Introducing Sisense Fusion 2021.11 — Easy-to-use UI, seamless installations, and More!
Dig into the latest features to discover what the new features in Sisense’s analytics platform can do for you. Learn how to install Sisense using Provisioner and Helm, copy theme identifiers, define a time for support access users, and test email server configuration.Exploring RAG: A Sisense-Based Approach with BigQuery and Gemini
Exploring RAG: A Sisense-Based Approach with BigQuery and Gemini Continuing from our previous article, Automated Machine Learning with Sisense Fusion: A Practical Guide, where we discussed how Sisense and AutoML simplify complex machine learning workflows, I am now excited to introduce an advanced feature: a chatbot interface powered by Retrieval-Augmented Generation (RAG) and a large language model (LLM) like Gemini. This enhancement allows users to interact with their data in natural language through Sisense's Native Blox widget. Imagine asking questions like "How many products did I sell per category?" and receiving insightful answers instantly. This not only improves data accessibility but also bridges the gap between technical data repositories and business decision-making. In this article, I’ll cover the first step of enabling this functionality: setting up a seamless integration pipeline between Sisense and Google BigQuery, embedding table and column metadata, and preparing the data for efficient querying. Important Note: Experimental Project: This implementation is an experimental project and not an official Sisense feature. It demonstrates how Sisense’s functionalities can be utilized to build a custom Retrieval-Augmented Generation (RAG) pipeline or how you can use this example by importing the provided notebooks. Google Ecosystem: The example provided is specific to the Google ecosystem, utilizing services like BigQuery (BQ) and Gemini for query generation and processing. Associated Costs: Please note that each prompt sent to the chatbot incurs a cost, which will be billed to your Google Cloud Platform (GCP) account. This includes charges for LLM processing and database queries. Building the Foundation: Data Preparation and Embeddings To enable natural language interactions with your data, we first need to establish a robust data infrastructure. This includes creating a vector store for embeddings in Google BigQuery (BQ) and preparing metadata about tables and columns. The process is fully automated using a pre-built Sisense custom code notebook, which simplifies embedding generation and management. Dataset in BigQuery The journey begins with your dataset in Google BigQuery. The notebook retrieves metadata such as table and column names, descriptions, and schema information from BigQuery’s Information_Schema. If any descriptions are missing, the system automatically generates them, ensuring comprehensive metadata coverage. Setting Up the Vector Store in BigQuery The vector store acts as the backbone for similarity searches within the RAG system. Using the custom notebook, the system: Creates the vector store: A structured repository to store embeddings for tables and columns. Organizes metadata: Ensures all table and column descriptions are structured and accessible. Generating Table and Column Descriptions Missing descriptions can hinder data understanding. The notebook includes a Description Agent that automatically generates meaningful text for: Tables: Contextual descriptions based on schema and usage. Columns: Descriptions highlighting their role within the dataset. These enhancements ensure the metadata is both informative and ready for embedding generation. Creating Embeddings with Vertex AI To enable semantic search, metadata descriptions are converted into numerical embeddings using Vertex AI’s TextEmbeddingModel. This is facilitated by the EmbedderAgent, which: Accepts strings or lists of strings (e.g., column descriptions). Generates embeddings through Vertex AI. Handles both single and batch processing for efficiency. Efficient Embedding with Chunked Processing For large datasets, embeddings are generated in chunks using the get_embedding_chunked function. This ensures: Scalability: Handles datasets of all sizes without performance issues. Parallel Processing: Processes text chunks simultaneously to speed up the workflow. Structured Outputs: Embeddings are returned in a structured DataFrame for storage or analysis. Storing Embeddings in the Vector Store The final step is storing these embeddings in the BigQuery vector store. This ensures that: Similarity searches are fast and efficient. Metadata is always accessible for chatbot interactions. ALT text: A screenshot of a data table displaying various columns such as "table_schema," "column_name," "data_type," "source_type," and several other attributes. The table shows sample values and metadata related to a database structure, organized in rows and columns. How the Chatbot Interface Works Now that the foundation for embeddings and metadata storage is set, let’s explore the chatbot interface in action. Imagine opening the chatbot in the Blox widget and asking a question about your dataset. Within moments, the chatbot responds in natural language, providing actionable insights. But what exactly happens under the hood to generate this seamless interaction? RAG Notebook and the Chatbot Workflow The chatbot operates using a pre-built RAG custom code transformation notebook, which orchestrates the end-to-end process. With this notebook, the entire pipeline—from understanding the query to generating the response—is automated. The notebook uses multiple specialized agents, each responsible for a specific task, ensuring precision and efficiency at every step. SQL Query Builder Agent BuildSQLAgent This agent specializes in constructing SQL queries for BigQuery. It uses the LLM to analyze the user’s natural language query and matches it with table schemas and column details from the vector store. It outputs a fully formed SQL query tailored to the user’s dataset and question. SQL Validation Agent ValidateSQLAgent The ValidateSQLAgent validates the SQL query before execution using a Large Language Model (LLM). Validation ensures the query adheres to essential rules, including: The presence of all referenced columns and tables. Proper table relationships and join conditions based on the schema. Formatting and compliance with BigQuery-specific SQL standards. Validation occurs during the debugging process, specifically within the DebugSQLAgent, to identify potential errors before attempting a dry run or execution. It provides a detailed JSON response: If valid, the process moves to the next step (dry run or execution). If invalid, the DebugSQLAgent uses the error details to refine the query iteratively. SQL Debugging Loop Agent DebugSQLAgent This agent runs the debugging loop to refine queries that fail validation or execution. The process includes: Validation: The query is passed to ValidateSQLAgent to check syntax, schema compliance, and structure. If valid, the query is ready for execution. Dry Run: If validation passes, the query is tested using a dry run via the test_sql_plan_execution function to confirm execution readiness. Execution: Once validation and dry runs succeed, the final query is executed using the retrieve_df function, which returns results as a DataFrame. Iterative Refinement: If the query fails either validation or the dry run, the DebugSQLAgent uses the LLM to troubleshoot and generate an alternative query. The loop repeats until a valid query is generated or the maximum debugging rounds are reached. This agent ensures the final query is: Correctly structured and semantically valid. Optimized for performance and aligns with the original user intent. Response Agent ResponseAgent This agent translates the SQL query results into natural language. It bridges the gap between technical SQL outputs and user-friendly communication. By combining the query results with the user’s original question, it crafts a clear and relevant response. How the Workflow Executes Here’s the step-by-step process for generating a response: User Query Embedding The EmbedderAgent converts the user’s natural language question into a numerical embedding. Using BigQuery native vector search, the system retrieves similar embeddings from the vector store created in the first phase. Schema and Content Retrieval Based on the retrieved embeddings, the system fetches relevant table and column schema details from the vector store. SQL Query Generation The BuildSQLAgent uses the retrieved schema details to construct an SQL query that aligns with the user’s question. SQL Validation and Execution The ValidateSQLAgent checks the generated SQL for accuracy and potential errors. If the SQL passes validation, it is executed against the BigQuery database. Debugging (if needed) If the query fails or generates an error, the DebugSQLAgent refines it iteratively until a valid query is produced. Response Generation The ResponseAgent uses the query results and the user’s original prompt to generate a natural language response. If the system fails to generate a valid response, it communicates the issue to the user. Conclusion By combining the foundational embedding process with this RAG-powered workflow, the chatbot transforms how users interact with their data. From seamless SQL query generation to delivering natural language responses, the system exemplifies the power of Sisense Fusion and advanced AI tools to simplify data-driven decision-making. As always, please reach out to your Customer Success Manager (CSM) if you would like to implement this in your own environment.
Limiting Date Range Filters in Sisense Dashboards
Wide date ranges in Sisense dashboards can lead to performance issues, especially when using live models or querying large datasets. For live data models, large queries increase costs as more data is pulled from the data warehouse. For Elasticubes, this can cause performance bottlenecks. To avoid these issues, here is a quick solution to "limit" the date range users can select, ensuring both cost-efficiency and smooth performance. Read more to find out how!
Building Confidence (and BloX Widgets) with the Help of Sisense Trainers
My personal experience taking two interactive training sessions with Sisense trainers as a new Sisense user. How I’ve been able to apply new skills to my dashboard work and feel more confident in dashboard design and customizing Blox widgets to get exactly the look and feel I want for my customer experience.Format Seconds to Time (HH:MM:SS) in a Pivot Table
This article will explain how to convert seconds into date-time format on a pivot table. 1. Ensure your seconds column is of type integer. 2. In the widget editor, add the widget script below. Information on how to add a widget script here: https://sisense.dev/guides/customJs/extensions/#dashboard-and-widget-javascript-extensions var dateColumns = [4,5,6,7,8]; //select the time columns that should be transformed var thousand_separator = ',' //replace , with your local thousands separator if needed var time_separator = ':' //replace : with your local time separator if needed var returnedData = ""; for (var i = 0; i < dateColumns.length; i++) { returnedData += "td:nth-child(" + dateColumns[i] + ") div"; if (i < dateColumns.length - 1) returnedData += ","; } widget.on('ready', function(se, ev){ var e = element; var w = widget; var d = dashboard; _.each($(returnedData , e).not('.wrapper, .p-head-content'), function(cell){ var num = parseFloat($(cell).text().split(thousand_separator).join('')); var sign = num < 0 ? "-" : ""; num = Math.abs(num); // console.log('num') console.log(num) if(isNaN(num)){ return; } else{ var hours = (parseInt( num / 3600 )); console.log('hours') console.log(hours) var minutes = parseInt( (num - (hours * 3600)) / 60 ); console.log('minutes') console.log(minutes) var seconds = num - (hours * 3600) - (minutes * 60) ; console.log('seconds') console.log(seconds) var hoursText = hours < 10 ? "0" + hours : hours; var minutesText = minutes < 10 ? "0" + minutes : minutes; var secondsText = seconds < 10 ? "0" + seconds : seconds; $(cell).text( sign + hoursText + time_separator + minutesText + time_separator + secondsText); } }) }) 3. Adjust the dateColumns array (line 1) with the index of columns you would like to convert. 4. Save the script. Refresh the widget and click apply. You're all set! Before and after the widget script is applied: Notes: the script may reset the values in the second column to 00:00:00, if this happens simply refresh. This script is based off an older post, updated for the latest version of Sisense.
Assessing the Quality of Your Dashboard
Introduction The following article discusses how to assess the quality and adoption of your dashboard. Table of Contents How to Measure a Dashboard's Quality? Let's begin by defining what a "Good Dashboard" is. To assess the quality of your dashboard, you should measure the following aspects: Is the dashboard practical? Does it serve its purpose? Does the dashboard display relevant and correct information? Is the information in the dashboard displayed correctly? Was the dashboard adopted by the end-users? Is the dashboard intuitive for use? Planning Makes Perfect Correct dashboard planning is the key to its success! I recommend you read this article discussing a dashboard's (high-level) development cycle. It breaks the process into easy measurable steps that start from the initial KPI planning to maintaining and adjusting your end product. Reading through it might help identify flaws during the dashboard's initial rollout process. Assessing a Dashboard's Quality The "Practical" Test The practicality of a dashboard focuses on whether the dashboard serves its purpose. Going back to a dashboard's planning phase, try to recall the reason for building this dashboard. What were the end-users looking to achieve? What insight were they seeking? Possible answers are: Monitoring a person's or a group's achievements Monitoring a measurable process Help decide what to do next based on historical information Assess future steps in various scenarios (What-If Analysis) A good dashboard: Answers its purpose! How to Test This? To check whether your dashboard is practical, you should: Set up a call with one (or more) end-users and/or stakeholders using the dashboard Ask them whether they can make the intended decision based on it If they are - What is the process of making the decision? If they aren't - What are the obstacles that stand in the way of making it? (e.g., missing/incorrect data, bad user experience, etc.) Next, revisit the dashboard's goals: What question should it answer? What action can the users make based on it? Compare the information you got from your stakeholders and the plan made when designing the dashboard - Define the gaps. How to Resolve This Type of Issue? Work towards re-planning your dashboard to meet its goals. Focus on: Aligning the "Call for Action" Redefining the KPIs and the user story Redesigning and restructuring the KPIs Optimizing/adjusting the data model to be able to answer the different KPIs Resolving this type of issue will most likely require a complete dashboard redesign, followed by an entire UAT cycle. The "Data Relevance" Test The term "Data Relevance" breaks into two aspects: Am I displaying relevant data - Considering the audience of the dashboard? Am I displaying relevant data - Considering the decision to be made? Does It Address the Right Crowd? Like any deliverable, one of the fundamental things to know is its audience. The answer to this question will affect the following: The type of dashboard you create The granularity of data you present The way information is presented A good dashboard: Is built with the right audience in mind. Does it Contain the Right Amount of Data? Making a decision requires having the right amount of data: Too little data - Could prevent the person from making the right call. Too much data - Could confuse the person and throw them off-track. A good dashboard: Has the right amount of data required to serve its purpose. How to Test This? To check whether your dashboard has relevant data, you'll have to: Assess who are the end-users and/or stakeholders using this dashboard Set up a call with one (or more) end-users and/or stakeholders using it Ask them what the process of making their decision is? Check whether the information they require is presented in the dashboard. Check whether the dashboard has redundant or irrelevant data. How to Resolve This Type of Issue? Incorrect Dashboard Type Check your dashboard's type - Based on the person using this dashboard, should it be operational, analytical, tactical, or strategic? Compare the dashboard type to the intended crowd to determine if it was designed correctly (for example, a "Tactical Dashboard" is aimed towards upper management and will usually present long-term KPIs and high-level metrics). Resolving this type of issue will most likely require a complete dashboard redesign, followed by an entire UAT cycle. Missing/Redundant Information If you've identified that certain information is redundant or missing, consider adding, modifying, or removing widgets from the dashboard. Doing so might be a simple task (especially when removing data). However, it might also end up as a more extensive project requiring a partial redesign of the dashboard and a partial/complete UAT cycle. The "Data Correctness" Test Presenting inaccurate or outdated data is worse than displaying partial or excessive data. Two possible outcomes of stakeholders, basing their decisions on incorrect data, are: Experiencing negative events (such as revenue loss) Compromising the trust relationship between Sisense and the end-users Frequent causes for showing outdated data in your dashboard include: An ETL process that is not run frequently enough An ETL process that often/occasionally fails An ETL process using an incorrect table update behavior (e.g., "Accumulative" where it should be "Full") An ETL process that isn't synchronized with your Data Warehouse A source database being offline for an extended period Frequent causes for showing incorrect data in your dashboard: Wrong data modeling leading to unexpected Many-to-Many relationships Business questions that don't align with the data model (e.g., causing "Random Paths") Unexpected / Missing inheritance of filters Wrong data security configuration The use of multiple data models (in the same dashboard) built at different schedules A good dashboard: Displays precise and up-to-date data How to Test This? To check whether your dashboard has correct data, you'll have to: Assess the different widgets on the dashboard to see the granularity of data displayed For each widget, check what data model it relies on - Use the Usage Analytics "Usage - Builds" dashboard to monitor the behavior of historical builds. Set up a call with one (or more) end-users and/or stakeholders using this dashboard Ask them if they trust the data on the dashboard and if the data refresh frequency is sufficient. How to Resolve This Type of Issue? Showing Outdated Data Ask your stakeholder how frequently they expect the data to be refreshed. Perform the following actions: Check the data model's build frequency - Should the build frequency be modified? Check the build time - Should the data model be optimized? Is the data model too heavy? Check the build success rate - Why are builds failing? Is the system running low on resources? Should "Data Groups" be applied? If using a Data Warehouse (DWH) and an Elasticube - Check the DWH build frequency; Check the synchronization between the DWH ETL finish and the Sisense ETL beginning. If you can't achieve the required build frequency, check whether all widgets require the same data refresh rate - Can a "Live Model" or a "Hybrid Dashboard" be considered? Showing Incorrect Information Ask a stakeholder to generate a report with correct data or ask them to point you to "defective widgets." Perform the following actions: Check the formula behind the figure(s) showing the incorrect data and correct them. Consider the different filters, inheritance behavior, etc. Examine the query path used to calculate each figure (look out for unexpected "Many-to-Many relationships" or "Random Paths") - Use the "Visualize Queries" add-on (Link) to compare Sisense behavior against the expected query path. Run the calculation in the data model using a custom table and a SQL statement that emulates the dashboard's calculation. The "Visual Correctness" Test Data alone isn't enough; it must be visualized correctly. Visualizing data requires using the correct widget type and visual aids to convey the message. For example, say a KPI is showing the company's revenue. See the different visualization options below: Option #1 Option #2 Option #3 Transitioning from option #1 to #2 provides an added value of "Good" / "Bad" Transitioning from option #2 to #3 provides an added value of revenue ranges A good dashboard: Has widgets that are visualized correctly and convey a clear message. How to Test This? To check whether your widgets are visualized correctly, you'll have to analyze each widget individually: Categorize each widget to find out its aim (e.g., show a single figure, compare values, show behavior over time, visualize data to show distribution, etc.) Verify the visualization matches the widget type (e.g., An indicator widget is perfect for displaying a single figure, a line chart is ideal for showing a behavior over time, etc.) Make sure each widget has visual aids (such as conditional formatting) to convey its message clearly - Each widget should tell a small puzzle of the story (which the dashboard should put together to a larger picture) Set up a call with people who are not familiar with the dashboard Ask them to describe each widget (separately) and their conclusion from looking at it. How to Resolve This Type of Issue? Wrong Widget Type Resolve the issue by fixing the visualization type - Use the following chart to help out: Type Use Case Indicator Show a single figure (numerical) Show a single figure and a gauge representing its range Column Chart Show a comparison among different sets of data Track data sets over time Track individual values + Their sum (stacking) Bar Chart Compare many items Track individual values + Their sum (stacking) Line Chart Reveal trends, progress, or changes that occur over time The data set is continuous rather than full of starts and stops Area Chart Displaying absolute or relative (stacked) values over a time period Analyzing a Part-to-whole Relationship Pie Chart One static number, divided into categories that constitute Represent numerical amounts in percentages Table Display RAW granular data Pivot Display RAW granular data Display aggregative data in a table format Scatter Plot Comparing large numbers of data points without regard to time Identify a potential relationship between two variables Calendar Heatmap Show relative number of events for each day in a calendar view Missing Visual Aids Visual aids help the end-users identify "Good" or "Bad" values. Possible visual aids include: Adding colors to numerical labels (e.g., Green figure vs. a red one) Adding background colors to table cells (e.g., Color negative cells red) Adding a line chart representing a threshold (e.g., A red line to indicate a lower threshold) Converting numerical values with text (e.g., Showing a "Revenue increased" label rather than a positive figure) Adding trend lines Adding a forecast range The "Intuitiveness" Test Dashboard intuitiveness refers to the ability of a non-technical person to: Access the dashboard Understand and conclude the dashboard Understand how to customize the dashboard (using filters, drilling, etc.) Easy Access to Data There are two methods of accessing a dashboard: Sisense offers a complete HTML5 platform (a.k.a. Sisense Web Application) that can be white-labeled and customized to meet the company's look & feel (i.e., color pallet, logo, links to internal support and documentation pages, etc.). Sisense offers three different embedding deployment options (conventional iFrames, an embedding SDK, and a complete JS-based embedding solution). Embedding the dashboard (or individual widgets) allows infusing analytics into web pages and applications. Choosing the proper access method will affect how end-users access data and benefit the BI solution. To improve intuitiveness - Make sure to streamline the process of consuming data as much as possible. The need to switch between multiple applications will result in a lack of efficiency, a complex adoption, or even a lack of adoption. Ease of Comprehension The comprehension of individual widgets was discussed earlier. However, is "the whole" greater than the sum of the parts? Does the dashboard tell a story? An excellent example for a dashboard: Widget #1 shows the revenue is low. Widget #2 shows a revenue breakdown per department and points out one department losing money. Widget #3 provides an income/expanse category breakdown and points out non-proportional marketing expenses. Widget #4 provides a detailed transactional income/expanse breakdown and points out the individual expenses Customizability and Interactivity Stale reports tell one story. However, a person looking at the dashboard may want to filter, sort, and pivot the data to tell the same story about a specific segment or individual. The tools for customizing/interacting with a dashboard include: Filtering data to a specific segment of interest (Link) Drilling into a measure to be able to extract more information about it (Link) Adding explanations (Link) and narratives (Link) A good dashboard: Is easy to access, self-explanatory, easy to customize, and play around with. How to Test This? To check whether your dashboard is interactive, you'll have to: Set up a call with one (or more) end-users and/or stakeholders using it Ask them when they use this dashboard and their workflow of consuming its information (e.g., A customer requires the data when building a financial report, and his workflow includes opening another browser tab and logging into the Sisense Web Application). Open the dashboard and ask them to explain it. Try to identify widgets that were over-explained and widgets they skipped. Track the questions you have to ask to understand what you see. Track the amount of "mouse scrolling" they perform when explaining the dashboard flow. Ask them what interaction they have with their dashboard and lay out the tools to increase interaction. Note the tools they are interested in (e.g., extra filters, adding narratives to widgets, etc.) How to Resolve This Type of Issue? Data is Hard to Access Ease of access is easy to measure - Count the number of clicks the user has to make when they want to consume the data in this dashboard - Fewer clicks = Easier to access. If data is hard to access or requires too much "Clicking around": Integrate SSO or WAT to prevent the user from logging in to the Sisense Web Application Embed the dashboard (or a single widget) into the application/webpage Have additional data imported to Sienese - Allowing the user to have a "One Stop Shop" for their entire workflow Enable sending periodic reports to the user's email Enable NLQ to allow the user to interact with Sisense easily (Link) Infuse data into G-Suite applications (Link) Enable Pulse alerts to send push notifications to the user (Link) Data is Hard to Comprehend If your widgets tell the right story, but the complete picture doesn't make sense: Move widgets around to make the story clearer Add visual separators between certain widgets (requires scripting) Reduce the number of widgets by splitting your dashboard into multiple stand-alone dashboards Use add-ons to simplify dashboard navigation (e.g., Accordion) Bring in a UI/UX designer to help visualize correctly Redesign the dashboard to avoid scrolling the mouse The Dashboard isn't Customizable / Interactive Make the dashboard more customizable by allowing the user to: Filter data based on predefined filters Define hierarchies to make filtering more intuitive Integrate filtering abilities into the dashboard (e.g., BloX buttons, drilling options, clickable widget values) Add additional widgets (e.g., Accordion, Switchable Dimensions, Tabber, etc.) Add premium widgets (e.g., Advanced Input Parameters - Link) The "Adoption" Test So you've created a great dashboard; it has all the correct data and visualizations, a person can access it quickly and draw the proper conclusion by just looking at it, but it wasn't adopted. Has the company done enough to drive its adoption (e.g., sufficient training, decommissioning the old dashboard, integrating it into the users' application, etc.)? A good dashboard: Is measured by its adoption and usage. How to Test This? To check whether your dashboard has correct data, you'll have to: Find out who this dashboard's audience is Use the Usage Analytics "Usage - Dashboards" dashboard to monitor who accesses this dashboard and how often. Set up a call with one (or more) end-users and/or stakeholders who are not using this dashboard Find out why this dashboard isn't being used. Find out what alternative ways they use to collect the data How to Resolve This Type of Issue? Once you're sure the dashboard is perfect and the only thing missing is adoption: Involve users in the design meetings and UAT loop to make them feel they are part of the process Create an adoption plan! Get executives to buy-in - Causing them to promote their usage Monitor what people are using the dashboard and which aren't - Target the right people Reeducate your end-users on the benefits of the dashboard and the value of using it Decommission old dashboards and reports Refresh your dashboards from time to time (design, contents, etc.)Color the Row and Value fields of a Pivot Table
This article will show you how to quickly color a pivot table into two sections: by Row and by Value 1. Build your pivot table. 2. In the widget editor, add the widget script below. Information on how to add a widget script here: https://sisense.dev/guides/js/extensions.html widget.transformPivot( { type: ['value','member'] }, function setCellBackground(metadata, cell) { if(metadata.type == 'value'){ cell.style = cell.style || {}; cell.style.backgroundColor = '#9A94BC'; } else { cell.style = cell.style || {}; cell.style.backgroundColor = '#ffe6e6'; } } ); 3. Change the colors by changing the color codes above and you're all set! Note: If you'd like to target different parts of the pivot table, refer to our Pivot 2 API documentation.
