Data Discovery Workshop for Jet Reports

The Problem

Victoria, the Report Designer doesn't know where to find the field she is looking for. She knows what type of information she needs, but not the field name or table name per se. Victoria needs a way to more quickly and easily find the field that will produce the data she needs.

The Activities

1. Solution Brainstorming - Individually, everyone writes down as many solutions to the problem that they can think of, off of the top of their heads.
2. Sketching - Forming cross-functional pairs, participants choose one or more ideas to sketch on a whiteboard.
3. Present & Design Critique - Pairs present their ideas to the group, and all participants have a chance to contribute comments and questions.
4. Impact & Effort Evaluation  - Place ideas on a matrix to estimate their relative impact and effort, in order to clarify product prioritization decisions

Results of Participant Brainstorming & Sketching

Metadata

Metadata could give users a broader set of information to search, sort, and filter by. It could also be used to present the users with less information to dig through. This meta-data may apply to reports, tables, and fields.

Business Categories

Within metadata, the concept of 'Business Categories' was brought up multiple times. Metadata could help to organize reports, tables and fields in a way that is already familiar to the user, like by business function or role. Scott suggested inferring business categories from the role centers in NAV. Some examples of potential business categories included Sales, Finance, and Inventory.

Amy and Hughes suggested that the NL function could be filtered by business category. For example, if the user opened a "NLSales()" function, the function wizard would only display tables and fields relevant to a sales report.

Image: Hughes and Amy - Filter Tables by Type, and NLSales()

Popular Tables & Fields

Another type of metadata would indicate which tables and fields are most commonly used in reports. This would limit the set of total information available to choose from to just information the average user is likely to choose from. One group suggested this could work like the 'Favorites' feature, but that this classification could be pre-defined by Jet Reports.

Organize by Type

Matt and Sarah suggested that fields could be filtered by data type, so if the user knew that what they were looking for was a text, number, or date field, they could filter to see fields of only that type.

Hughes and Amy suggested that tables could be organized by whether they were classified as a base table or a transaction table.

Image: Matt and Sarah - Filter Fields by Type

Custom Metadata

Multiple groups stated that it would be good to pre-define metadata, but let users customize the metadata, like for example, to create their own business categories. Brian suggested that users could also modify which tables and fields belonged in which business categories.

Wireframe

Below is a mock up illustrating how the various types of metadata could be used in the Browser or Jet Function Wizard.

Artificial Intelligence

With AI, the software obscures the complexity needed to produce a report so that the user can communicate in human or business terms, instead of in database terms.

Natural Language Search

The concept of natural language search was primary in two groups' ideas (Matt and Sarah, as well as Sam and Bryan). In a natural language search, the software would be able to read questions worded however the user writes normally, and would make intelligent guesses about what report that user wants.

Here were some examples from the brainstorm:

• "How many customers bought pink flamingos last month?"
• "What's in our Australian warehouses right now?"
• "I want sales by month by salesperson"
• "I need a fixed asset report that shows depreciation by month"
• "I want a balance sheet for the current fiscal year"
• "I want an A/R aging report"
• "Outstanding invoices with a due date next month"
• "I want to see item sales by category, grouped by customer & territory"

Image: Matt and Sarah - The "Magic Box"

Another idea related to this was displaying web/help results so the user could go learn how to do the thing they are requesting. This would search all of our web properties and bring back relevant results. 

Clarifying/Refining Questions

Sam and Bryan commented that the users may sometimes ask questions that are too vague for an AI to produce a useful report. In this case, they suggested, the AI could ask the user to make clarifying choices, for example, if a user types "I want a balance sheet", the software could respond "What year do you want?".

Image: Sam & Bryan - The "Digital Jet Guru"

Scott and Brian described another use for the software asking questions. Instead of clarifying natural language search, these series of questions would help the software make decisions to structure a report.

For example, it could ask a series of questions like:

• Do you want to make a list? (Y/N)
• Of what? (Customers;Documents;Items)
• Do you want to make another list? (Y/N)
• Of what? (Document No., Item)...etc
• What additional data do you want to show in this list? (Name;City;State)

Image: Brian & Scott - The "Jet Application"

Templates

Ideas that involved artificial intelligence implied that the software could make use of templates to make intelligent guesses about how the report should be structured, which fields are related and should be included together, and common row and column formats based on the type of report. 

Live Preview

Ideas that involved a report building tool suggested that the user should be able to view a preview to get timely feedback on their choices, and be given a chance to change their choices.

Brian and Scott said that to be able to do this report building effectively, the application would need to live outside of Excel, and include the option to send the resulting report to Excel.

Linking Tables

One major issue related to finding data involves finding data that lives on a different table than the table which contains the first field the user selected.

Highlighting Foreign Keys

Matt and Sarah suggested indicating within a list which fields could serve as a key to link tables together. 

Automatically Offer to Include Related Fields from Other Tables

Bryan suggested that the software could abstract table linking one degree further by suggesting fields from other tables that are commonly used in combination with the field the user just selected, and offer to add these, taking care of linking the tables on the back end.

Other ideas

These ideas were generated during brainstorming, but didn't appear in sketches or discussions.

• Search filtering in the browser
• Right click to learn about a table in browser
• Add-in to NAV to show a field's table there
• Show related tables in database when a table is selected
• Place to look up database schema and see table and field relationships
• Help chat
• Generate a Jet Function for a selected field in NAV
• In Browser, display the NAV page where the table or field is used
• In-product tutorial
• Export table/field info from browser

Results of Impact/Effort Prioritization

In addition to evaluating the impact and effort of these ideas, we identified themes for ideas, which are color-coded below.

The Participants

Facilitators

• JD
• Loren

Development

• Matt
• Sam
• Hughes
• Kim
• Scott

Customer Engagement

• Brian
• Bryan
• Sarah
• Amy
• Josh