About
Monitoring tells you a pipeline broke. Decim finds out why. Describe the incident in plain language, and Decim maps the pipeline from your GitHub code, reads each step's logs from Google Cloud Logging, and imports the Jira ticket. It weighs up to five competing explanations and rules most of them out. It confirms a cause only when the evidence supports it, and every claim links to the code line or log line behind it. It then drafts the fix as a patch, SQL repair or draft pull request for your team to review.
Features
- Problem
- An incident report does not identify which pipeline or step caused the problem.
- How it works
- Discovers pipelines and their steps from selected GitHub repositories, then ranks matches to the incident and cites relevant code.
- Result
- An identified pipeline and step, with the code behind the match.
- Why
- A code-linked pipeline map gives the investigation a grounded starting point.
- Problem
- A successful job status can hide missing rows or other abnormal behavior.
- How it works
- Reads logs step by step and compares them with successful runs from up to 30 days earlier.
- Result
- Findings grounded in log evidence and comparisons with healthy runs.
- Why
- A baseline helps distinguish unusual behavior from normal variation.
- Problem
- A plausible explanation can be mistaken for a confirmed cause.
- How it works
- Keeps up to five competing explanations, weighs evidence for and against them, and proposes checks when more evidence is needed.
- Result
- A confirmed, likely, unclear, or access-limited outcome, with rejected explanations retained.
- Why
- Testing alternatives and requiring corroboration helps separate evidence from speculation.
- Problem
- Incident conclusions are difficult to verify after the investigation.
- How it works
- Links findings to exact log excerpts, code, commits, tickets, baselines and submitted check results.
- Result
- An auditable investigation record and a PDF report.
- Why
- Traceable evidence lets engineers inspect what each claim rests on.
- Problem
- A diagnosis does not by itself provide a safe, reviewable path to remediation.
- How it works
- Proposes a repository patch and, where relevant, a SQL repair; can open a draft pull request. It does not run queries or merge changes.
- Result
- A proposed fix for review, with any data repair left for a person to run.
- Why
- Keeping execution and approval with the team preserves human control over production changes.
Product questions
What does Decim do?
Decim investigates data pipeline incidents after an alert or report, tracing symptoms through pipeline code and logs to identify a cause supported by checkable evidence.
Who is Decim for?
Data engineers and engineering teams responsible for data pipelines and on-call incident investigations.
How much does Decim cost?
Free during the design-partner pilot; Team and Enterprise priced per environment (https://decim.dev/pricing).
Which platforms does Decim support?
- Web
Screenshots
