How I Got AI To Audit My Klaviyo Flows In 5 Minutes

You know what's not a vibe?

Wading through Klaviyo flows in analytics view, trying to piece together whether your Welcome Flow is actually doing anything. Or trying to compare which two version of a flow was actually better.

You know what is a vibe?

Having AI spit out a full analysis of all your flows—top performers, what’s stale, what to fix, and even a dashboard to show your boss.

This takes 4 steps. About 5 minutes.

Let's go.

Step 1: Export your Klaviyo flow data

We can’t audit what we don't have data for, so we need to get our flow data!

Here’s how to get the data we need:

  1. Log into Klaviyo

  2. Go to Flows

  3. Hit OptionsExport analytics

Then for export settings:

  • Last 12 months

  • Aggregate monthly

  • Include A/B tests (yep, even the messy ones)

Download the file. Done.

Step 2: Give the data to ChatGPT

This is where AI becomes your marketing data analyst, your strategist, and your e-mail expert.

We’re going to upload that CSV to ChatGPT, drop in a prompt, and let it go full data-marketing-strategist-nerd for you. It’s going to:

✅ Analyze every active and historical flow
✅ Break down performance by month
✅ Look at the conditionals in each flow (up to 3rd email sends for now)
✅ Check for best practices (like A/B tests)
✅ And give you 5 actual recommendations

Here’s how to make that happen:

  • Go to ChatGPT

  • New chat → use either o1 or o3-mini. Do not use 4o.

  • Upload the CSV you just downloaded from Klaviyo

  • Copy & paste the prompt below

  • Hit enter

Done right, it'll give you beautiful pages of analysis that look something like this:

What the prompt is actually doing

For those who want to learn more on the prompt engineering side, here's a quick rundown of what we have the prompt doing for you.

  1. It sets context: tells ChatGPT what the data is

  2. It defines its role: makes it act like a DTC email pro

  3. It tells it how to analyze flows and emails, month-by-month

  4. It asks for real recommendations based on the data

  5. It locks in an output format so everything stays organized and readable

The analysis prompt:

# Context 
Use the attached CSV (or pasted data) of the email automations (flows) setup for a brand, as well as how they have performed. 

# System 
You're an expert at direct to consumer email marketing. You are going to analyze email automation flows that are setup, and how they have performed based on their analytics. You're trying to determine current flow performance, including performance of conditional or A/B steps. When you’ve finished your flow analysis, share your results in the OUTPUT_FORMAT. Consider recommendations based on #RECOMMENDATION_ANALYSIS

# Flow Analysis
As part of your analysis, you're going to detail: 
1. What flows are live currently 
2. What flows were live but aren't anymore and how they performed 

For each flow, you're going to cover the following stats, showing month over month for the last 12 months: 

## Flow 
- Total Revenue 
- Total Recipients 
- Total # of Deliveries 
- 1st Step Conversion Rate (whether the first message is where it converted)
- 2nd Step Conversion Rate  (whether the second message is where it converted)
- 3rd Step Conversion Rate  (whether the third message is where it converted)

## Email 
- Email Revenue 
- Total Recipients 
- Email Conversion Rate 
- Email Orders Placed 
- Email AOV 
- Total Unique Email Opens
- Total Unique Email clicks
- Total Email Unsubscribe Rate 
- Total Email Unsubscribes 
- Email Bounce Rate 
- Email Spam Complaint Rate 

As part of the analysis, you’re also going to check for common best practices in flows, such as A/B tests with conditional splits, multi-step flows to capture more customers, and whether there’s A/B testing happening on subject and preview lines.
Ignore draft flows. 

#RECOMMENDATION_ANALYSIS
When done, come up with 5 data-backed recommendations of how to update flows, and reasoning why.

#OUTPUT_FORMAT
First list all the currently live flows, ranked by most revenue generating to least.

Second, list all previously live flows, ranked by most revenue generating to least (and ignore draft flows).

Next, each live flow should have its own section. To lead off that section will be the flow name, along with a summary of what the trigger is, and what the flow does. For instance: Triggered based on last purchase, waits 30 days, sends an email, waits another 30 days, sends a discount.

Within that same section should be a table where the earlier mentioned stats are the columns and each month is a row. It should then provide a total summary (such as total revenue generated in the last 12 months). This will give us a good view into how each flow has been performing.

IMPORTANT: 
- Make sure the table has a row for EACH MONTH. No skipping out.
- Actually parse the CSV data and do a REAL, FULL analysis of the data. Not just a simulation or example. I want the real deal.

Lastly there will be a section on recommendations. It should identify the flow the recommendation applies to, along with a few actionable steps with reasoning as to why it’s making the suggestion

Important Notes

  • WATCH OUT for ChatGPT being lazy. Sometimes instead of doing the full analysis it does a "simulation". I’ve tried to tune the prompt to avoid it, but if you see ChatGPT being lazy, tell it to do the full analysis.

  • If you want to chat about the data, I would take the prompt results, copy & paste it into 4o, and then have the conversation. Most reasoning models are bad at back-and-forth conversation compared to their non-reasoning counterparts. Plus you'll have already put a lot into the context window.

  • I personally use Claude, and it does this whole process better, but ChatGPT will work great if that's what you have.

Step 3: Prep your dashboard data

Now that you’ve got the analysis, let’s make it visual.

Stay in the same chat, and drop in this quick follow-up prompt:

Give me the table summaries from each section above as a CSV I can download

I know, super advanced, right? This just gives you clean table data, ready for charts. Nothing fancy.

Step 4: Build the dashboard

Almost done. Let’s turn those tables into a beautiful dashboard.

Here’s how:

  1. Start a new chat using the GPT-4o model specifically (it has canvas support)

  2. Upload your newly generated CSV file

  3. Paste the prompt below

  4. Once the prompt is done writing the code, Hit “Preview” in the top right (this switches from code mode to visual)

What the prompt is actually doing

A couple of callouts for those who might want to modify this:

  1. "Using canvas" is an important callout. That tells 4o to actually open the canvas and use it for coding. This is what lets you preview the dashboard vs. trying to host it yourself

  2. If you want to slice the data differently, just adjust or add additional charts to the list

  3. It sounds crazy but the number notation is important to mention. Otherwise you get a bunch of data that's very difficult to understand. If you want to represent months in a different format, you'll have to tweak that, too

  4. If you don't have a CSV and you're copy and pasting your data, put it under the #DATA section to make sure it's all fully understood

Using canvas, code a dashboard I can view based on the table data attached. Make sure all charts and numbers in the dashboard appropriately use commas to notate dollar figures.

Break it up into the following charts: 
1. Bar chart of top performing live flows (last 12 months) 
2. Donut-hole Pie chart of revenue by flow, with total revenue overall in the middle (last 12 months) 
3. A line chart with an X-axis of each month, charting each of the flows where the Y axis is revenue.

#DATA (if not attached)
<paste your data here if needed, or remove this if uploaded

That's it! You’ve got a performance dashboard your boss will love. Or if you're an agency, an awesome dashboard your clients will actually understand about flows.

Need help?

DM me on LinkedIn or Twitter if anything breaks.

And if you want to skip the DIY and get super-fast, integrated retention automation, check us out at raleon.io.

We're all about letting AI work harder so you can focus on more strategic things.

Ready for your AI Retention Team? 👇

Related Posts

Unlock the power of precision targeting with AI-driven retention marketing. Learn how modern algorithms turn complex customer data into revenue-generating insights—boosting engagement, repeat purchases, and ROI through hyper-personalized campaigns.

AI isn’t replacing agency creatives—it’s making them more powerful. Discover how modern marketing teams are using AI to streamline workflows, supercharge performance, and focus more on strategy, storytelling, and client impact.

Move beyond outdated email segments. Discover how AI-driven predictive segmentation unlocks smarter targeting, real-time personalization, and 300%+ gains in revenue per recipient—without manual list management or guesswork.

Get Started Today!

Get Started Today!

Ryan

GreenEZ

I don't even know where to begin. From the exceptional customer support to the ease of use, Raleon has surpassed every current paid software available.

Copyright © 2024 Raleon. All Rights Reserved.

Copyright © 2024 Raleon. All Rights Reserved.

Copyright © 2024 Raleon. All Rights Reserved.