How Do I Use Claude to Analyze My Klaviyo Account? (Prompts Included)

Build-in-public post — the actual prompts we run on client Klaviyo accounts. Part of our AI workflows pillar.

The customer pain point

Every Klaviyo account has the same problem: the dashboards show you what happened (open rates, revenue per flow, campaign performance) but not what to do next. Founders stare at the flow performance screen and can't tell:

Which flow is underperforming relative to its potential (not just its absolute number)
Where the segmentation is too broad and diluting engagement
Whether the campaign cadence is too aggressive (driving unsubs) or too timid (leaving revenue)
Which single change would move the most revenue this month

That's an analysis problem, not a data problem. The data is all there. Claude is good at turning that data into a prioritized action list — if you prompt it right.

The Workflow
Prompt 1: Flow Performance Audit
Prompt 2: Segmentation Gap Analysis
Prompt 3: Campaign Cadence Review
How to Get the Data Out of Klaviyo
What to Watch For

Key Takeaways

Point	Details
The shift	Klaviyo dashboards report; Claude prescribes. Feed it the export, ask the right question, get a ranked action list.
3 core prompts	Flow performance audit, segmentation gap analysis, campaign cadence review.
Guardrail	Always feed Claude benchmarks + context. Without them it pattern-matches on generic "email best practices."
Output	A prioritized list of 3–5 changes ranked by revenue impact — not a wall of observations.

The Workflow

Export the relevant Klaviyo data (flow performance, segment sizes, campaign history — see the section below).
Feed it to Claude with a prompt that includes benchmarks and the business context (category, AOV, list size).
Demand a ranked output — the prompt forces Claude to prioritize by revenue impact, not just list problems.
Human review — an operator sanity-checks the recommendations against what's actually feasible before acting.

The single biggest mistake people make: pasting a Klaviyo screenshot and asking "how can I improve my email?" That gets you generic advice. The prompts below give Claude the structure to be specific.

Prompt 1: Flow Performance Audit

You are auditing a Shopify brand's Klaviyo flows. Here is the flow
performance data (last 90 days):

{paste flow performance: flow name, recipients, open rate, click rate,
placed-order rate, revenue per recipient, total revenue}

Business context:
- Category: {e.g. beauty / apparel / supplements}
- AOV: ${X}
- List size: {N}
- Email % of total revenue: {Y}%

Benchmarks for this category (use these, not generic ones):
- Welcome series: 8–12% placed-order rate, RPR $1.50–$4.00
- Abandoned cart: 10–15% recovery, RPR $4–$10
- Browse abandonment: 2–4% conversion, RPR $0.50–$1.50
- Post-purchase: drives 15–30% repeat-rate uplift, RPR $2–$5
- Healthy total: flows should be 50–60% of email revenue

For each flow:
1. State whether it is ABOVE / AT / BELOW benchmark.
2. If below: the single most likely cause (structure, copy, timing,
   segmentation, or it's simply not built out enough).
3. The specific fix.

Then give me a RANKED list of the top 3 changes by estimated revenue
impact, with a rough $/month estimate for each based on the AOV and
list size above. Be concrete. No generic advice.

The benchmark block is what makes this work. Without it Claude compares your flows to a vague mental average. With it, it tells you "your abandoned cart RPR is $2.10 vs the $4–$10 benchmark — that's the #1 fix."

Prompt 2: Segmentation Gap Analysis

Here are the active Klaviyo segments and their sizes:

{paste: segment name, size, definition/criteria}

Total list size: {N}. Engaged (opened/clicked last 60d): {M}.

Identify:
1. Segments that overlap heavily (sending the same people twice).
2. Missing segments that this brand should have but doesn't
   (engaged vs lapsed vs cold, VIP by lifetime spend, browse-only
   non-buyers, post-purchase by product category).
3. Whether the brand is sending to cold/unengaged subscribers in a
   way that risks deliverability.

Output: the 3 segmentation changes that would most improve engagement
and deliverability, ranked, with the reasoning for each.

Prompt 3: Campaign Cadence Review

Here is the campaign send history (last 60 days):

{paste: send date, campaign name, segment sent to, recipients,
open rate, click rate, revenue, unsubscribe count}

Analyze:
1. Is the overall cadence too aggressive (rising unsub rate, declining
   opens over the period) or too timid (low send frequency, leaving
   revenue on the table)?
2. Which campaign TYPES (promo / content / product / story) perform
   best for this brand by revenue per send?
3. Is there list fatigue — are opens declining over the 60 days
   independent of subject line?

Output: a recommended weekly cadence (sends per week, by segment
engagement tier) and the campaign-type mix that maximizes revenue
without accelerating unsubscribes.

How to Get the Data Out of Klaviyo

Three ways, in order of effort:

Manual export (fastest to start). Klaviyo → Analytics → Flows / Campaigns → export the performance table to CSV. Paste relevant columns into the prompt. Good for a one-time audit.
Klaviyo API (repeatable). Klaviyo's reporting API returns flow + campaign metrics as JSON. A small script pulls it on a schedule and feeds the prompt. Good for monthly reviews.
MCP server (most integrated). An MCP server wrapping the Klaviyo API lets Claude pull the data itself during the analysis conversation. Most seamless, most setup.

Start with #1 to prove the prompts work for your account, then graduate to #2 for a repeatable monthly cadence.

What to Watch For

Feed benchmarks or get generic output. This is the #1 failure. Claude without category benchmarks gives you "improve your subject lines" — useless. With benchmarks it gives you "your welcome flow RPR is 60% below benchmark, here's why."
Context changes the answer. A $30 AOV beauty brand and a $400 AOV furniture brand have completely different "good" — always include AOV, category, list size.
It estimates, you verify. The $/month impact numbers are directional, not forecasts. Use them to rank fixes, not to set revenue targets.
Human gate before acting. Claude might recommend a change that's technically right but operationally hard (e.g., a segmentation overhaul mid-BFCM). An operator filters for feasibility. We cover this discipline in what AI-run means when the AI screws up.

Talk to Branva

We run this Klaviyo analysis monthly for every client as part of transparent monthly marketing — plus we build and run the 4 core flows themselves. Book a free call.

Frequently Asked Questions

Can Claude connect to Klaviyo directly?

Not natively. You either export data manually, pull it via the Klaviyo API with a script, or wrap the API in an MCP server Claude can call. The prompts work with any of those — the data just has to get into the conversation.

Will Claude's revenue estimates be accurate?

They're directional, meant for ranking fixes, not forecasting. "Fix A is worth ~3x more than Fix B" is the useful signal. The absolute dollar figures are rough.

Is this better than Klaviyo's own benchmarks feature?

Klaviyo's benchmarks tell you where you stand. This tells you what to do about it, ranked by impact, with the specific fix. Different jobs — use both.

What model should I use?

A frontier reasoning model for the analysis (the prioritization is the hard part). The data extraction is cheap; the reasoning is where quality matters. Don't cheap out on the analysis call — it's a few cents either way and the recommendation quality difference is large.