How to export Klaviyo campaign data and find your winning patterns with AI

Why export Klaviyo data instead of using the dashboard?

Because comparing a lot of sends inside any dashboard is slow. When I tried Klaviyo's Composer, nothing was wrong with it exactly, but comparing revenue across time periods took a long time. I could have downloaded the CSV, or had Claude pull and analyze it, much faster.

I love using Claude, so I don't see a reason to use Composer myself. If you live in Klaviyo and don't use any AI tools, Composer can make sense. My take on it is in the Klaviyo Composer review, and the bigger picture of Klaviyo's own AI features is in Klaviyo AI.

The point of the export isn't the file. It's being able to ask questions like "what do my best 20 campaigns have in common?" and get an answer fast.

How do you export campaign data from Klaviyo?

Klaviyo's help center lays out the steps:

1. Go to the Campaigns tab.
2. Click the three-dot Options menu next to "View library."
3. Choose Export analytics. If you also see "Export analytics (legacy)," use the one that isn't legacy.
4. Pick a time range. Use a preset or set custom dates.
5. Optional: filter by tags.
6. Pick channels: email, SMS, push or WhatsApp (only the ones turned on in your account show up).
7. Decide whether to include all A/B test variations.
8. Click Export. The file downloads when it's ready.

If you need it again later, Klaviyo keeps exports from the last 30 days under your account name > Settings > Other > Downloads.

What's in the Klaviyo campaign export?

One row per campaign message, with these columns: campaign ID, message ID, campaign name, message name, tags, audiences, send date and time, send day of week, subject line, channel, total recipients, unique conversions, conversion rate, conversion value, revenue per recipient, average order value, unique opens, open rate, total opens, unique clicks, click rate, total clicks, unsubscribes, unsubscribe rate, spam complaints, spam complaint rate, deliveries, bounces and bounce rate.

A few things to know before you analyze it:

• SMS opens are always 0. Klaviyo doesn't track SMS opens, so don't let the AI compare email and SMS on opens.
• A/B tests add rows. If you include variations, each one gets its own row, plus the winner. A test with two variations shows up as three rows.
• It's campaign-level data. The file is about your sends, not individual customer profiles.

How do you export flow data from Klaviyo?

Flows work the same way. Klaviyo's help center says Owners, Admins and Analysts can do it:

1. Go to the Flows tab.
2. Click the Options dropdown at the top.
3. Choose Export analytics.
4. Pick a time range (all-time, a preset or custom).
5. Optional: tick "Aggregate analytics by day, week, or month" if you want to see trends.
6. Click Export.

To export a single flow, Klaviyo's tip is to tag that flow and export by tag.

The flow export is the one I'd pull to see where volume is. It shows you which emails actually get sent the most, which tells you where a test or a rewrite will matter. A test on an email that barely gets sent won't teach you much.

How do you use AI to find your winning subject line patterns?

Upload the campaign file to Claude or ChatGPT and ask it to find your winning subject line patterns. I suggest doing this before you plan a big month like November, so your calendar is built on what has worked for your list.

Prompts that work well:

• "Rank these campaigns by revenue per recipient. What do the top 20 subject lines have in common that the bottom 20 don't?"
• "Group the subject lines by type: news, offer, question, brand voice, emoji. Compare average click rate and revenue per recipient for each group."
• "Does subject line length relate to click rate? Split them into under 20 characters, 21 to 40, and 41 to 60."
• "Which send day and hour has the highest revenue per recipient, and how many sends is that based on?"

Two rules for the prompts. First, ask it to judge on clicks and revenue per recipient, not opens. Across 4 consumable brands and 2,897 campaigns, click rate tracked revenue about twice as closely as open rate (0.65 vs 0.32). Second, ask how many sends each finding is based on. A pattern from three campaigns isn't a pattern.

What patterns should you look for?

Here's what came out of a subject line study of 2,821 campaigns across 7 brands:

• News beat brand voice. Subject lines that said what was new or what you get did better than clever lines in the brand's voice.
• Vague questions did worst. Across brands, question subject lines earned about 18% less revenue per email. My guess, and it's only a guess, is that they reach people who are interested but confused, not ready to buy.
• Short lines can win clicks. For one sports tech brand, subject lines of 20 characters or fewer got more than 2x the clicks of lines between 41 and 60 characters.
• Emoji are brand-specific. Across 4 brands in one category, emoji helped two and hurt two.

That last one is the whole reason to analyze your own file. A general rule can be wrong for your list. More on writing them in email subject line best practices.

How do you turn what the AI finds into a test?

Pick one pattern and test it on purpose:

1. Write two subject lines that differ only in that pattern (news vs question, short vs long).
2. Split the send 50/50.
3. Let at least 100 emails go out before you decide, and judge a subject line test on opens as a rough signal, then confirm with clicks and revenue.
4. Repeat it across about five campaign sends at your normal send size before you treat it as a rule.

Then export again next quarter and see if the pattern held. If you want a regular rhythm for this, the weekly Klaviyo check is a good place to slot it in.

FAQ

WRITTEN BY Jack White

Founder of Bardown Digital, a Chicago-based Klaviyo email and retention agency for Shopify brands. He's worked with brands doing $500K to $100M a year, with $100M+ in email-attributed revenue across them.

Instagram @jackintheinbox LinkedIn