Talk to Data: The Prompting Guide for Moby

The next generation of Triple Whale is here.

Don't worry, all of the dashboards you know and love are still here. But now, Triple Whale is more than just dashboards.

Over the last year, we worked with hundreds of customers and partners to create the smart data platform.

That meant:

• Rebuilding Triple Whale from the ground up
• Creating a unified data language for ecommerce
• Building a full enterprise data platform (data warehouse, ETL)

It’s more flexible, more accurate, and more powerful. But, that’s just the tip of the iceberg. 

All this hard work built the foundation for Moby, the only AI that understands your business.

Moby understands your data and can convert natural language into SQL queries better than any other LLM in the world. And then, it can help you analyze that data to make better decisions.

Enough of the big fancy words. Let’s show you how to use Moby.

The Basics of Talking to Moby

When Meeting Moby for the first time, you’ll see this screen (without the red boxes):

There are 4 things to bring your attention to here:

1. Under the Triple Whale logo, there are two buttons: “Prompt Guide” and “Getting Started”. Clicking on either of these is a great way to get familiar with Moby and build your understanding of what Moby is capable of.
2. Above the chat box, you will see four suggested prompts. These are randomly pulled from our verified prompt library, but as you start typing a prompt, the shown prompts will update to better reflect the key words you’re typing. 
3. On the left side of the screen, you will see a list of categories and prompts. These are all verified prompts that you can use to explore your data. You can use the search box as well to find a prompt that suits your question.
4. Above the verified prompts, you will see 3 buttons:some text
   • Prompts: where you can explore verified prompts.
   • History: where you can view your chat history.
   • Sequences: where saved sequences will appear. (A sequence is a series of prompts, similar to a conversation, that will help you get deeper analysis and insights than just prompts alone.
     Now, let’s chat with Moby.

How to Run Your First Prompt

Running your first prompt can be intimidating. 

Let's start with something simple:

“Give me my new customer revenue over the last 7 days from Shopify.” 

And just like magic, Moby’s got your answer: the new customer revenue over the last 7 days from Shopify was $529,341.78. 

After you’ve run your first prompt, you’ll see new buttons pop up on the screen:

Here is what the new buttons can do for you:

Best Practices for Prompting Moby

Writing a quality prompt is key to getting the most value out of Moby. Here are some best practices for writing good prompts:

1. Define your need: Start by clearly identifying the specific information you’re looking for, such as total sales, new customer counts, or channel specific ad spend.
2. Choose your data source: Specify whether your data will come from Shopify data, ad platform data, First-party pixel data, Web analytics, Email/SMS, product analytic, customer journey, and more. Moby will interpret your prompt to choose the best data source, but clarifying can only help.
3. Specify the time frame: Clearly define the time period for your query, such as "last 7 days," "yesterday," or a specific quarter to ensure the data is relevant and actionable.
4. State your output preference: Decide how you want the information presented—whether as a list, a summary, broken out by day or aggregate, in tables or charts, to suit your analysis needs.
5. Craft your prompt: Combine the details you've outlined into a clear and straightforward prompt. For example, "Show me my total sales from Shopify shop data for last month broken out by day."
6. Review and refine: After receiving your data, review it to ensure it meets your expectations. If necessary, refine your prompt to improve accuracy and relevance.

Tips for Success:

• Keep it simple: Focus on one query at a time to avoid overwhelming yourself and to ensure data accuracy.
• Be patient: Crafting effective prompts is a skill that improves over time. Learn from each attempt to refine your approach.
• Use examples: Refer to provided examples as a starting point and adjust them as needed to fit your specific requirements.

How to Ask Follow-up Questions

Let’s start by asking one of the suggested follow-up questions. These are automatically generated by Moby:

“What is the ROAS for Meta Ads over the last 12 months by month?”

This question adds more data to our conversation, which provides Moby with more context. We can continue by crafting our own question:

“What is my revenue and new customer revenue for Meta Ads over the last 12 months by month.”

My grammar wasn’t perfect, but that’s okay!

Now that we added some data to our conversation, let’s ask Moby to analyze that data:

Here’s where it starts to get interesting! Not only can Moby help you get and visualize your data, Moby can help you analyze that data to make better decisions. Just imagine how Moby can help you analyze your ad performance on the campaign level!

Conversations with Moby

To get the most value out of Moby, you should think of your interactions as a conversation. Instead of just telling Moby to do something for you, It’s important to start your chat by providing context. 

Moby becomes increasingly powerful as you feed it more context, more data, and ask follow up questions around analyzing the data.  

Perfect. Now let’s add some more data:

And lastly, some high level data:

Now that our chat has the data and context, we can ask some questions to analyze it:

“Analyzing all the data above. What are some of your key observations around channel spend and performance for the last 12 months?”

“Based on the data above, how do you recommend I allocate my budget for the next 12 months?”

“What months showed the strongest ROAS for Meta, TikTok, and Google Ads respectively?”

I’ll let you explore those for yourself!

Code Words for Prompting Moby

As you can see above, Moby does its best to interpret what kind of prompt you are asking, and uses the right tool to get your answer. But knowing some specific code words and using them effectively will help with prompt accuracy.

Here are some examples of key phrases you can use to prompt Moby more efficiently:

1. Give Me / Show Me

• Use case: to request specific data, information, or examples.
• Example: "Give me my Meta ROAS and CPA by campaign for the last 30 days."

2. Forecast

• Use case: to predict future trends, sales, or any other metric.
• Example: "Forecast my total revenue for the next 6 Months. Use the last 30 months of data."

3. Web / Search

• Use case: to access information that is not stored in local databases but can be found on the internet.‍
• Example: "Search the web for the latest studies on consumer behavior in online shopping."

4. Help Center

• Use case: to get detailed explanations or help understanding how the Triple Whale product works. This code word will search Triple Whale’s Knowledge Base. 
• Example: "Help center: explain how NC ROAS is calculated."

5. Using Code

• Use case: to perform calculations, generate graphs, or conduct analysis that requires a script in Python.
• Example: "Using code, add a column for '2024 target revenue' = 2023 Revenue * 1.2."

6. Revise/ Modify

• Use case: to request changes or corrections in the data, queries, or scripts provided.
• Example: "Revise the SQL query to include data from the last 12 months instead of 6."

7. Analyze / Analysis

• Use case: to dive deeper into your data.‍
• Example: "Analyze the above data and share any key insights to guide my decision-making."

8. Retrieve / Pull Data

• Use case: to pull specific data from a database or a dataset.
• Example: "Retrieve all $100+ orders from yesterday"

9. Compare / Comparison

• Use case: to compare data sets, metrics, or performance across different parameters.
• Example: "Compare the performance of our two best selling products over the past year."

Data Dictionary

We get it. It can be difficult to understand which data Moby has access to. That’s why our team has been working hard to build a Metrics Library to show you the key data points you can access with Moby, where they come from, and how you can query them.

How to Build Visuals For Boards

When you run a prompt that pulls data from the warehouse, Moby will create an out-of-the-box visualization for you.

But you have the power to customize how this data is visualized!

In this example, Moby generated a table for the “Daily Meta ROAS and AOV analysis”. It might be easier to visualize trends in the data by changing this to a line graph: (Like that BIG spike in ROAS at the end of March)

Using the widget editor, we can customize how each axis is visualized. 

You can continue to format and customize your widget to your liking. Once you’re happy with how the widget looks, click Save.

You didn’t do all of that work for nothing - add the widget to any dashboard to access it later!

This is where the power of Moby begins to intersect with all the dashboards you know and love. Moby makes Triple Whale flexible and unlocks the ability to customize and create new dashboards. 

Did we mention that Moby also enables you to chat directly with your dashboards to get insights? We’ll save that for the next blog… 

Conclusion

We built the next generation of Triple Whale to be something you love, so you can stop doing things that you hate. While there’s a bit of a learning curve (see above), once you get familiar with the basics of Moby, you won’t want to go back to clunky dashboards.

Ready to join the future? Give Triple Whale a try today!