Preparing for Spotter
To fine-tune Spotter’s understanding of your data, you must coach it to recognize your organization’s business terms and common use cases before your users start using it. This article documents the order in which you’ll want to prepare Spotter before you begin coaching.
Step 1: Identify who you’re coaching Spotter for
Start by thinking about your target personas (the types of users) and the kinds of questions they typically ask. Spotter is most effective when you coach it based on real user needs.
| Team | Role | Use cases |
|---|---|---|
Sales |
Sales Ops, Regional heads |
Pipeline metrics, conversion, regional performance |
Marketing |
Campaign managers |
Spend ROI, campaign attribution, top-performing channels |
Customer success |
CSMs |
Churn risk, product usage patterns, account health scores |
| Focus on one team or user group at a time, preferably one that has urgent data needs and could see high value from using Spotter. |
Step 2: Collect real questions and group them
Gather the actual questions your target business users are asking. This step is crucial because analysts, who are typically more familiar with the data set’s structure and terminology, may ask questions very differently than business users.
Collecting real questions directly from business users helps you understand their natural language, the specific terms they use, and the context behind their queries. This collection will form your Spotter Coaching Scope. Discovering how your business users actually phrase their questions early on will help you develop a clear understanding of how much coaching Spotter will require and precisely what aspects you need to focus your coaching efforts on.
| Group | Sample questions |
|---|---|
Pipeline metrics |
“How much pipeline do we have for this quarter?” |
Conversion funnel |
“How many deals were won last month?” |
Team comparisons |
“Which sales rep has the highest win rate?” |
If you’re unsure how to begin categorizing these questions or want a structured way to organize them for different use cases, you can use this Spotter coaching scope template to get started.
Step 3: Optimize your data model
Once you have a clear understanding of your target users and the real questions they ask, the next crucial step is to ensure your data model is optimized to answer these questions effectively. This might involve creating a new data model specifically for this use case, or optimizing an existing one for which you have already created Liveboards and Answers.
| Associated content such as Liveboards on a Model help “warm up” Spotter, as it assists in column association and helps Spotter pick columns more accurately for topics on which content has been created. For more information, see Spotter Model readiness. |
Step 4: Test Spotter
After you’ve created or optimized your data model, perform an initial round of testing before diving deep into creating specific Spotter coaching like reference questions.
The purpose of this testing phase is to take a baseline assessment of how well Spotter can answer your collected questions with just the optimized data model, and to identify gaps or questions that Spotter struggles with or gets wrong. This helps determine if the issue lies with the data model itself (as in, you might need to tweak column names and synonyms as part of optimizing the data model), or if you need to add explicit Spotter coaching.
To test how Spotter responds to your Model, take a representative sample of the real questions you collected, and ask these questions in Spotter against the data model. Review the answers generated by Spotter, and note which ones are accurate, which are close, and which are entirely off. This review informs whether you need to revisit data model optimization or if you’re ready to proceed to targeted Spotter coaching for the problematic queries.
Related information
