Klaviyo is a customer data and marketing platform for consumer brands that combines email, SMS and WhatsApp messaging with predictive scoring, Flows automation and AI agents that build campaigns and answer support tickets. It targets ecommerce and retail teams that want segmentation, lifecycle automation and service in one customer profile.
Used this tool? Rate it
Used this tool? Rate it
Klaviyo began life as an email marketing tool for online stores. That is no longer an accurate description of what it is, and using the old framing will lead you to evaluate it against the wrong competitors. Today the homepage now leads with the autonomous B2C CRM and describes the product as the AI marketing and service platform for ambitious consumer brands.
The distinction matters practically. An email tool asks "what should we send this week?" A customer relationship platform asks "who is this person, what have they bought, what are they likely to do next, and which channel should reach them?" Klaviyo's architecture is built around the second question: a unified customer profile fed by store events, with messaging channels, predictive scoring, automation and now customer service all reading from and writing back to that same profile.
The operating entity is Klaviyo, Inc., with terms governed by the Commonwealth of Massachusetts and venue in Boston. It is a public company — third-party reporting confirms it completed its IPO in September 2023 at a 9.2 billion dollar valuation — which means its financial disclosures and product roadmap are subject to a level of external scrutiny that private vendors avoid.
On its own numbers, Klaviyo reports 205,000+ relationship-driven brands across 100 countries, 7.3 billion customer profiles and 2.5 billion events processed daily, all vendor-reported and unaudited. Two of those figures deserve different weight. The brand count is a commercial metric; the event volume is an architectural one, and it hints at the real product: an event ingestion and profiling pipeline that happens to have messaging attached, rather than a mail sender that happens to store contacts.
Klaviyo is explicitly B2C. Its data model, predictive scores and automation patterns assume repeat consumer purchasing — order histories, browsing sessions, product catalogues, replenishment cycles. If you sell to businesses through a long sales cycle with named accounts and opportunity stages, this is the wrong shape of tool, and the ordinary B2B CRM category is where you should be looking.
The foundational object is the segment. Rather than exporting a static list, you define conditions — placed an order in the last 90 days, viewed a product but did not purchase, predicted to churn — and membership updates continuously as events arrive. This sounds like a small distinction and is not: it means an automation targeting "customers about to lapse" always addresses the people who currently qualify rather than those who qualified when the list was built.
Flows are the automation engine. A flow starts from a trigger — an abandoned cart, a first purchase, entry into a segment, a date-based event like a subscription renewal — and then branches through conditional logic, time delays and channel choices. The canonical revenue cases are well known in ecommerce: cart and browse abandonment, welcome series, post-purchase sequences, winback campaigns and replenishment reminders.
What makes flows more than a drip tool is that their conditions can read anything on the profile, including the predictive scores described below. A branch can send a discount only to customers whose predicted lifetime value is high enough to justify the margin sacrifice.
This is where Klaviyo's own mechanism is worth stating precisely rather than describing generically. The published scoring dimensions are predicted customer lifetime value, RFM analysis and churn risk. Beyond those, the platform also projects an expected next order date and average order value, and those predictions become filter conditions inside segments.
That last point is the important one. Prediction is only useful if it changes an action, and here the predicted values are first-class profile attributes, so "message customers whose expected next order date has passed" is a segment definition rather than an analysis exercise. Note the practical prerequisite: predictive models need purchase history to train on, so a new store with little order data will not get meaningful scores immediately.
Composer is the marketing-side AI agent. Per the official description, Composer analyzes performance to find opportunities, then builds the campaign for you from a prompt and supplies the context behind its decisions.
The design intent is worth noting: it does not simply generate copy on request. It starts from performance analysis to identify where revenue is being left on the table, then assembles the campaign — audience, content, channel — and explains its reasoning. The explanation matters because an agent that makes targeting decisions without exposing them is difficult to trust with a customer list.
Customer Agent handles the post-purchase half. It is described as operating around the clock across support and sales, pre-trained on your customer data, working across the existing tech stack through integrations, and enriching profiles through the conversations it has.
That final capability is the structurally interesting one. A support conversation is a rich source of customer information — preferences, complaints, intent — and routing it back into the same profile that marketing reads from is what makes the marketing-plus-service combination more than two products sold together.
The messaging surface spans email, SMS, RCS, WhatsApp, mobile push and social, with review management and a helpdesk alongside. These are not equivalent: email remains the volume channel, SMS is metered separately and carries regulatory obligations, and WhatsApp has its own template approval and session rules. But they share the same profile and segment layer, which is what allows channel choice to become a branch inside a flow rather than a separate campaign in a separate tool.
On what powers the AI, the company states that Klaviyo builds in-house machine learning capabilities and also collaborates with service providers like OpenAI and AWS Bedrock, stating those providers are contractually prohibited from using customer prompts for training. This mixed approach is worth knowing: the predictive scoring is proprietary modelling on your event data, while the generative features route through external foundation models under contractual restrictions.
The highest-return automation in ecommerce remains cart and browse abandonment, because it targets people who have already demonstrated purchase intent. Klaviyo's version benefits from event granularity: it can differentiate between someone who viewed a category, viewed a product, and reached checkout, and vary the message accordingly.
For brands with repeat purchase patterns — consumables, beauty, supplements, pet supplies — the predicted next order date turns retention from a calendar guess into a per-customer estimate. A reminder timed to an individual's actual replenishment cycle outperforms a blanket monthly send, and this is the clearest example of the predictive layer paying for itself.
Rather than emailing the whole list about a new product, the segment engine allows targeting by prior category purchase, engagement recency and predicted value. This matters for deliverability as well as conversion: consistently mailing unengaged recipients damages sender reputation, and suppression by engagement segment is standard practice.
The combination of Customer Agent, Customer Hub and Helpdesk addresses a common structural problem: marketing sends a promotion to a customer whose complaint is unresolved, because the two systems do not share state. Putting both on one profile is the intended remedy.
For teams running separate tools for email, SMS, reviews, help desk and customer data, the consolidation argument is about eliminating sync latency and reconciliation work as much as about licence cost. The counterargument is concentration risk, which the Limitations section addresses.
Start by integrating your ecommerce platform so historical orders and live events flow in. This step is not merely setup: predictive scores and meaningful segments both depend on this history, so the platform is genuinely less capable on day one than it will be after data accumulates.
Configure domain authentication before sending volume. This is unglamorous infrastructure work — publishing the DNS records that let mailbox providers verify your mail — and skipping it is a leading cause of inbox placement problems that later get misattributed to the platform.
Automations earn continuously, campaigns earn once. Set up welcome, abandonment and post-purchase flows first; they generate revenue every day without further work, while a campaign requires a new decision each time.
Create segments for engaged and unengaged subscribers at the outset and route sends accordingly. Deliverability is a function of consistent engagement, and establishing suppression discipline before problems appear is far easier than repairing a damaged sender reputation afterwards.
Predictive attributes need sufficient order history to become reliable. Treat early scores as directional, and revisit any flow branch that depends on them once the store has accumulated more transactions.
If you enable Composer or Customer Agent, start on a limited scope with human review before letting them act broadly. Composer explains its reasoning specifically so it can be checked; use that. For Customer Agent, review transcripts on a defined set of query types before widening its remit.
Because messaging is metered and agent features consume credits, your bill responds to activity as well as list size. Build the habit of checking consumption alongside campaign results.
Optimise for revenue per recipient, not open rate. Open tracking has been unreliable since mail privacy protection began inflating it. Attributed revenue and conversion are the metrics that survive scrutiny, and the platform reports on them directly.
Prune aggressively. Removing chronically unengaged subscribers usually improves both deliverability and cost, since billing scales with active profiles. A smaller engaged list frequently outperforms a larger indifferent one on absolute revenue.
Treat SMS as a scarcer resource than email. It is more expensive per message, more intrusive, and more tightly regulated. Reserve it for time-sensitive or high-value moments and keep consent records rigorous.
Let flows do the heavy lifting. Teams that plateau usually over-invest in weekly campaign production and under-invest in automation. Auditing and improving existing flows typically yields more than another broadcast.
Test one variable at a time and wait for significance. With reasonable list sizes, A/B testing is meaningful — but only if you change one thing and let the test reach a real sample rather than declaring a winner after an hour.
Check the agents' reasoning rather than only their output. Composer explains why it targeted a segment; that explanation is the audit surface. Reviewing it is how you catch a plausible-looking campaign built on a faulty assumption.
Direct-to-consumer ecommerce brands are the core audience and the best fit, particularly those on Shopify, where integration depth and the volume of shared documentation are greatest.
Retention-driven and subscription businesses get disproportionate value from the predictive layer, because replenishment timing and churn risk are directly actionable for repeat-purchase categories.
Multi-channel consumer brands benefit when email alone is insufficient and messages need to reach customers across SMS, WhatsApp and push from a single profile and segment definition.
Teams consolidating marketing and support are the audience for the newer service modules, where sharing state between the two functions is the point.
Larger consumer brands with compliance requirements are served by the certification posture described below, and by the enterprise tier and professional services.
Who it fits less well: B2B companies with account-based sales motions, which need pipeline and opportunity management rather than consumer lifecycle automation; very small or new stores without enough purchase history for the predictive features to work; businesses with tiny lists where the free tier's constraints bite quickly; and any team that wants a simple newsletter tool, for whom the platform's depth is cost and complexity they will not use.
Klaviyo is a cloud application accessed through the browser, with mobile apps for monitoring and approvals rather than full authoring. There is nothing to install and no self-hosted option.
The integration surface is where platform availability actually matters for this category. The ecommerce connection is the load-bearing one, and the third-party listing confirms it syncs Shopify data in real time and integrates with 350+ applications, with more than 117,000 brands using Klaviyo and Shopify together. Beyond Shopify, connectors cover other major commerce platforms, point-of-sale systems, review tools, loyalty programmes, advertising platforms and help desks.
For anything unsupported, the developer platform exposes REST APIs and webhooks, which is how custom event tracking and bespoke integrations are built. Given that the entire product is driven by events landing on profiles, API access is a more consequential capability here than it would be in a simple mail tool.
Channel availability carries geographic caveats: SMS and RCS support varies by country and is subject to local carrier registration and regulatory requirements, and WhatsApp operates under its own messaging rules. Confirm channel availability for your specific markets before assuming a global rollout.
Klaviyo's pricing has two moving parts: which modules you buy, and how large and active your audience is. Understanding both prevents unpleasant surprises.
There is a genuine free tier, and its ceilings are published precisely: the free tier is capped at 250 active profiles, 500 emails per month, five dollars of mobile messages per month and 10,000 Composer credits, with email support limited to the first 60 days. Crossing either ceiling matters immediately because your ability to send may be restricted once you surpass 250 contacts or 500 monthly emails.
Paid pricing is modular rather than a single subscription: the pricing page presents Marketing, Data and Analytics, Service, Composer, Customer Agent, Professional Services and Enterprise as separately priced components. Within each, cost scales with active profiles and message volume, which is why the official page uses an interactive calculator rather than publishing a static table.
For concrete entry-level figures, the third-party marketplace listing is more specific than the official page: an SMS plan at 15 dollars per month for up to 1,250 SMS or MMS credits, and an email plan at 20 dollars per month for 251 to 500 contacts. These are the first paid steps, and cost rises substantially as profile counts grow into the tens of thousands.
The contractual terms deserve attention before committing: service orders automatically renew, fees are charged in advance and are non-refundable, and cancellation requires five days notice on monthly billing or thirty days on any other cycle. Diarise the notice period at the point of signing rather than discovering it at renewal.
A budgeting note specific to this platform: because billing follows active profiles, list hygiene is a direct cost lever. Removing subscribers who never engage reduces spend and typically improves deliverability at the same time.
Broad marketing automation suites aimed at general small business use are usually cheaper and simpler, and are the better choice if you send newsletters and do not need deep commerce event data or predictive scoring.
Enterprise marketing clouds offer greater breadth across B2B and B2C, more elaborate governance and deeper integration with large CRM estates, at substantially higher cost and implementation effort. Large enterprises with existing commitments to those ecosystems often stay there.
Commerce-platform-native marketing tools are cheaper and require no integration work, but offer materially less sophisticated segmentation and automation. They are a reasonable starting point that many brands outgrow.
Specialist SMS platforms can be more capable or cheaper on messaging alone, at the cost of splitting the customer profile across systems — which is precisely what the consolidated approach is trying to avoid.
Customer data platforms plus separate messaging tools give maximum flexibility and control for organisations with data engineering capacity, at the cost of building and maintaining what Klaviyo delivers as a package.
The fair summary: Klaviyo's distinguishing combination is deep commerce event data, predictive scores usable directly as segment conditions, multi-channel messaging and, more recently, service — in one profile. Competitors typically win on price, on enterprise breadth, or on a single channel, but rarely on that combination for B2C.
Costs scale with success. Pricing follows active profiles and message volume, so a growing list means a growing bill. Brands with large but weakly engaged audiences can find the economics unfavourable, and the mitigation — disciplined list hygiene — needs to be a routine, not a rescue measure.
The free tier is a trial, not a plan. With 250 profiles and 500 monthly emails, it is sized for evaluation. Any real store will exhaust it quickly, and support access lapses after 60 days.
Predictive features require data. Scores depend on purchase history. New stores, low-frequency purchase categories and businesses with sparse order data will get limited value from the predictive layer initially, despite paying for a platform whose differentiation partly rests on it.
Renewal terms are strict. Automatic renewal, advance non-refundable billing and short cancellation notice windows mean an unnoticed renewal date is expensive. This is standard SaaS contracting, but the five-day monthly notice period is tighter than many expect.
Deliverability remains your responsibility. No platform can rescue poor list practices. Buying lists, mailing unengaged recipients and neglecting authentication will damage inbox placement regardless of vendor.
Channel regulation sits with the sender. SMS and WhatsApp carry consent, opt-out and registration obligations that vary by jurisdiction. The platform provides consent tooling, but legal compliance is the brand's duty, not the vendor's.
Agent output needs supervision. AI-generated campaigns and automated support replies can be confidently wrong. Composer's explanation of its reasoning exists to be checked; treat both agents as capable drafters requiring review rather than autonomous operators, particularly on pricing, promises and policy statements.
Consolidation concentrates risk. Running marketing, messaging, customer data and support on one vendor simplifies operations and increases exposure — to outages, pricing changes and migration difficulty. Retaining an export path for your customer data is prudent.
Complexity has a learning cost. The depth that makes the platform powerful also makes it harder to learn than a basic newsletter tool. Small teams should budget onboarding time honestly.
For a platform whose entire function is accumulating customer profiles, data governance is a first-order concern rather than a footnote — and the contractual position here is clearer than most.
On ownership, the terms are unambiguous: customer data is the exclusive property of the customer as between the two parties, and the customer also owns all outputs generated for its account. That second clause matters as AI features expand, since ownership of generated content is not universally granted by vendors.
On model training, the terms state that customer data will not be used to train third-party foundation models, while reserving the right to use anonymized derived data for product development. Read both halves. The commitment about third-party foundation models is meaningful and specific; the derived-data reservation is broader and is the clause to raise with your legal team if aggregate analysis of your data is a concern. The AI documentation adds that external providers are contractually prohibited from using customer prompts for training.
On external validation, the trust centre lists SOC 2, ISO 27001, ISO 27017 and PCI DSS, with annual third-party audits of internal controls. These are independent attestations rather than self-declarations, which places the platform ahead of many vendors in this category on verifiable security posture.
For your own compliance obligations, the platform provides profile consent management and rights request tooling to satisfy both access and deletion requests. This is the correct division of responsibility to understand: Klaviyo acts as processor for the customer data you load, you remain the controller, and the tooling helps you discharge duties that remain legally yours. A data processing agreement is published alongside the other legal documents.
One gap worth noting: the public trust page does not specify data residency regions. If your obligations require data to remain in a particular jurisdiction, confirm that directly with the vendor rather than assuming it.
There is a free tier limited to 250 active profiles and 500 emails per month, plus a small monthly allowance of mobile messages and Composer credits, with email support for the first 60 days only. It is suitable for evaluation rather than for running a real store, and sending may be restricted once either limit is exceeded.
Cost depends on which modules you take and how many active profiles you have. Entry-level paid pricing starts around 20 dollars per month for 251 to 500 contacts on email, and 15 dollars per month for an SMS allowance of up to 1,250 credits. Marketing, analytics, service, Composer and Customer Agent are priced as separate components, and cost rises with audience size — check the official calculator for your specific numbers.
It is built for B2C relationships and is strongest with ecommerce data, since its predictive models and automation patterns assume repeat consumer purchasing. Retail, restaurant and wellness businesses are also served. It is a poor fit for B2B sales processes based on named accounts and opportunity pipelines.
Predicted customer lifetime value, RFM classification, churn risk, expected next order date and expected average order value. Crucially, these are profile attributes you can use directly as segment conditions, so they drive targeting rather than only reporting. They require sufficient purchase history to become reliable.
Composer is the marketing agent: it analyses performance to find opportunities and builds complete cross-channel campaigns from a prompt, explaining its reasoning. Customer Agent is the service agent: it handles support and sales enquiries continuously, is pre-trained on your customer data, and writes what it learns back into customer profiles.
Email, SMS, RCS, WhatsApp, mobile push and social messaging, alongside review management and a help desk. Availability of SMS, RCS and WhatsApp varies by country and is subject to carrier registration and local regulation, so verify coverage for your markets.
Yes, and it is the deepest integration in the ecosystem — real-time data sync plus more than 350 other application integrations. The marketplace listing reports over 117,000 brands using the two together, with a 4.7 star rating across nearly 3,000 reviews.
The terms state that customer data will not be used to train third-party foundation models, and the AI documentation says external providers are contractually prohibited from using customer prompts for training. Note the separate reservation allowing anonymised derived data to be used for product development, which is worth reviewing if aggregate use of your data concerns you.
The trust centre lists SOC 2, ISO 27001, ISO 27017 and PCI DSS, with annual third-party audits. Consent management and rights-request tooling are provided to help you meet GDPR and CCPA duties, which remain yours as the data controller. Data residency regions are not published publicly, so confirm those directly if they matter to you.
Subscriptions renew automatically and fees are charged in advance and are generally non-refundable. Cancellation requires five days notice on a monthly billing cycle or thirty days on any other cycle, so note the deadline when you sign rather than at renewal. Export your customer data before terminating, since the profile database is the asset you most need to retain.
