Develop the question together
Give analysts and domain experts a shared context for the data, the study scope and the assumptions that still need resolving.
Scientist · Collaborative research & visual data science
Follow the question.
Keep the evidence in view.
Turn your data into an investigation others can inspect. Scientist brings shared research, visual workflows, statistical analysis and AI-assisted reasoning together—from the first hypothesis to findings linked to their supporting work.
Scientist · Renewal cohortsResearch workflow
What changes when the cohort changes?
Follow a preparation choice into the interpretation.
Define the comparison
QuestionHow does the renewal pattern differ by cohort?
Unit of analysisSubscription within an observation period
Before the runAgree eligibility with a domain specialist
Research intentInvestigate a pattern. Do not assume it explains why a customer renews.
Make preparation inspectable
InputsAccount records + Subscription activity
Visible pathJoin → Eligibility → Derive cohort
Definition to inspectHow paused subscriptions enter the population
Working methodUse visual operators, with specialist code where the question needs it.
Change a methodological choice
CheckpointSubscription preparation
Current treatment
Review before aggregationInspect incomplete records and affected cohorts
Why it matters
Compare the method—not just the output
Earlier approachExclude records with missing renewal dates
Current approach
Check before interpretingWorkflow revision + Run context + Source changes
Research reviewA changed output needs an explanation of what changed in the investigation.
Write the finding with its limits
Report contextCohort definition + Observation period
Supporting workPrepared data + Workflow + Relevant run
Conclusion boundaryObserved pattern ≠ Causal explanation
A useful handoffLink the evidence, state the limitation and name the next question.
The method matters
A result is only as useful as the path behind it.
A joined table can change the population. A missing-value rule can change the comparison. A model can suggest an explanation the evidence does not support. Scientist keeps those choices inside the investigation.
Build a path you can inspect
Connect preparation and analytical operators. Use custom logic where it belongs, then inspect meaningful intermediate results before interpreting the output.
Explain the finding with its limits
Keep experimental revisions and supporting evidence with the research discussion. Make the report useful to someone who needs to challenge it—not just read it.
Explore Scientist
One investigation. Thirteen connected capabilities.
Start where the question needs work: shared context, a visible analytical path, a defensible method or a finding ready for review.
Investigate together
Build & inspect
Work with data & AI
Analyst, domain expert and AI
Different contributions. A shared standard of evidence.
Visual workflows make the method discussable. Code brings specialist logic. Model-assisted reasoning can prepare interpretations. None removes the need to examine the research design.
Define what the records mean.
Agree the business definitions and exceptions before a transformation turns them into an analytical population.
Make the comparison defensible.
Review source relationships, feature availability and evaluation boundaries. Keep a workflow revision distinct from proof of exact reproducibility.
Challenge the proposed interpretation.
Use model suggestions as research material. Inspect excerpts, ambiguous cases and contradictory evidence before accepting a theme or writing a conclusion.
Scientist in the DO Stack
Research informs the work. The next owner decides what follows.
Scientist develops the analytical evidence. Other parts of DataObserve support company knowledge, engineering, operations, approvals, security and training around it.
Coworker
Bring company knowledge into the question and make reviewed findings useful to teams. Define what context and evidence should be exchanged.
Explore Coworker ↗Developer
Take specialist logic or a research-informed requirement into engineering review. Agree the tests and implementation contract separately.
Explore Developer ↗Unified Observability
Investigate instrumented services when runtime behavior affects the work. Do not confuse a service incident with an analytical-method issue.
Explore Unified Observability ↗Flow
Put a consequential decision or follow-up with a named owner. A research finding is evidence for review, not permission to act.
Explore Flow ↗SAFE
Review the research environment and connected material against security requirements. Validate evidence for each required control.
Explore SAFE ↗Model Training
Plan a separate training or infrastructure effort when the research requires it. Confirm datasets, compute and evaluation before operational use.
Explore Model Training ↗These are planning relationships, not a claim of automatic native integrations. Confirm interfaces, source access, artifact transfer and implementation responsibilities for your deployment.
From hypothesis to supporting work
Keep the analytical thread connected.
The research question, the preparation path and the evidence-linked brief belong to the same investigation.
Plan the first investigation
Start with the question. Validate the operating requirements.
Use Scientist’s existing collaborative research, visual execution, data and AI, and sharing capabilities as the starting point—not assumptions borrowed from another platform.
Is Scientist a replacement for researcher judgment?
No. Scientist helps develop the analytical workflow and supporting evidence. Researchers still need to define the study, inspect the method and decide which conclusions the material supports. Model-assisted suggestions require the same scrutiny.
Can visual workflows and custom code work together?
The existing offering includes visual operators and custom Python, R and Java/Scala logic. Confirm language versions, libraries, packaging and runtime limits for your environment. A supported language is not a promise that every package or execution mode is available.
Can we exactly reproduce an earlier experiment?
Workflow and analytical-run history helps explain how an investigation changed. Exact reproducibility also depends on source data, dependencies, model versions and execution conditions. Validate which of these are captured and retained for your study.
Which connections and collaboration controls are available?
Bring the sources, intended researchers and access requirements to the implementation discussion. Confirm supported connections, workspace permissions, concurrent editing behavior, identity and retention rather than assuming a universal connector catalog or particular security guarantees.
Can a reviewed analysis become a recurring tool?
Data apps and automation provide a framing for reusable analytical work. Confirm delivery, scheduling, triggers, source access and failure handling for your use case. A recurring analysis should still have a method owner and a clear boundary for further review.
What belongs in the first research discussion?
Bring one question, the data definitions, an intended observation period and the people who understand the domain. Agree what would count as useful evidence, identify access constraints and choose a preparation or interpretation checkpoint to inspect first.
Start with a question worth investigating.
Bring the question, the available data and the people who understand it. We’ll work through the analytical path, the review points and the implementation requirements.