Podcast: Download (Duration: 28:31 — 39.4MB)
The counterintuitive fact about email deliverability is that inbox providers know nothing about your sales and do not care. A customer who buys from you every month but never opens an email is, from Gmail’s perspective, evidence that you are a bad sender.
Jake Cohen works on deliverability at Klaviyo, my email platform of choice. I recorded this with him at the Klaviyo conference in Boston immediately after his keynote.
Below is how different inbox providers evaluate you, which subscribers to suppress, and the story of a brand that went from 16 million email addresses to 30,000 and made more money afterward.
Get My Free Mini Course On How To Start A Successful Ecommerce Store
If you are interested in starting an ecommerce business, I put together a comprehensive package of resources that will help you launch your own online store from complete scratch. Be sure to grab it before you leave!
Table of Contents
Key takeaways
- Inbox providers judge you on opens alone. Purchase history is invisible to them.
- Hotmail penalizes you for sending to anyone who has not opened in six months.
- Gmail is more granular and penalizes you per recipient, then overall if enough recipients ignore you.
- Build segmentation logic requiring an open within the last three to six months depending on send frequency.
- Suppress anyone who has received 10 or more emails in two years and opened none.
- Automated flows have higher open rates than campaigns because they are triggered by customer behavior.
- JackThreads cut from 16 million addresses to 30,000, rebuilt to 300,000, and earned more than at 16 million.
- Sending to unengaged subscribers signals to inbox providers that you are not cleaning your list.
Why your sales data does not affect deliverability
Jake’s central point reverses how most store owners think about their list.
As a store owner, someone who buys from you is obviously engaged and obviously interested. From an email perspective they may never open anything you send.
Inbox providers have no visibility into your sales. They cannot see that this person is your best customer, so all they observe is a recipient who consistently ignores your messages.
Taken to the letter, your sales data does not matter at all for deliverability. What matters is opens.
That means you can actively harm your deliverability by mailing people purely because they are customers rather than because they engage with email.
How different inbox providers evaluate you
| Provider | How it evaluates senders |
|---|---|
| Hotmail | Looks at engagement over roughly six months with little granularity beyond that. Sending to anyone who has not opened in six months gets you flagged and penalizes your placement. |
| Gmail | Substantially more sophisticated and granular. A daily sender whose recipient does not open for 7, 14, or 21 days gets penalized for that individual, and enough of those produces an overall penalty. |
In both cases the signal is opens.
The segmentation rule that protects you
Jake’s practical recommendation is including logic in your segments requiring an open within a recent window, typically three to six months depending on how often you send.
That keeps your sends going to people who are generally engaged, which keeps you out of deliverability trouble.
How to use flows to earn the right to send campaigns
The cleanest strategy Jake describes uses automated flows to do the work campaigns cannot.
Flows have consistently higher open rates than campaigns, because they are triggered by events the customer actually went through rather than by your calendar.
The sequence: someone signs up through a form and enters a welcome series. When they open those messages, they enter the segment defined by recent engagement, and only then do they start receiving campaigns.
Separately, build a flow targeting people who have not opened in a long time, offering them something to bring them back into the engaged group.
Some senders go further and send campaigns only to people who have opened, letting automated flows handle everything else.
The failure mode is the opposite: sending campaigns constantly to a large list including people who have not opened recently. That produces enough volume and enough negative signal for inbox providers to classify you as a bad sender.
Sequencing sends by engagement
My own observation is that emailing people more likely to open first improves overall deliverability, so I mail the most engaged group and reach less engaged people later.
Jake confirmed the logic works, and it is a matter of how strategic you want to be about ordering.
Which subscribers to suppress
Every three to six months there is a group of profiles you should suppress entirely, meaning stop sending to them.
Jake acknowledges this is emotionally hard. There is apparent value sitting in those addresses and the feeling that one of them might eventually buy.
His conservative rule: build a segment of people who have received 10 or more emails over the past two years and opened none of them, and suppress those.
He calls this very conservative and says it will not materially harm your business while helping your deliverability.
The reason it matters is what continued sending signals. Mailing people who never open tells inbox providers you are not cleaning your list.
Jake attends M3AAWG, the Messaging, Malware and Mobile Anti-Abuse Working Group, where the people who effectively set email norms meet. His characterization is that they are actively looking for ways to penalize senders who do not respect inboxes.
His summary principle: think in terms of earning the right to send, remove people who do not want it, and be respectful.
The JackThreads story
The cautionary example Jake shares came from a panel the previous day.
JackThreads had 16 million email addresses and discovered they were largely unengaged. They cut down to 5 million, then 1.1 million, then 300,000, then 30,000 addresses.
Along the way they were mass blocked by Gmail, with every email going straight to spam because they had not respected the inbox.
They rebuilt to 300,000 and now make more money with 300,000 addresses than they did with 16 million.
Why list size is a poor metric
That story is the argument against optimizing for list size.
An unengaged address is not a neutral asset sitting in reserve. It actively costs you placement with every send, which reduces how many of your genuinely interested subscribers see anything at all.
Jake’s closing reassurance is that a good platform monitors sending across its shared infrastructure, removing bad senders and promoting good ones, so a sender who respects inboxes should not need to obsess over the mechanics.
What Klaviyo announced at the conference
The rest of the conversation covered features Klaviyo announced in 2018. These have since shipped and evolved considerably, so treat this as context for how the platform thinks rather than a current feature guide.
Custom objects
The feature that interested me most lets a store pass arbitrary business-specific data into Klaviyo and use it to trigger flows.
Jake’s example is gift cards. Many businesses issue them and recipients frequently forget the balance or never spend it fully.
Unspent balance, called breakage, is not necessarily a problem since you already have the cash. What Klaviyo found from store owners is more interesting: a meaningful share of recipients spend beyond the card value, using it as a reason to try more products.
The gift card effectively subsidizes the start of a customer relationship, and if the products are good those people return as full-price buyers.
The mechanism passes the gift card as an object with its creation date, current value, and last change. A flow can then trigger when the balance changes, and 30 days later email anyone with a remaining balance above zero, including recommended products.
Customizable analytics and report builder
Klaviyo polled customers on what to improve and the answer was overwhelmingly reporting: the data was all there and impossible to see.
The response was a fully customizable analytics dashboard with sensible defaults, plus a report builder for more specific questions.
Jake’s example question: when someone buys shoes first, what do they most often buy next. The data already exists in the platform, so the problem was purely exposing it.
The relevant point for store owners is that your email platform holds complete sales data, where Google Analytics is harder to extract from and can be incomplete when pixels fail to fire.
Predicted lifetime value
Klaviyo’s data science team built models predicting what each individual profile will spend over the next year, computed on demand so the results reflect every recent purchase rather than a stale stored value.
Jake’s step-by-step recommendation for using it: create a segment of all profiles, export it with the predictive analytics included, find the average predicted lifetime value, identify the top 20% threshold, and build a segment above that number.
Then push that segment to Facebook as a custom audience, build a lookalike from it, and target that lookalike for acquisition.
On emailing that segment directly, his caution is against simply blasting it. The better use is staggering a launch or sale, starting with top spenders, with content acknowledging they are being given first access before everyone else the following week.
On-site personalization
The feature in design at the time used the same profile data to personalize the storefront, showing recommended products based on what someone had viewed or purchased, pulling best sellers, excluding prior purchases, and reordering by observed interest.
You can find Jake’s work at Klaviyo.
Frequently asked questions
Does buying from a store improve email deliverability?
No. Inbox providers have no visibility into your sales data, so a frequent customer who never opens your email still counts as an unengaged recipient and can harm your sender reputation.
How does Gmail evaluate email senders?
Gmail is granular and penalizes you per recipient. If someone does not open your messages over 7, 14, or 21 days as a daily sender, Gmail penalizes delivery to that person, and enough of those produces a penalty across your whole list.
How does Hotmail differ from Gmail on deliverability?
Hotmail evaluates engagement over roughly a six month window with little granularity beyond that. Sending to anyone who has not opened within six months gets you flagged and hurts your inbox placement.
Which subscribers should you suppress?
A conservative rule is suppressing anyone who has received 10 or more emails over two years and opened none of them. This will not materially damage your business and will improve deliverability.
Why do automated flows have better open rates than campaigns?
Because flows are triggered by something the customer actually did rather than by your sending schedule. That makes them a reliable way to move new subscribers into your engaged segment before you start sending campaigns.
Is a bigger email list always better?
No. JackThreads cut from 16 million addresses to 30,000 after being mass blocked by Gmail, rebuilt to 300,000, and now earns more than they did at 16 million.
How recently should someone have opened before you email them?
Build segments requiring an open within the last three to six months, depending on how frequently you send. More frequent senders need a tighter window.
How can you use predicted lifetime value for advertising?
Export your profiles with predictive analytics, identify the top 20% by predicted lifetime value, and build a segment above that threshold. Push that segment to Facebook as a custom audience and build a lookalike from it for acquisition.
