What Optics BI does and how it helps your business

How it works and why that matters

Optics BI separates data calculations from AI interpretation. It works in two stages:

1. The AI uses information about your dataset, including column names and data types, to translate your question into a structured request. This identifies the measures, filters, groupings and comparisons needed to answer it.
2. A separate data-processing engine executes that request. The calculated results are then provided to the AI to explain the findings and suggest relevant next steps.

This means totals, averages and other measures are calculated from your data rather than relying on the language model to perform the arithmetic.

Privacy controls restrict which information is shared with the AI. The AI still receives information needed to interpret the results, so these controls matter at both the question-planning and explanation stages.

Why it helps

You can move from a business question to a useful explanation more quickly. Instead of manually assembling each calculation and chart, you can focus on understanding the findings and deciding what to investigate or act on next.

How you use it

Upload a dataset, open the Senior Analyst panel, and ask questions such as:

• “What are my top products by revenue?”
• “How have customer sales changed this quarter?”
• “Which states have the highest return rates, and what is the value of those returns?”

You can continue in the same conversation with follow-up questions such as “Why might that be happening?” or “Break that down by sales channel.” Optics BI uses the conversation context to continue the analysis.

Worth knowing

The quality of an answer depends on the data available and how the question is interpreted. Optics BI is designed to flag missing information or questions it cannot answer reliably.

Calculated results support the explanation, but recommendations and possible causes still require judgment. Where the data shows an association, treat it as a lead to investigate rather than proof that one factor caused another.

How it works

Optics BI uses the structure of the calculated results to select a visualization, such as:

• Line charts to show changes over time.
• Bar charts to compare products, locations or other categories.
• Distribution charts to explore how values are spread.
• Maps to display geographic patterns when suitable location data is available.

Formatting helps distinguish dates, currencies, percentages and counts. Depending on the chart, you can hover for values, zoom into an area or show and hide individual series.

Why it helps

You can move quickly from a question to a visual explanation, making patterns easier to understand and findings easier to discuss with colleagues.

How you use it

Ask a question in the Senior Analyst panel. When the result supports a visualization, Optics BI generates a chart.

Pin useful charts to a dashboard and rearrange them to bring related findings together. Export your dashboard to PDF on any plan, or to PowerPoint on Growth and Enterprise plans.

Worth knowing

Charts reflect the data, filters and calculations used in the analysis. Review titles, units, date ranges and category groupings before sharing. Some answers are clearer as a table or written explanation and may not need a chart.

How it works

Optics BI examines your dataset for missing values, duplicate rows, inconsistent data types and unusual values. It presents identified issues with severity levels, the number and percentage of affected rows, and recommended actions.

For example, it can flag a numeric column stored as text, missing values that could affect a calculation, or duplicate records that may inflate totals.

Why it helps

You can spot potential problems early, reduce manual preparation and understand where your data needs attention before using it to make decisions.

How you use it

Open the Clean panel to review the assessment. Apply individual fixes, such as removing duplicates, handling missing values or standardizing a column, or use Auto-clean to apply the recommended changes together.

Conservative mode is the default. Review the proposed changes before choosing more aggressive cleaning options, particularly those that remove records or replace missing values.

Worth knowing

An unusual value is not necessarily an error, and repeated records may represent legitimate transactions. Cleaning decisions should reflect what your data means. Review the changes and their effect on your dataset before continuing your analysis.

How it works

Optics BI supports four types of analysis:

• Forecasting: Uses historical dates and values to estimate future results, helping you explore the likely direction of a measure such as sales or demand.
• Anomaly detection: Highlights observations that differ from typical patterns, helping you focus attention on unexpected changes or potential data issues.
• Segmentation: Groups records with similar characteristics and describes their differences, helping you understand customer groups or other patterns in your data.
• Driver analysis: Uses correlation analysis and predictive modeling to identify variables associated with an outcome and assess their contribution to the model’s predictions.

Why it helps

These analyses can support planning, prioritize investigations and help you tailor business actions to different groups. They give you evidence to explore before committing resources.

How you use it

Open the Predict panel, choose the analysis type and select the relevant columns. Review the results alongside any available explanations and model performance measures.

Worth knowing

Forecasts are estimates, and their reliability depends on the quality, quantity and relevance of the historical data. Unusual observations deserve investigation but are not automatically problems. A variable that helps predict an outcome does not necessarily cause it.

Use the results alongside your business knowledge, and review model performance before relying on predictions for important decisions. Treat it as a well-informed starting point, not a promise.

How it works

Business memory has two parts:

• Business profile: Information you provide about your company, key metrics, targets and terminology.
• Analysis memory: Findings from previous analyses that are retained to provide context for future questions.

This context helps Optics BI tailor its explanations to your business. For example, a saved definition of churn can help it interpret future churn questions consistently.

Why it helps

Your team can spend less time explaining the business and more time exploring results. Shared definitions also support more consistent conversations about performance.

How you use it

Complete your business profile during setup and update it as your priorities change. Review the memories created as you work. You can edit or delete individual memories, or clear them together.

Worth knowing

Saved findings are context, not permanent facts. Review them when your data, definitions or business conditions change. Profiles and memories are encrypted at rest and processed through the platform’s privacy controls.

How it works

Optics BI uses your dataset’s structure and selected transformations to generate Python code for extracting, transforming and loading data.

Supported destinations include Snowflake, AWS S3 and Redshift, Google BigQuery, and Azure Synapse.

The generated code incorporates your dataset’s column names and data types, reducing the amount of initial setup your team needs to write.

Why it helps

Your technical team gets a starting point for building a repeatable data workflow, saving time on the first draft of integration code.

How you use it

Open the Warehouse panel, choose your destination and transformations, and generate the code. Have your technical team review it, configure credentials and dependencies, and test it before deployment.

Worth knowing

Generating code does not automatically deploy a pipeline or transfer your data. The code may need adjustments for your infrastructure, access permissions and operational requirements.

How it works

Access is managed through two separate controls:

• Roles determine which actions a member can perform, such as inviting colleagues, changing settings or managing datasets.
• Sharing determines which datasets a member can open and use.

An administrator’s role does not automatically grant access to the contents of every colleague’s dataset. Dataset access requires sharing.

Administrators can view an inventory containing dataset names, sizes and owners to support workspace management without opening the underlying data.

Why it helps

Teams can collaborate while keeping access to business data limited to the people who need it.

How you use it

Invite colleagues by email, assign appropriate roles and share the datasets they need.

Included seats are:

• Starter: 1 seat.
• Growth: 3 seats.
• Enterprise: 10 seats.

Invited members use the workspace’s subscription and included features; they do not need a separate subscription.

Worth knowing

Review roles and shared access when responsibilities change or someone leaves the team. Dataset names and other inventory details may be visible to administrators, so avoid placing sensitive information in dataset names.

How it works

• Encryption: Stored questions, answers, chart data, business profiles and connector credentials are encrypted using AES-256-GCM, in addition to the hosting provider’s disk encryption.
• Privacy filtering: A privacy classifier evaluates columns before their values are included in model requests. It is designed to withhold identifying information and restrict other sensitive values according to the platform’s privacy rules.
• Server-side value restoration: Where supported, withheld values are restored in the response on Optics BI’s servers, allowing users to see the relevant details without including those values in the model request.
• Sensitive-column controls: You can mark additional columns as sensitive for each dataset.
• Two-factor authentication: Protect accounts with an authenticator app and recovery codes. Organizations can require two-factor authentication for their members.
• Session management: Revoke individual sessions or all sessions, with idle-timeout controls to limit inactive access.
• Tamper-evident audit logs: Linked cryptographic hashes support checks for changes to recorded audit entries. Verification is available through the app.
• Retention and account controls: Configure retention settings and use the available data-export and account-deletion options.

Why it helps

These controls give your team greater visibility and control over how information is stored, accessed and used during analysis.

How you use it

Enable two-factor authentication, review active sessions, mark sensitive columns and choose retention settings that suit your organization. Use the audit log to review recorded activity.

Worth knowing

Automated privacy filtering may not recognize every sensitive field. Review your dataset and apply additional sensitive-column settings where needed. These controls support your organization’s security practices; they do not, by themselves, establish regulatory compliance.