Klaviyo AI Features: What Works and What Doesn't

Klaviyo has been adding machine learning features at a rapid pace. Predictive analytics. Subject line suggestions. Send time optimization. Audience recommendations. Generative content tools.

Some of these features genuinely improve performance. They save time, surface insights you would miss manually, and optimize in ways that human intuition cannot match.

Others are still half-baked. They sound impressive in a product demo but underdeliver in practice — giving generic suggestions, requiring too much data to work well, or producing results you could achieve faster manually.

I have tested every feature Klaviyo offers across multiple eCommerce brands. Here is the honest assessment of what works, what does not, and what is promising but not quite there yet.

What Works: Predictive Analytics

Klaviyo's predictive analytics are genuinely useful. They calculate forward-looking metrics on each customer profile:

• Predicted next order date: When this customer is likely to buy again
• Predicted lifetime value: Expected total revenue from this customer
• Average time between orders: Typical purchase cadence
• Predicted gender: Inferred from purchase behavior
• Churn risk: How likely they are to never buy again

Why it works:

These predictions are based on actual purchase history across your entire customer base. The model looks at patterns — customers who buy product X tend to reorder every 45 days; customers who only buy during sales have lower LTV — and applies those patterns to individual profiles.

How to use it effectively:

• Segment by predicted LTV. Your top 20% by predicted LTV should get VIP treatment: early access, better offers, personal outreach.
• Segment by churn risk. High churn risk customers need a proactive retention campaign before they disappear.
• Trigger flows on predicted next order date. Send a replenishment reminder 5-7 days before their predicted next purchase.
• Allocate marketing spend by LTV. Spend more to retain and upsell high-LTV customers. Spend less acquiring low-LTV profiles.

Caveats: Predictions require at least 500 customers with repeat purchases to generate useful data. For newer brands or brands with mostly one-time buyers, the predictions are less accurate. The model improves as your data grows.

Verdict: Use it. This is one of Klaviyo's strongest differentiators and genuinely drives revenue when you build segments and flows around the predictions.

What Works: Smart Send Time

Smart Send Time optimizes delivery timing at the individual subscriber level. Instead of blasting your entire list at 10 AM, it delivers each email at the time that specific subscriber is most likely to engage.

Why it works:

Different people check email at different times. A subscriber who always opens emails at 7 PM will miss your 10 AM send (it is buried under 50 other emails by evening). Smart Send Time delivers to that person at 6:45 PM when they are about to check.

Real results we have seen:

• Open rate improvements of 3-7 percentage points
• Click rate improvements of 1-3 percentage points
• These improvements compound over time as the model learns

How to use it:

Enable Smart Send Time on campaigns where timing matters and the campaign is not time-sensitive. A flash sale that ends at midnight needs everyone to receive it at the same time. A content email or product launch can benefit from personalized timing.

Caveats: It needs data. For new subscribers with no open history, it defaults to your selected send time. The more engagement data Klaviyo has on a subscriber, the better the timing optimization works. After 4-6 weeks of engagement data, predictions become reliable.

Verdict: Use it for non-urgent campaigns. Keep manual timing for time-sensitive promotions.

What Partially Works: Subject Line Suggestions

Klaviyo offers subject line generation and scoring. You can ask it to generate subject line options or score your written subject line for predicted performance.

What works about it:

• The scoring feature is useful as a gut-check. If your subject line scores poorly, it is worth reconsidering.
• Generated suggestions sometimes spark ideas you would not have considered.
• It learns from your brand's historical performance data.

What does not work:

• Generated subject lines are often generic. They lack the brand voice, personality, and specific knowledge that makes great subject lines great.
• The scoring is based on general patterns, not your specific audience. A subject line that scores low might perform well with your particular niche.
• It cannot replicate the curiosity, humor, or cultural references that your best-performing subject lines use.

How to use it:

Use subject line suggestions as a brainstorming starting point, not a final answer. Write your own subject lines first. Then check the score as one data point (not the deciding factor). If the score is extremely low, reconsider. If it is moderate, trust your instincts and test.

Verdict: Use as a tool, not a crutch. Your best subject lines will come from understanding your audience deeply, not from a model trained on aggregate data.

What Partially Works: Audience Recommendations

Klaviyo suggests audience segments for campaigns based on product type, purchase history, and engagement data.

What works:

• Useful for newer marketers who do not know where to start with segmentation
• Surfaces segments you might not think of ("customers who bought X but not Y")
• Helps identify cross-sell opportunities

What does not work:

• Recommendations are often obvious to experienced email marketers
• Cannot account for your specific brand knowledge (seasonal patterns, upcoming launches, customer feedback)
• Sometimes suggests sending to segments that are too small to generate meaningful revenue

How to use it:

Check the recommendations weekly as inspiration. If one suggests a segment you had not considered, test it. But build your primary segmentation strategy yourself based on your knowledge of your business.

Verdict: Useful for beginners, limited value for experienced teams.

What Does Not Work Well: Generative Email Content

Klaviyo's content generation (writing full email body copy) is the weakest feature in our testing.

The problems:

• Generated copy is bland and generic. It reads like every other brand's email because it is optimized for broad performance, not your specific voice.
• It cannot reference specific product benefits, customer stories, or brand history without significant manual editing.
• The tone is consistently "corporate friendly" regardless of your brand's personality.
• You spend as much time editing the generated content as you would writing from scratch.

Where it might help:

• Generating first drafts of transactional emails (order confirmation, shipping updates)
• Writing product description snippets for dynamic product blocks
• Creating basic email outlines that you heavily rewrite

Verdict: Skip it for campaigns and brand emails. Write your own copy or hire a copywriter. Generated content does not convert at the same rate as thoughtful, brand-specific writing.

What Does Not Work Well: Campaign Recommendations

Klaviyo suggests when to send campaigns, what topics to cover, and which products to feature.

The problems:

• Recommendations are based on general eCommerce patterns, not your specific brand calendar
• Cannot account for your inventory levels, upcoming launches, or marketing strategy
• Timing suggestions often conflict with your planned calendar
• Product recommendations are purely based on data (bestsellers, trending) without considering margin, inventory, or strategic priority

Verdict: Ignore these unless you have zero marketing strategy. If you have a content calendar and a plan, your judgment about what to send and when will outperform algorithmic suggestions.

What Is Promising But Early: Flows Optimization

Klaviyo is adding features that suggest flow improvements: additional branches, better timing between emails, optimal number of emails in a sequence.

Where this could be valuable:

• Suggesting that your welcome series should be 6 emails instead of 4 (based on engagement drop-off data)
• Identifying the optimal delay between flow emails
• Suggesting which conditional splits would improve flow performance

Current limitations:

• Still requires manual testing to validate suggestions
• Cannot account for your specific content strategy or brand voice
• Suggestions are conservative (optimize for engagement, not necessarily revenue)

Verdict: Watch this space. The technology is improving. For now, treat suggestions as hypotheses to test, not directives to follow.

The Honest Assessment

Klaviyo's machine learning features fall into three buckets:

Genuinely useful (use these):

• Predictive analytics (CLV, next order date, churn risk)
• Smart Send Time
• Segment-level performance insights

Partially useful (use with judgment):

• Subject line scoring (as a gut check, not gospel)
• Audience recommendations (for inspiration)
• A/B test optimization (auto-selecting winners)

Not ready yet (skip these):

• Full email content generation
• Campaign topic recommendations
• Generalized flow suggestions

The Right Way to Think About These Tools

Machine learning is excellent at finding patterns in large datasets. It is terrible at understanding context, brand voice, and creative strategy.

Use the tools for what they are good at:

• Timing optimization (when to send)
• Prediction (who will buy, when, how much)
• Pattern recognition (which segments perform differently)

Do not rely on them for:

• Creative decisions (what to say, how to say it)
• Strategic decisions (what campaigns to run, what flows to build)
• Brand voice (no model knows your brand like you do)

The brands making the most money from Klaviyo use the data and prediction features extensively while keeping creative and strategic control in human hands. That combination — machine precision on timing and targeting, human creativity on content and strategy — is where the magic happens.

The Bottom Line

Klaviyo's best features are the ones that give you better data to make better decisions. Predictive analytics and Smart Send Time genuinely improve performance with minimal effort.

Their weakest features are the ones that try to replace human creativity and judgment. Content generation and campaign recommendations are too generic to drive real results for brands with a distinct voice and strategy.

Use what works. Ignore what does not. And test everything before trusting it.

---

Want us to optimize your Klaviyo setup and configure the features that actually drive revenue? Book a free strategy call and we will audit your account and show you what to enable, disable, and improve.