Your CAC did not double because of the auction
Almost every article on this subject opens the same way. CPMs are exploding, paid media is broken, build your list. The first two claims are rarely true as stated.
Here is what the auction actually did. Meta reported average price per ad up 9% year over year for full year 2025, and up 12% year over year in the second quarter of 2026, both in its own results releases. That is real inflation. It is not a doubling.
So if your blended acquisition cost doubled, the auction took a slice and something else did the rest. Look at creative decay first, where the winners stopped winning and nothing replaced them, then the offer, then whether a measurement change moved the number without moving the business.
The case for owning your audience is therefore not that paid is dead. It is duller and better. Paid media is a variable cost with no accumulation: stop paying and it stops. The list is the only line in the P&L where last year’s work still shows up this year.
You do not own a list. You own permission. Permission is the only marketing asset that decays quietly enough that you do not notice until the quarter you need it. That is our position, not a quotation.
What owned actually means
An owned audience is a set of contact records you hold, with consent you can prove, reachable without paying a third party for distribution. You own the permission and the data. You do not own the attention, and you do not own the inbox. Google, Apple and Microsoft own the inbox, and they keep tightening the terms.
Klaviyo’s BFCM 2025 report, filed with the SEC on 2 December 2025, put more than $3.8 billion in Klaviyo Attributed Value across the five days, up 27% year over year, and revenue from repeat customers growing 13.5% year over year, outpacing new buyers. The second number is the stronger one, because the first is a last touch attributed figure.
Flows before campaigns, and the arithmetic behind it
The reason to build automation before sending more newsletters is not craft. It is revenue per recipient.
| Metric | Flows | Campaigns |
|---|---|---|
| Share of email sends | about 5.3% | the balance |
| Share of email revenue | about 41% | the balance |
| Revenue per recipient | $1.94 | $0.11 |
| Placed order rate | 2.11% | 0.16% |
| Click rate | 5.58% | 1.69% |
Those are the figures in Klaviyo’s 2026 Omnichannel Benchmark Report as consulted in September 2026, so check them against the edition you read before they reach a board deck. Watch the denominator too. Agency benchmarks published in 2026, Darkroom’s among them, put flow revenue per recipient at $0.15 to $0.25, an order of magnitude lower, because they measure a different base. Pick one and stay inside it. And leave out the $36 for every $1 line. Litmus publishes it as an average, not as the top of a range, and it comes from a survey in which marketers state the return they believe email produces. It is a statement about what marketers believe, not a measured return, and it is not a benchmark for your account.
The build order, and why it is that order
Build flows in order of how close the person already is to buying, then move outward. The demand at the top of this list already exists. You are collecting it, not creating it.
- Checkout abandonment. Highest intent in the account: they entered an email address and a shipping form. Two messages, the first inside an hour, no discount in the first unless you have decided to be a discount brand.
- Welcome. Three to five messages over seven to ten days. Message one delivers what you promised at signup. The rest do the job nobody builds for, explaining why the brand exists, so the second purchase has a reason beyond price.
- Post purchase and second order. The least glamorous flow and the most underbuilt. Shipping reassurance, then use and care, then a second product at the point where your data says people come back.
- Browse abandonment. Lower intent, much larger audience, so it earns on volume. Cap frequency and exclude anyone already in a cart or checkout flow.
- Replenishment or winback. Replenishment if the product has a consumption cycle you can measure. Winback if it does not, timed off your real median repurchase gap rather than a round number.
Past flow five, the return comes from segmentation and campaign discipline, not from adding a sixth flow. And none of this rescues a product page that does not convert. Flows move people to the store. The store still has to close.
Attributed revenue is not incremental revenue
As of September 2026, Klaviyo’s default attribution model is co-operative last touch. For accounts opened after 9 October 2024 the default conversion window is 5 days for email opens and clicks, 5 days for SMS clicks and 1 day for SMS opens, with 24 hours applying to push. All of them are adjustable in account settings, up to 90 days or 720 hours. With both channels running, Klaviyo credits whichever message was most recently clicked or opened inside its window.
So “email is 35% of our revenue” is a statement about credit allocation, not about what would have happened if the emails had never been sent. The flows reporting the biggest numbers, cart and checkout, have the weakest incrementality, because they intercept people already closest to buying.
The honest test is a holdout. Hold back 10% to 20% of eligible profiles at the profile level, persistently rather than per send. Run it 14 to 28 days depending on your purchase cycle, then compare completed purchases per profile between the treated and held groups. That difference is your incremental lift. Everything above it is credit, not causation. The number always lands below the dashboard and is usually still good enough to justify the work.
A list is not a CRM
A list is addresses plus consent. A CRM is identity resolved across email, phone, device and order, with an event history you can query and a consent record you can prove. Different assets, different costs.
Klaviyo now sells the second thing. It announced a B2C CRM on 20 February 2025, built on the Klaviyo Data Platform with deterministic identity resolution. Marketing Agent and Customer Agent followed on 25 September 2025, and Composer entered public beta on 30 June 2026.
The honest read for a brand between $500k and $20M: the CRM tier earns its price once you have a genuine second purchase motion or a real service load, and it does not while your event schema is a mess, because identity resolution over bad events gives you confident nonsense faster. Fix the schema first. If your store runs on Shopify, the customer record and the Klaviyo profile have to agree on who a person is before any of this is worth building. The same rule applies to any agents you run inside your own stack: an agent inherits your data quality, it does not repair it.
Segmentation that is worth the effort
Four cuts pay for themselves and most of the rest is theatre. Engagement recency. Purchase count at zero, one, and two or more. Value, meaning your top decile by revenue. Category affinity, where categories mean something to a customer.
Klaviyo’s own guidance, as of September 2026: a 90 day unengaged window as standard, 30 days if you send daily or several times a week, a never engaged segment at 180 days with at least five emails received, unengaged profiles excluded from campaigns on an ongoing basis, and a sunset flow capped at three messages before suppression.
One rule matters more than the bands. Define engagement on clicks, site activity and purchases, never on opens alone. Apple’s Mail Privacy Protection has been pre-fetching tracking pixels since its rollout with iOS 15 in September 2021, inflating opens for Apple Mail users whether or not anyone read the message. An engaged segment built on opens is a segment of Apple Mail users.
Deliverability is infrastructure, not best practice
As of September 2026, Google applies its bulk sender rules to anyone sending more than 5,000 messages a day to Gmail accounts. Yahoo publishes closely matching requirements for bulk senders without naming a volume threshold, so staying under Google’s threshold exempts you nowhere. Both want SPF and DKIM, a DMARC record with alignment, a spam complaint rate below 0.30%, and a working list unsubscribe header supporting one click unsubscribe, with opt outs honoured within two days. Google requires the RFC 8058 POST method for that header, while Yahoo, as of September 2026, calls that method highly recommended rather than mandatory. Google’s own guidance says to aim below 0.10%. Read that correctly: 0.30% is where enforcement starts, not where you sit. Microsoft applied comparable requirements to Outlook.com consumer domains from 5 May 2025, and non compliant mail is rejected with permanent 550 errors rather than filtered into junk. It is the cheapest insurance in the stack.
The five things that break silently in a Klaviyo account
1. Double opt in is on, so the welcome flow never fires
Symptom: signups climbing, welcome flow recipients flat. As of September 2026, every list in Klaviyo is double opt in by default, and Klaviyo’s own help documentation states it plainly. A person subscribes, a profile is created, they never click the confirmation link, and they are never added to the list, so a flow triggered by added to list never fires for them. Fix: either keep double opt in and treat the confirmation email as message zero of the welcome, or turn it off on a clean source and trigger on the subscription event instead. What is not defensible is the default, plus a list trigger, plus nobody checking.
2. The checkout consent box is not doing what you think
In Shopify, the marketing consent checkbox at checkout is a setting. It is one of the first things we check when we open an account, and it is often set to off. Turned off, a large share of buyers never join the list, so the post purchase flow runs on a fraction of your customers and retention looks worse than the business is. In the EU or UK there is a second constraint: a pre ticked box is not valid consent under GDPR, which requires a clear affirmative action, and the route that works for existing customers is the soft opt in under Article 13(2) of the ePrivacy Directive. Take that to a lawyer in your market rather than to a blog, this one included.
3. Smart Sending is quietly dropping message three
Symptom: a flow reports far fewer recipients than people who entered it. Smart Sending suppresses a message when the profile already received another on the same channel inside the smart sending window, which is why message three vanishes during a heavy campaign week. Fix: open Recipient Activity, read the Skipped reasons, then set smart sending per message rather than per account.
4. Flow filters and trigger filters do not behave the same way
Profile filters, which the Skipped reasons still label flow filters, are re-evaluated before every single send. Metric trigger filters are evaluated only at entry and never re-checked. The one exception is a custom object trigger filter, which Klaviyo checks at every step. So a profile can qualify on entry and still get dropped before message three, for a reason the canvas never displays. Fix: conditions that keep pruning, such as has not placed an order since starting this flow, belong in profile filters. Conditions describing the moment of entry, such as cart value above a threshold, belong in trigger filters.
5. The sunset flow is suppressing your good customers
If your unengaged definition runs on opens, Apple’s pixel pre-fetch has been marking dead profiles as engaged since September 2021 while readers who use privacy tools look inactive. You clean the list and keep the wrong half. Fix: rebuild every engagement segment on clicks, site activity and orders, then re-run the sunset.
The cost nobody puts in the model: how Klaviyo billing works
Klaviyo bills on active profiles, meaning any profile that can still receive marketing. Suppressed, unsubscribed and deleted profiles do not count towards billing, and as of September 2026 the free tier is 250 active profiles with 500 email sends a month. The paid ladder moves, so check klaviyo.com/pricing on the day you plan. A bloated list therefore charges you twice, once on the invoice and once in inbox placement, because sending to people who never engage is exactly the behaviour that complaint signals punish. Cleaning the list is not hygiene. It is a margin decision, and one of the few that improves two numbers at once.
The honest limits
Email does not create demand you have not earned. It collects demand, sequences it and stops it leaking. And a list built on a 20% welcome discount is a list of discount buyers, which shows up in your margin rather than your revenue chart, which is why it takes so long to notice.
Owned is not free either. It costs content, consent and deliverability maintenance forever, which is a recurring production cost, not a one off.
And owned is conditional. Regulators decide what consent means, inbox providers decide what arrives, Apple decides what you can measure. You own the permission. You rent everything around it, at a far better rate than you rent an audience on Meta.
Five things to check before lunch
- Compare people who entered your welcome flow against people who received message one. If the gap is large, check double opt in.
- Open Recipient Activity on your longest flow, click Skipped, read the top three reasons.
- Check the marketing consent setting at checkout, then what share of last month’s buyers carry consent.
- Open your engaged segment definition. If it contains the word opened, rebuild it on clicks and orders.
- Compare your active profile count against people who clicked anything in the last 180 days. The difference is your invoice.
At EDEUS we run this work inside client accounts rather than advising from outside, which is why this reads as a list of settings rather than principles. If you want the same audit on yours, the checklist above is the first hour of it.
Frequently asked questions
Which Klaviyo flows should I build first?
Build them in order of how close the person already is to buying: checkout abandonment, then welcome, then post purchase and second order, then browse abandonment, then replenishment or winback. The order matters because the first flows intercept demand that already exists rather than trying to create it. Past the fifth flow, more revenue comes from segmentation and campaign discipline than from adding another flow.
Why is my Klaviyo welcome flow not sending?
The most common cause is double opt in. As of September 2026, every Klaviyo list is double opt in by default, so someone who subscribes but never clicks the confirmation link gets a profile created but is never added to the list, and a flow triggered by added to list never fires for them. Compare the number of signups with the number of people who received message one. If there is a large gap, either trigger the flow on the subscription event instead of list membership, or keep double opt in and treat the confirmation email as the first designed message of the welcome.
Why does my Klaviyo flow show skipped recipients?
Most skips come from Smart Sending, which suppresses a message when the profile already received another message on the same channel inside the smart sending window. Profile filters, which the Skipped reasons still label flow filters, are re-evaluated before every single send, so a profile that qualified on entry can be dropped at message three. Metric trigger filters, by contrast, are evaluated only at entry and are never re-checked. The one exception is a custom object trigger filter, which Klaviyo checks at every step. The Skipped tab inside Recipient Activity gives the specific reason for each profile.
Should double opt in be on or off in Klaviyo?
Turn it off if your signup sources are clean and you trigger the welcome on the subscription event. Keep it on if you buy or import traffic you cannot vouch for. Double opt in protects deliverability at the cost of losing everyone who never confirms. Single opt in captures more subscribers and requires a clean source plus real suppression discipline. What breaks accounts is leaving the default on, triggering the welcome flow off list membership, and never checking the gap between subscribers and recipients.
Is Klaviyo flow revenue real, incremental revenue?
Not entirely. As of September 2026 Klaviyo attributes revenue on a co-operative last touch basis, with a default conversion window of 5 days for email and 5 days for SMS clicks in accounts opened after 9 October 2024, so attributed flow revenue includes purchases that would have happened without the message. Cart and checkout flows are the most inflated, because they intercept people already closest to buying. The only honest measure is a holdout: withhold 10% to 20% of eligible profiles persistently at the profile level for 14 to 28 days, then compare completed purchases per profile between the treated and held groups.
What percentage of revenue should come from email?
There is no correct number, and every figure you see is an attributed figure rather than a measure of incremental contribution. For context, as of September 2026 the Klaviyo 2026 Omnichannel Benchmark Report puts flows at about 5.3% of email sends and about 41% of email revenue, while agency benchmarks published in 2026, Darkroom’s among them, put email and SMS combined at 30% to 40% of total revenue. Do not mix the two, because the denominators differ. Note also that a rising email share can simply mean paid acquisition shrank.
Why did my Klaviyo bill go up?
As of September 2026, Klaviyo bills on active profiles, meaning any profile that can still receive marketing, so a list growing in unengaged profiles raises the bill without raising revenue. Suppressed, unsubscribed and deleted profiles do not count towards billing, which means suppressing dead profiles lowers the invoice and improves inbox placement at the same time. Check current tiers on klaviyo.com/pricing rather than a comparison blog, since the pricing changes.
What do Google, Yahoo and Microsoft require from bulk senders?
SPF and DKIM, a DMARC record with alignment, a spam complaint rate below 0.30%, and a working list unsubscribe header supporting one click unsubscribe, with opt outs honoured within two days. As of September 2026, Google applies these rules to anyone sending more than 5,000 messages a day to Gmail accounts, requires the RFC 8058 POST method for that header, and says in its own guidance to aim below 0.10%. Yahoo publishes closely matching requirements for bulk senders without naming a volume threshold, and as of September 2026 calls the RFC 8058 method highly recommended rather than mandatory, so staying under Google’s threshold does not exempt you there. Microsoft applied comparable requirements to Outlook.com consumer domains from 5 May 2025 and rejects non compliant mail with a permanent 550 error rather than filtering it to junk.
Klaviyo and the Klaviyo logo are trademarks of Klaviyo, Inc. EDEUS Studio is a Klaviyo partner. Photography by EDEUS Studio.
Content created with the support of artificial intelligence
