Hey guys, super excited to be diving into Triple Whale Agents with you today. This is truly a game-changer for how you'll analyze data, make decisions, and drive growth for your eCommerce business.

So what exactly are Triple Whale Agents? Think of them as your own personal analysts working 24/7, automating all those repetitive tasks you've been doing manually.

Agents are essentially automated workflows that can pull data from multiple sources, analyze it, and deliver insights and recommendations - all without you having to lift a finger once they're set up.

What makes Agents different from your standard dashboards and reports is that they don't just show you data - they actually process it, compare it, analyze it, and give you actionable recommendations based on your specific business rules.

The core benefits of using Agents are pretty incredible:

First, there's the automation aspect. Those weekly reports you've been manually putting together? Those hours spent comparing performance data across different time periods? Agents can handle all of that automatically.

Second, Agents can synthesize data from multiple sources. They can pull information from your Shopify store, your ad platforms, your email marketing - basically anywhere Triple Whale has access to your data - and bring it all together in one place.

Third, and this is where things get really exciting - Agents can generate actual insights and recommendations. They don't just show you that your ROAS dropped - they can tell you why it dropped, which products or campaigns are responsible, and what you might want to do about it.

Let me quickly walk you through the Agent ecosystem within Triple Whale:

We've got the Agent builder, which is where you'll create and edit your Agents. Don't worry - we'll get into the details of that in later videos.

We have an extensive Agent library with templates you can use right away or customize for your specific needs.

There are multiple ways Agents can deliver outputs - through the feed in your Triple Whale account, via email, through Slack, or directly to Google Sheets if you're using that for reporting.

And one of my favorite features - Agent-powered AI recommendations that live right in your attribution dashboards, giving you instant guidance on which campaigns, ad sets, or ads to scale, reduce, or monitor.

Let me share a couple of quick real-world examples to show you the power of Agents:

Imagine you're a media buyer managing multiple ad platforms. Every Monday morning, you sit down to analyze performance, compare it to previous periods, and make budget allocation decisions. With Agents, you can have that analysis waiting in your inbox when you arrive, complete with recommendations on which campaigns deserve more budget and which ones should be scaled back.

Or maybe you're managing inventory for hundreds of SKUs. An Agent can monitor stock levels, sales velocity, and seasonality patterns to alert you when products need to be reordered, taking into account historical trends and forecasted demand.

Or perhaps you're an agency managing multiple clients. Agents can automatically generate client-ready reports with your agency branding, showing the most relevant KPIs and insights for each client, saving you hours of report preparation time.

In this course, we're going to take you through everything you need to know to leverage the full power of Agents:

In our next video, we'll cover getting started with Agents and Moby, which is our conversational AI interface.

Then we'll walk through building your first Agent from scratch.

We'll dive into media buying optimization Agents to help you make better ad spend decisions.

We'll explore product and channel performance Agents to get deeper insights into what's working in your business.

And finally, we'll cover some advanced techniques and best practices for those ready to take their Agent game to the next level.

I'm super excited to be on this journey with you guys. Agents have honestly transformed how our customers analyze data and make decisions, and I can't wait to show you how they can do the same for you. Let's dive in!

Hey guys, welcome back! Super excited to be diving deeper into Triple Whale Agents with you today. In this video, we're going to get you up and running with the Agent interface and show you how to start leveraging Agents right away.

So let's jump right into the Triple Whale platform. When you log in, you'll see Moby here on the left navigation. This is your starting point for all things AI in Triple Whale.

Now, if you click on Agents here in the nav, you'll see your Agent library. This is where all your Agents live once they're created. You can see I've got several already set up for this shop.

Under "My Agents," you'll see all the Agents currently running for this specific store. And under "Agency Agent Collection," you'll find our library of pre-built Agent templates that you can start using right away.

Before we go further, let's talk about the relationship between Moby and Agents. This is important because they serve different but complementary purposes.

Moby is your conversational interface for asking one-off questions about your data. You can think of Moby as your data analyst who's always ready to answer specific questions. "What were my top 10 products last week?" "How did my Facebook campaigns perform yesterday?" Moby will answer these questions directly.

Agents, on the other hand, are like setting up automated workflows. An Agent will run through multiple steps, pull various data points, compare them, analyze them, and then generate insights based on all that information. While Moby is great for exploring data and getting quick answers, Agents shine when you need regular, multi-step analysis or want to automate reporting.

Let's take a second to talk about using Moby effectively, since it's such a key part of the Agent ecosystem.

The magic of Moby is that you can just type natural language questions. Let me show you a couple of examples:

[DEMO: Type a question like "Show me my top 10 products by revenue for the last 30 days"]

See how quickly Moby pulled that data? Now let's try something a bit more complex:

[DEMO: Type "Compare my Facebook ROAS this week versus last week broken down by campaign"]

The key to getting great results from Moby is being specific about:

And here's a pro tip: if you find yourself asking Moby the same questions repeatedly, that's a sign you should probably turn that into an Agent!

Now let me show you one of the most powerful features in Triple Whale - Deep Dive research. This is where Moby really flexes its muscles.

Deep Dive allows you to ask complex business questions that require Moby to search through your entire data set, looking for correlations and insights. Let me show you:

[DEMO: Type a complex question like "What products lead to highest customer LTV and what channels drive the most sales for these products?"]

When you ask a Deep Dive question, Moby will create a research plan and then execute it step by step, pulling all the relevant data before giving you a comprehensive analysis. This is incredibly powerful for understanding complex relationships in your business.

And guess what? You can turn any Deep Dive analysis into an Agent with just one click, so it can run automatically on a schedule. That's the kind of seamless integration we've built between Moby and Agents.

Now let's look at how to use the pre-built Agent templates in our library. Click back on Agents in the navigation, and then click on the "Agency Agent Collection" tab.

You'll see we've organized our templates into different categories based on use case. For media buyers, we've got campaign analysis, creative analysis, budget allocation. For inventory management, we've got forecasting, stock alerts. For reporting, we've got daily summaries, weekly rollups, and much more.

Let's take a look at one of these templates. I'll click on "Daily Marketing Pulse" here. This gives you a quick overview of what the Agent does and what outputs it provides.

To add this Agent to your shop, just click "Get Agent" and it will be added to your "My Agents" section. From there, you can run it as is, or customize it to better fit your specific needs.

Speaking of customization, let's talk about managing your Agents. Back in the "My Agents" section, you can see all the Agents currently set up for your shop.

For each Agent, you've got a few key management options:

If you want to edit an Agent, just click the three dots and select "Edit Agent." This will take you into the Agent builder where you can modify any aspect of the Agent. We'll cover the Agent builder in detail in our next video.

One of the great things about Agents is the flexibility in where outputs can be delivered. Let me show you the main options:

The Agent Feed is your centralized place to view all Agent outputs. Think of it as your personalized newsfeed of business insights. You can access it directly from the Moby tab.

Email delivery is perfect for reports you want delivered directly to your inbox or shared with team members.

Slack integration lets you send Agent outputs directly to specific Slack channels, which is great for team collaboration.

Google Sheets export is ideal if you're incorporating Agent outputs into existing reporting spreadsheets or if you need to manipulate the data further.

And finally, you can send Agent outputs directly to custom dashboards within Triple Whale, embedding those insights right where you need them.

The last thing I want to cover is how to schedule your Agents. When you create or edit an Agent, you'll see scheduling options at the bottom.

You can set Agents to run:

The scheduling flexibility means you can have daily reports waiting for you each morning, weekly analyses for your Monday planning sessions, or even real-time alerts for critical metrics.

So that's the foundation you need to get started with Agents and Moby. We've covered navigating the interface, understanding the difference between Moby and Agents, using Moby effectively for data exploration and Deep Dive research, finding and using Agent templates, managing your Agents, and setting up scheduling.

In our next video, we'll dive deeper into the Agent builder and walk through creating your first Agent from scratch. This is where things get really exciting, as you'll see how to build completely customized automated workflows tailored to your specific business needs.

I'm super pumped to continue this journey with you guys. See you in the next video!

Hey guys, super excited to be back with you for video three of our Triple Whale Agents course. Today is where things get really hands-on - we're going to build an Agent from scratch together. By the end of this video, you'll understand all the building blocks of an Agent and how to create your own custom automated workflows.

Let's start by understanding the core building blocks of every Agent. No matter how simple or complex your Agent is, it's going to be made up of some combination of these five types of steps:

First, you've got your Get Data steps. These pull raw data from your connected platforms like Shopify, Meta, Google, and so on.

Next, there are Analysis steps, which apply different analytical techniques to that data - things like forecasting, MMM analysis, or even visual analysis of creative assets.

Then you've got Logic steps, which let you create conditional flows - basically "if this, then that" type logic within your Agent.

Fourth are Reporting steps, where you define how you want insights presented, either as plain text or using our GenUI visual interface.

And finally, there are Destination steps, which determine where your Agent outputs will be delivered - to the feed, email, Slack, Google Sheets, or dashboards.

Alright, let's create a new Agent together. I'll walk you through building a basic performance analysis Agent that compares current performance to previous periods.

Let's go to the Agents section and click "New Agent" in the top right corner.

This brings us into the Agent builder interface. You can see we start with a blank canvas here. Let's add our first step by clicking the plus button.

For our first step, we need to get some current performance data. Let's click "Get Data" from the options on the right.

There are several ways to get data in an Agent. We could use:

For most users, Plain English is the easiest way to start, so let's use that.

In this text box, we're going to write a prompt that describes what data we want. Let's keep it simple and type:

"Show me my store's key performance metrics for the last 7 days including revenue, orders, AOV, ROAS, and new customer ROAS"

Now, let's click "Run only this step" to test it and see what data we get back.

Great! We've got our current period data. Now let's add another step to get comparison data from the previous period.

Let's add another Get Data step and write:

"Show me my store's key performance metrics from 8-14 days ago including revenue, orders, AOV, ROAS, and new customer ROAS"

Let's run this step too... perfect!

Now that we have our data, let's add a step to analyze and report on it. Click the plus button again, but this time let's select "Report" and then "Plain Text."

In this step, we're going to tell the AI how to analyze our data and what format we want the report in. Here's a good prompt:

"Create a performance analysis report comparing the last 7 days to the previous 7-day period. Include the following sections:

Format the report in a clear, professional style with headers for each section."

Let's run this step to see the output... and there we go! The Agent has analyzed our data and created a formatted report with all the sections we requested.

The final piece is deciding where we want this report to go. Let's add a destination step by clicking the plus button again and selecting "Choose Destination."

You can see we have several options:

For this example, let's select "Feed" and "Email." For email, we'll need to add the email addresses where we want the report sent.

Now let's click "Finish up" in the top right corner.

Here we'll give our Agent a name - let's call it "Weekly Performance Analysis" - and add a description so we remember what it does.

Below that, we have scheduling options. We can run it:

Since this is a weekly analysis, let's set it to run every Monday morning at 9 AM, so it's waiting in our inbox when we start the week.

Click "Save" and we're done! We've just created our first Agent that will automatically pull performance data, compare it to the previous period, analyze it, and deliver a report every Monday morning.

Before we wrap up, let me show you a few troubleshooting tips for when your Agents don't work exactly as expected.

The most important troubleshooting technique is testing each step individually. If your Agent isn't producing the expected output, go back to the builder and run each step one by one to see where things are going wrong.

Common issues include:

If a step isn't returning what you expected, try refining your prompt to be more specific. For data steps, specify exactly what metrics, time periods, and groupings you want. For reporting steps, be explicit about what analysis you want and how you want it formatted.

Finally, let's talk about editing and refining your Agents. Once you've created an Agent, you can always go back and improve it.

To edit an Agent, go to Shops Agents, find the Agent you want to modify, click the three dots, and select "Edit Agent."

This brings you back to the builder where you can change any step, add new steps, or remove steps you don't need.

One approach I recommend is to start simple with your Agents and then gradually add more complexity as you see what works well for your specific needs.

So that's how you build your first Agent! We've covered the basic building blocks, created a simple performance analysis Agent, set up destinations and scheduling, and talked about troubleshooting and refinement.

In our next video, we'll get more specific and look at creating Media Buying Optimization Agents to help guide your ad spend decisions. This is where Agents can really start driving ROI for your business.

Super excited to continue this journey with you guys. See you in the next video!

Hey guys, super excited to be diving into video four of our Triple Whale Agents course today. In this video, we're going to focus specifically on Media Buying Optimization Agents - and this is where things get really interesting for anyone managing ad spend.

These Agents can fundamentally transform how you analyze campaign performance, make budget decisions, and ultimately drive better ROAS for your business. I've seen these types of Agents save media buyers hours every week while simultaneously improving their decision making. Let's dive in!

Media buying is all about making smart decisions with your ad budget - knowing when to scale campaigns that are performing well, when to pull back on underperformers, and when to maintain what's working.

The challenge is that these decisions require analyzing a ton of data across multiple platforms, comparing performance across different time periods, and applying complex decision rules. This is exactly the kind of work that Agents excel at automating.

Let's build a Media Buying Optimization Agent together that will analyze your campaign performance and make recommendations on budget adjustments.

We'll start by creating an Agent focused on Meta ads optimization. Go to Agents and click "New Agent."

For our first step, we need to get blended performance data for our store. Click the plus sign, select "Get Data," and let's use Plain English:

"Show me my store's blended ROAS and blended new customer ROAS for the last 7 days"

This will give us our overall store performance, which we'll use as a baseline.

Now let's add another step to get our Meta campaign data. Add another "Get Data" step:

"Show me the following metrics for my Meta campaigns using the last 7 days of data: campaign name, ad set name, spend, revenue, ROAS, new customer ROAS, new customer orders, and cost per acquisition. Include only campaigns with spend greater than $50."

That minimum spend threshold is important – we don't want to make decisions based on campaigns with insignificant spend.

Let's test this step... perfect! Now we have all our campaign-level data.

For ad set level, let's add another step:

"Show me the following metrics for my Meta ad sets using the last 7 days of data: campaign name, ad set name, spend, revenue, ROAS, new customer ROAS, new customer orders, and cost per acquisition. Include only ad sets with spend greater than $20."

Great, now we have our ad set data too.

Now comes the key part - setting up the decision rules for our optimization recommendations. Add a "Report" step and let's define our rules:

"Analyze the campaign and ad set data and provide budget recommendations based on the following rules:

For campaigns:

For ad sets:

Format the results in a clear table showing current metrics, the recommendation, and the reasoning behind each recommendation. Sort the tables with the highest performers at the top."

These thresholds are just examples - you'll want to customize them based on your specific business goals and profit margins. The power here is that you can codify your own media buying strategy directly into the Agent.

Let's run this step and see the output...

Awesome! We now have a nicely formatted report with budget recommendations for each campaign and ad set.

Now let me show you another super powerful way to use Agents for media buying - adding AI recommendation columns directly to your attribution dashboards.

This feature gives you instant guidance on which campaigns, ad sets, and ads to scale, reduce, or monitor, right where you're already analyzing performance.

Let's go to Dashboards > Attribution.

In the column selector in the top right, under presets, you'll find "AI Agent Columns." Select that and save.

You'll now see a new column called "AI Recommendation" that will analyze each row and give you a recommendation - scale, reduce, or monitor.

What's really cool is that you can click on any recommendation to see the full analysis behind it. This shows you all the metrics that went into the decision, giving you total transparency into why the Agent is making that recommendation.

By default, this uses some standard thresholds, but you can customize these to match your specific buying strategy. To customize, click the edit pencil when hovering over the column name, then "Configure Column," and you'll see options to edit the Agent that powers these recommendations.

Let's quickly go back to our Agent and add Google Ads optimization. The process is very similar to what we did with Meta.

Add another "Get Data" step for Google campaigns:

"Show me the following metrics for my Google campaigns using the last 14 days of data: campaign name, spend, revenue, ROAS, new customer ROAS, new customer orders, and cost per acquisition. Include only campaigns with spend greater than $50."

Notice I'm using a 14-day window here instead of 7 days - this is because Google campaigns often need a longer data window for reliable analysis due to attribution differences.

Now let's update our report step to include Google campaigns as well. Just add a section for Google with appropriate thresholds - these might be different than your Meta thresholds based on how the platforms perform for your business.

For many media buying teams, having this data in a spreadsheet where they can work with it is really valuable. Let's add a Google Sheets export.

Add a "Choose Destination" step and select Google Sheets. You can either create a new spreadsheet or select an existing one, and specify a worksheet name.

This will send all the raw data and recommendations to your spreadsheet, where you can then use it as part of your media buying workflow.

Finally, let's set up scheduling for our Agent. Click "Finish up" and name our Agent "Media Buying Optimization."

For a media buying Agent, I recommend running it at least weekly, but many teams find it valuable to run it 2-3 times per week to stay on top of performance trends.

Let's schedule it for Monday, Wednesday, and Friday mornings at 8 AM, so it's ready before your media buying team starts work.

Before we wrap up, I want to share a few customization tips for your media buying Agents:

First, make sure you're focusing on the metrics that actually drive your business decisions. For some businesses, that's ROAS or NC ROAS. For others, it might be CPA or CPL. Your Agent should reflect your specific KPIs.

Second, consider adding historical comparison. Instead of just looking at the last 7 days in isolation, compare to the previous 7 days to identify trends.

Third, you can add more sophisticated logic, like different thresholds for different campaign types. Maybe brand campaigns have different targets than prospecting campaigns.

Fourth, consider adding creative analysis. You can have your Agent identify which creative assets are driving the best performance, which can inform both budget decisions and creative strategy.

So that's how you build Media Buying Optimization Agents! We've covered building a comprehensive Agent that analyzes campaign and ad set performance, provides specific budget recommendations, adds AI recommendations directly to your attribution dashboards, and exports everything to Google Sheets for your team's workflow.

These types of Agents have been absolute game-changers for our customers who manage significant ad spend. They save hours of analysis time each week and often lead to better performance as well.

In our next video, we'll explore Product and Channel Performance Agents to help you get deeper insights into what's working across your entire business. Super excited to continue this journey with you guys - see you in the next video!

Hey guys, super excited to be back with you in our Triple Whale Agents masterclass. Today we're diving into some more advanced use-case Agents - powerful tools that will give you unprecedented insights into what's driving revenue in your business.

So whether you’re looking to understand which products are performing well, analyze your marketing creatives, or forecasting your inventory, we’ll show you how Agents can automate this analysis to deliver consistent, data-driven insights without the manual effort. Let's dive in!

Let's start by building a Product Performance Analysis Agent that will identify your top performers, underperformers, and seasonal trends.

Go to Agents and click "New Agent."

For our first step, we'll get current product performance data. Click the plus sign, select "Get Data," and let's use Plain English:

"Show me the top 20 products performance for the last 30 days in terms of revenue, including product title, variant title, units sold, revenue, profit margin, and number of new customer purchases. Include only products with at least 1 unit sold. Sort by revenue descending."

Let's run this step... Great! Now we have our current product performance data.

Let's add another step to get historical comparison data:

"Show me the top product in terms of revenue performance for the same 30-day period last year including product title, variant title, units sold, revenue, profit margin, and number of new customer purchases. Include only products with at least 1 unit sold. Sort by revenue descending."

Perfect. Now we can see year-over-year performance.

For seasonal analysis, let's add one more data step:

"Show me monthly product sales for the last 2 years broken down by product title and month. Include revenue and units sold."

Excellent. Now we have all the data we need to analyze product performance and seasonality.

Let's add a "Report" step to analyze this data:

"Create a comprehensive product performance analysis with the following sections:

Let's run this step... Amazing! We now have a comprehensive product analysis that automatically highlights top performers, growth trends, seasonal patterns, and strategic recommendations.

Building on our product analysis, let's add an inventory forecasting component to predict stock needs and create alerts for potential stockouts.

Add another "Get Data" step:

"Show me current inventory levels for all products including product title, variant title, current stock quantity, and average daily sales over the last 30 days."

Now let's add a "Forecasting" step. Select "Analyze Data" and then "Forecasting."

In the forecasting step, we'll configure it to:

Let's add another "Report" step to format the inventory alerts:

"Create an inventory alert report with the following sections:

This inventory forecasting Agent can be a huge time-saver for operations teams and helps prevent costly stockouts or excess inventory.

One of the most powerful applications of Agents is analyzing creative performance. I want to take you into an agent our team built that leverages computer vision to analyze creative performance on meta.

Add a "Get Data" step:

Show me the 10 FB ads with the largest decrease in pixel NC ROAS between two weeks ago and last week. For each ad, please give me the ad name, ad ID, copy, creative (images only), as well as spend and pixel conversions, ROAS. Also give me the NC ROAS from two weeks ago and the NC ROAS from last week. Only use ads where pixel conversions is at least 5. For attribution model please use linear all.

Now add an Analyze "Computer Vision" step. This allows the Agent to analyze visual elements of your creative assets.

Configure the vision step to:

Here’s my prompt:

Analyze the images in each ad and then give me a table where the first column is the ad name, the second column contains a description of the image, and the third column contains a comma separated list of colors contained in the image

Add a "Report" step for the creative analysis:

"Create a creative performance analysis with the following sections:

This creative analysis can provide invaluable insights for your creative team, helping them understand what's resonating with your audience.

So that's how you build Agents! We've covered:

These Agents give you unprecedented visibility into what's driving your business performance, automatically surfacing insights that would take hours or days to compile manually.

In our next video, we’ll dive more deeply into Deep Dive and Build Mode and take your Agent game to the next level. See you there!

Hey guys, super excited to be back with you for our next installment of our Triple Whale Agents course. Today we're going to focus on two incredibly powerful features that can take your analysis to the next level - Deep Dive Research and Moby's Build Mode.

These tools are absolute game-changers for answering complex business questions and creating sophisticated Agents without having to manually configure every step. I've seen teams unlock incredible insights using these approaches, often discovering patterns and opportunities they didn't even know to look for. Let's dive in!

Let's start with Deep Dive research - one of the most powerful but sometimes overlooked features in Triple Whale.

While regular Moby queries are great for straightforward questions with direct answers, Deep Dive is designed for complex, multi-layered business questions that require comprehensive analysis across your entire dataset.

Here's what makes Deep Dive different - when you ask a Deep Dive question, Moby doesn't just try to answer it directly. Instead, it creates a research plan and executes it step by step, pulling relevant data from multiple sources, analyzing it from different angles, and then synthesizing everything into a comprehensive answer.

Let me show you how it works. Let's go to the Moby chat interface and click on the "Deep Dive" button in the bottom left.

Now let's ask a complex business question:

"What factors are driving the recent changes in our ROAS, and how do these factors differ across our top product categories?"

When we hit enter, Moby first confirms our research question and presents a proposed plan. This plan shows all the steps it's going to take to answer our question.

[SHOW PLAN ON SCREEN]

Look at the depth of this plan - it's going to analyze our ad performance, customer behavior, conversion rates, product performance, and more. This is the kind of comprehensive analysis that would take a human analyst hours or even days to complete.

Once we approve the plan, Moby goes to work, executing each step and collecting data. This might take a few minutes depending on the complexity of the question and the amount of data being analyzed.

And here's the result - a comprehensive analysis that not only answers our question but provides supporting data, visualizations, and actionable insights.

[SHOW DEEP DIVE OUTPUT]

The power of Deep Dive is that it can surface insights and connections you might never have thought to look for. It's like having a team of data scientists working on your business questions around the clock.

The key to getting valuable results from Deep Dive is asking the right kinds of questions. Here are some tips:

First, focus on "why" and "how" questions rather than just "what" questions. Instead of asking "What's our best performing channel?", try "Why is our Facebook ROAS declining while TikTok ROAS is improving?"

Second, specify the business context and why you're asking. For example: "We've noticed a drop in conversion rate over the last month. What factors might be causing this and how do they relate to our recent product launches?"

Third, make your questions specific enough to be actionable but broad enough to allow for discovery. Too narrow, and you'll miss unexpected insights. Too broad, and you'll get surface-level analysis.

Some great examples of Deep Dive questions include:

Now let's talk about another incredibly powerful feature - Build Mode in Moby. This allows you to create complex Agents through natural conversation, without having to manually configure each step in the Agent builder - and, simultaneously, iterate on your desired report output.

To access Build Mode, go to Moby and click Build Mode, then just type "build me an agent that..." followed by your description of what you want the Agent to do.

Let me show you an example:

"Build me an agent that analyzes our top-performing products each week, compares them to the previous week and same week last year, identifies significant changes in performance, and wraps that into a beautiful visual report with recommendations."

When we hit enter, Moby will begin a conversation to clarify exactly what we want this Agent to do.

[SHOW CONVERSATION]

Moby is asking clarifying questions about which specific metrics we want to include, what thresholds we consider significant for performance changes, and who should receive the report.

As we respond to these questions, Moby is building out our Agent in the background, creating all the necessary steps - data pulls, comparisons, analysis, reporting, and scheduling.

When the conversation is complete, Moby will present us with a fully configured Agent that's ready to run. We can review it, make any final adjustments, and then activate it with a single click.

This conversational approach to building Agents is not only faster but often leads to more comprehensive Agents because Moby can suggest elements you might not have thought to include.

When using Build Mode to create Agents, here are some best practices:

First, start with a clear goal statement. Be specific about what analysis you want the Agent to perform and what output you expect.

Second, include key details in your initial request - metrics that matter, time periods to compare, output format, and scheduling preferences.

Third, be prepared to provide clarification. Moby will ask follow-up questions to ensure it's building exactly what you need. The more specific your answers, the better your Agent will be.

Fourth, review the proposed Agent carefully before activating it. While Build Mode is very powerful, you might want to make small adjustments to fine-tune the Agent for your specific needs.

One of my favorite power moves is combining Deep Dive and Build Mode to create automated insight pipelines. Here's how:

First, use Deep Dive to explore a complex business question and discover valuable insights.

Once you've found an insight pattern that's valuable to track over time, click the "Create Agent" button at the top of the Deep Dive report.

This will automatically convert that one-time analysis into a recurring Agent that will continuously monitor those factors for you and alert you to important changes.

You can then customize the scheduling, outputs, and alert thresholds to fit your workflow.

This approach lets you go from a question, to a discovery, to automated monitoring in just a few clicks - it's an incredibly powerful workflow that can transform how you manage your business.

Whether you're using Deep Dive research or Build Mode, the quality of your prompts significantly impacts the quality of your results. Let me share some advanced prompt engineering techniques:

Structure and formatting are crucial. Break complex requests into clear sections with explicit instructions for each part. For example:

"Analyze our Facebook campaign performance with the following sections:

Including context and business rules helps produce more relevant insights. For example:

"Our target ROAS is 2.5, but we're willing to accept lower ROAS for new customer acquisition campaigns. We consider a CPA under $30 to be good performance. We typically reallocate budgets on Mondays based on the previous week's performance."

For multi-step analysis, clearly outline the analytical process you want to follow:

"First, identify our top performing products by revenue. Then, analyze which marketing channels drive the most sales for these products. Finally, recommend budget adjustments to maximize revenue from top products."

And always be specific about output formats:

"Present the results as a concise executive summary followed by supporting data in clearly labeled tables. Include visual charts for key metrics and highlight critical insights in bold."

Let's finish by talking about how to set up automated reporting workflows using these advanced techniques.

The ideal approach is to create a system of complementary Agents that work together to deliver the right insights to the right people at the right time.

For example, you might have:

For each audience, customize not just the content but also the format and delivery method:

You can also create cascading workflows where one Agent triggers another based on specific conditions. For example, if a daily monitoring Agent detects a significant performance drop, it could trigger a Deep Dive analysis Agent to investigate the causes.

That covers Deep Dive Research and Moby's Build Mode - two incredibly powerful tools that can transform how you analyze your business and create Agents.

We've covered how to formulate effective Deep Dive questions, how to use Build Mode to create Agents conversationally, best practices for prompt engineering, and how to set up automated insight workflows.

These approaches allow you to spend less time configuring reports and more time acting on insights. They enable you to go from a business question to automated monitoring in just a few minutes, and they help you discover connections and opportunities you might never have thought to look for.

In our next and final video, we'll explore Creative Analysis Agents and advanced techniques for power users. Super excited to continue this journey with you guys - see you in the next video!

Intro

Hey everyone, and welcome back to the next video in our Triple Whale Agents Masterclass. Today, we’re diving into one of the most important skills for long-term success with Agents—troubleshooting and optimization. Even the most experienced Agent builders run into occasional issues. Maybe you're seeing a "No Data" error, missing expected metrics, or struggling with formatting in your final report. These challenges are completely normal when working with a flexible, powerful tool like Agents.

In this video, I’ll walk you through the most common troubleshooting scenarios we encounter. I’ll share a step-by-step methodology for diagnosing problems and demonstrate how to optimize your Agents for consistent, high-quality performance.

Let’s start with one of the most frequent issues: the dreaded “No Data to Show” message.

This typically means your query is too restrictive. For example, let’s say you're filtering for campaigns with over $5,000 in spend and names containing "CBO." If no campaigns meet both criteria, the Agent returns no data. Relaxing your filters and rechecking your date range often solves the issue.

Another potential cause? Typos or mismatched naming conventions. For instance, if your structure uses “ABO” but your query references “CBO,” the Agent won’t return results. Correcting the naming term usually resolves this.

Also, special characters—like vertical bars (|)—in naming conventions can cause SQL errors. Try simplifying the query first. Once it runs successfully, you can gradually reintroduce those characters to test whether they’re the cause of the issue.

Sometimes your Agent returns data—but it doesn't match what you expected. This is usually due to one of three issues:

Let’s look at a case: An Agent displays ad spend and revenue using Triple Whale Pixel data, but the dashboard shows different values. A closer look reveals the Agent uses the default "Triple Attribution" model, while the dashboard uses Linear Attribution. Changing the dashboard to match the Agent—or adjusting the prompt to specify the attribution model—resolves the inconsistency.

Your Agent returned the right data, but the final report is hard to read. Tables might be missing headers, charts may not appear, or the layout isn’t actionable.

This usually happens when the reporting prompt is too vague. Instead of writing, “Make a nice report with the above data,” try something like:

“Present the performance metrics at the top, followed by insights and recommendations at the bottom. Format everything into a clean, readable table with clear headers and color-coded highlights.”

The more structured your prompt, the more consistent your report output will be—especially important when rerunning the Agent frequently.

If your Agent isn’t delivering reports to Slack, Google Sheets, or email, it might be due to integration issues. Being logged into Triple Whale doesn’t guarantee that permissions are set up.

Head to Settings > Integrations and ensure the correct Google account or Slack workspace is properly linked. Only then will those options appear when you add a destination step in the Agent builder.

One of the most effective troubleshooting techniques is to run each Agent step individually. This isolates where the error occurs.

For example, let’s say we’re working on “Ahmed’s Report” and it keeps erroring out. When opening it in the builder, we see that the report includes tens of thousands of rows. The volume exceeds token limits. Ask yourself: Do I really need that much data?

Try limiting the scope:

This improves performance and increases the chance of a successful run.

Let’s say you're asking for “creative tables” from Meta, but get an error. The system might not understand what “creative tables” means.

Instead, refine your prompt:

“Show me ad performance grouped by image URL from Meta over the last 30 days. Include spend, ROAS, and new customer ROAS.”

Now the Agent can generate a useful preview table with relevant data. To reduce token usage, add a spend threshold (e.g., over $100) to filter out low-spend assets.

You can instruct the Agent to group data by naming conventions. For instance:

“Group creative performance into four segments: shorts-video, shorts-image, non-shorts-video, and non-shorts-image.”

Be specific. Define what to look for (e.g., naming patterns) and how to format the output. This produces a more meaningful analysis.

If you’re writing in plain English and hitting errors, Mobi’s build mode is a great tool. You can:

If your SQL breaks, ask Mobi to fix it by pasting the query and saying “fix this query.” Mobi understands Triple Whale’s data infrastructure and can revise your query accordingly.

Specificity matters. Replace vague prompts like:

“Analyze our campaigns and tell me what’s working.”

With:

“Analyze campaign performance over the last 7 days. Define top performers by ROAS > 3, underperformers by ROAS < 1. Show spend, ROAS, and NC ROAS. Sort with top performers on top. Use green for top performers, red for underperformers.”

You can even break this down into markdown sections to improve readability and prompt clarity.

Sometimes it’s more effective to break large, complex Agents into multiple smaller ones. Each Agent:

Then use the Run Sub Workflow option to combine them in a larger Agent. Just click the plus sign, add logic, and select the sub-agent you want to include. Make sure the final report integrates the insights from all sub-agents.

Don’t troubleshoot alone. Join our Narwhal Nation Slack Community, where you can:

You can also reach out via live chat or the SQL team. When doing so, be specific: include screenshots, a link to the Agent, and what you’ve already tried.

That wraps up our Agents Masterclass. You’ve now learned:

Remember, Agent building is iterative. Start simple, test, learn, and refine. I’m excited to see what you build with Triple Whale Agents. Thanks for joining us—and happy building!

Let me know if you'd like this delivered as a downloadable PDF script, repurposed for a teleprompter, or divided into separate training modules.