R&D evidence review
Bring technical and financial records into one scoped engagement. Receive an initial evidence review showing what the records support, what is missing and which questions to answer next.
What you do. What you receive.
Start with one engagement and supply only records you are permitted to share. Collection status and missing inputs remain visible while your initial review is prepared. You can return to the same engagement as the work develops.
Create your account, identify the company and record the period and jurisdiction. Save the engagement and return when you are ready to supply records.
Upload project notes, experiment evidence, payroll and expense records. Where enabled, collect selected GitHub, Linear, Jira or GitLab records and capture permitted Notion or Google Drive documents. Save exact evidence versions separately from live collection. Uploads work on their own.
Submit your available records for preparation. A preparer can assemble the initial review manually, linking supported project notes and available cost records to their sources and identifying what still needs an answer.
Open the prepared review in your engagement, inspect its source references and download the saved version. Use its coverage, gaps and next questions to continue the work with your adviser.
The initial review supports preparation; it does not determine tax eligibility or replace an approved annual report. Cost allocation and formal delivery need their own agreed method and responsible reviewer.
Software reliability · January–June
The project notes describe comparing keyword and embedding retrieval for ambiguous queries. Final benchmark results have not been supplied.
This sample explains the format; it is not a customer result or an available allocation feature. Your review uses supplied records and identifies its preparer. An initial review is not a tax eligibility determination or an approved annual report.
The broader platform
Proforma also brings data engineering, AI and experimentation into a programmable workspace. The examples and capabilities below show how teams can investigate, build Apps and run Actions across sales, marketing, operations and technical R&D. They are available separately from the focused customer collection flow.
Business Experiment Lab
Investigate conversion and customer quality across sales, marketing and product. Compare a change against its baseline, inspect the tradeoffs, and use the evidence to plan the next test.
3 business questions / inspectable sample results / proposed next tests
Swipe the capture to inspect it, or open it full size.
Broader platform workflows
A sales message, an operating process and a technical design can all improve through investigation and testing. The evidence each needs is different. Start with the job in front of you, then keep what helps the next decision. A direct analysis or automation task does not need a formal experiment.
In Business Experiment Lab, a broader landing-page promise produces 102 signups instead of 72, but only 42 qualified leads instead of 48. The larger number hides a weaker business outcome.
Keep the qualification definition and full-visitor denominator. The proposed next test keeps the specific promise and changes the supporting proof.
In Experiment Bench, fixed retries deliver all 120 packets, but p95 delay is 15 minutes against a 10-minute target. A repeatable simulation makes the trade-off visible.
Keep the fixture and delay target. A proposed follow-up varies the retry interval. Sensor Calibration Study asks a different question: how well does the same fitted method hold up under drift?
Both examples use synthetic records. They illustrate a method and a proposed next step, not a proven business improvement or a completed follow-up. Explore the calibration example.
Start in Chat with a question: why are more signups producing fewer qualified leads? Can a retry policy recover packets within the delay target? Inspect the available Data and agree on the outcome and constraints.
Inspect identifiers, missing events, units and freshness. Define joins, qualification rules and conversion denominators in SQL Models. Keep reusable checks and computation in Actions, and build an App around the comparison.
Use Actions and configured Connections for repeatable work: refresh data, run checks and coordinate tasks on a schedule or event. Inspect the execution and its outputs. Keep the audience or test fixture, changes and review window with the result.
Compare qualified demand with signup volume and customer quality. Compare packet delivery with delay and transmission attempts. Keep denominators, targets and uncertainty visible; a better headline number can still miss the goal.
Record what happened, what remains uncertain and what to try next. Keep useful definitions in Context, calculations in Models and repeatable execution in Actions. Reuse and adapt the App as evidence develops.
Save the meaning of a qualified lead, a delivery target or an acceptable error in Context. Keep the source evidence in files and Data so the next investigation can inspect the same starting point.
Keep calculations in SQL Models, interaction in Apps and repeatable execution in Actions. Reuse transformations, quality checks and instrumentation as well as the analysis, with definitions and assumptions available for review.
Retain the evidence, decision and limits of a finding. A colleague or agent can review what you tried, see what remains uncertain and choose the next change without starting from an empty conversation.
The computational engine
Connect the data, define what matters and build the App around the work. Use the same workspace to answer a one-off question, test a change and run the useful routine it becomes.
Work with warehouse tables, connected services, documents and spreadsheets. Inspect source schemas and sync history, reconcile records and prepare reusable data for analysis and automation.
Find the datasets available to you and inspect their structure and descriptions.
Query imported data and inspect its provider, schema, and sync history.
Sync supported services so your analysis can work with updated data.
Use CSVs, workbooks, and Google Sheets alongside warehouse data and source documents.
Explore patterns, check assumptions, and write reusable analyses. Inspect the calculation before interpreting what the result means.
Define reusable transformations and calculations in SQL, then query them throughout the workspace.
Inspect calculation definitions and their data dependencies.
Investigate patterns in revenue, retention, costs, research data, and operations.
Collect behavioral events and analyze them beside your other evidence.
Combine customer, product, and commercial records to investigate a more complete picture.
Create Apps for inspecting data, comparing results and operating a workflow. Use Actions to refresh data, call connected services and coordinate repeatable work, including configured AI calls for text or structured output.
Build an interface for exploring results, changing inputs, or reviewing a body of evidence.
Run saved routines for reports, reconciliations, checks, and alerts on a schedule or event.
Receive external events and use them to trigger a routine.
Ask a question, work through an analysis with AI, and create files, data models, Apps, and Actions.
Start with a working application, inspect its setup, and adapt it to your question and data.
Follow task progress, inspect outputs, and respond when your agent needs input.
Use connected tools to collect context, prepare communications, and carry an investigation into the places your team already works.
Work with conversation context and bring an analysis into a team discussion.
Prepare and send messages through a connected account with the controls your workflow needs.
Use supervised browser work to investigate or interact with systems beyond a data connection.
Collect company and contact research for market analysis and outreach.
Bring recorded demonstrations and reviews into your research material.
Capture observations, decisions, and open questions from a conversation.
Prepare, simulate, review, and place calls through supported voice connections.
Use APIs and coding agents with scoped workspace access. Connect a specialist executor when an investigation needs work in a repository.
Give external software a scoped path to workspace data and operations.
Let a coding agent work with the data and tools in an approved Proforma workspace.
Delegate a defined repository task to a connected coding system and follow its progress and outputs.
Use workspace data and operations from your project shell.
Use one native MCP connection to read governed data, install complete packages, build Apps and run approved Actions in the same Workspace.
Configure https://app.proformalabs.ai/mcp as a Streamable HTTP server in your compatible host, then authorize your Workspace in the browser. Use a plugin or a direct registration once; /mcp/analysis remains an alias to the same catalog. Keep execution receipts and follow continuations and readiness before claiming completed work. Scopes and Model policy still govern each operation.
The technical model
Files hold your notes and code. Data lets you discover and query your datasets. These five building blocks turn the workspace into tools you can adapt to your own questions and workflows.
Start with your R&D evidence
Create a scoped engagement for your company, supply the records you have and return to a prepared initial review. Discuss follow-up and commercial arrangements separately through the advisor email.