On Tuesday, October 6, at its Agentic World conference in Miami, Treasure AI said it will start charging for email based on what people do with a message, such as clicking, instead of how many messages a brand sends. The company used to be called Treasure Data and is best known for its customer data platform. Now it wants to be your email platform too, and it picked a fight to get there.
“As the industry talks about outcome-based pricing, the real test is whether a platform’s economics align with the value it helps create,” Rafa Flores, Treasure AI’s chief product and growth officer, said in the announcement. To MarTech he put it more simply: “We’re betting on the click-through rate.” To CMSWire he named the targets. Braze and Klaviyo, he said, can change their pricing or risk losing customers. “The bet is made, so show me your cards now.”
I like the bet. Paying for volume has always rewarded the wrong thing. A platform that earns more when you send more has no reason to tell you to send less.
But there is a problem sitting right in the middle of this idea, and it is worth solving before anyone signs a contract. Today, an email click is a number in a report. Under this model it becomes the unit on your invoice. And a real share of the clicks your email platform records are not made by people at all.
Here is my read. When the click becomes the price, the definition of a click becomes the contract. Whoever writes that definition decides what you pay.
What Treasure AI announced, and what it did not
The announcement itself is short. The press release says the new model “ties email costs to customer actions, such as clicks, rather than the number of messages sent,” and frames it as a dare to the rest of the market: “if our intelligence doesn’t help you drive engagement and conversions, why should you pay for send volume alone?”
What it does not include is almost everything a buyer would need. CMSWire reported that the company did not disclose rates, tiers or a start date. Two days later, the same outlet noted Treasure AI had not said whether the model includes a guaranteed click rate or simply charges per click with no floor. As of Sunday morning, Treasure AI’s public pricing page still describes annual subscriptions priced on customer profiles and behaviors, plus credits for AI features. There is nothing about what a click costs, which clicks count, or how the company will tell a person from a machine.
That is normal for a conference announcement. It is also the part that will decide whether this is a good deal.
Why the click, and why now
The click did not become the currency of email by accident. In 2021, Apple’s Mail Privacy Protection started preloading tracking pixels for people who turned it on, which made open rates unreliable as a signal. Mailchimp’s own guidance since then has been blunt: focus on clicks, because clicks and purchases are stronger signs of engagement than opens. So the click is the last engagement number most email teams still trust.
At the same time, the way marketing teams pay for software is moving. In Gartner’s 2026 CMO Spend Survey of 401 CMOs, 56 percent said they had increased the share of their martech budget going to consumption-based pricing over the past year, against 9 percent who cut it. Customer service AI went first: Zendesk and Intercom price their AI agents per resolution, and in April HubSpot moved its Customer Agent from $1 per conversation to $0.50 per resolved one.
Email was the obvious next place to try it, because email already has a clean, countable event at the end of every message. That is exactly the thing worth looking at more closely.
Machines click first
Most email platforms already know this, and their help pages say so plainly. Klaviyo’s documentation explains that inbox providers, security software and mobile carriers “use bots to click links in emails before any human user,” to check that the links are safe. On SMS, the link preview your phone draws in a message thread is generated by programmatically clicking the link. “There is no way to prevent bot clicks performed by inbox providers entirely,” Klaviyo adds.
Microsoft is open about it from the other side. Its Safe Links documentation for Defender for Office 365 says that while the feature is on, “URLs are scanned prior to message delivery,” and links without a good reputation are “detonated asynchronously in the background.” Through Microsoft’s built-in protection policy, it covers every recipient by default in any organization with at least one Defender for Office 365 license. In June 2023, a change in how Safe Links scanned URLs caused many Office 365 mailboxes to generate clicks on almost every link in the messages they received, according to the email platform ClickDimensions, which investigated it with Microsoft.
How big is the effect? Nobody has a single number, partly because the scanners are built to be hard to spot. The industry anti-abuse group M3AAWG concluded, as ClickDimensions quotes it, that it is “hard to quantify a summary effect, as most filter agents and ISPs work to mask detection.” Its research, as summarized by ClickDimensions, put the impact under 10 percent when sending to big consumer mailbox providers like Gmail and Yahoo, and heavier for business, nonprofit and university inboxes that run stricter security. Mailchimp says its internal testing finds roughly one in five clicks are bots with basic detection, and close to 50 percent once it adds third-party intelligence.
Pick whichever number fits your list. It is not zero, and for a B2B sender it can be a lot.
When a bot click becomes a line item
Here is the sentence from Klaviyo’s help page that made this click for me: “Bot clicks do not influence the cost of a campaign.”
That is true today because email platforms do not bill on clicks yet. Klaviyo prices its email plans on active profiles with send capacity included. Most of the market charges on some mix of contacts, users and volume. Under those models, a scanner that taps every link in your newsletter is a reporting nuisance. It inflates your click rate, it can quietly pick the wrong winner in an A/B test, and it can steal last-click credit for a sale. Annoying, but it never shows up on the invoice.
Under click pricing, that same scanner is spending your money, unless the contract says it is not. Mailchimp notes that security bots tend to click every link in a message, “including your logo and the Unsubscribe button.” Think about what that means for a bill priced per click. A corporate firewall checking a newsletter with twelve links could, in principle, generate twelve billable events before a human has seen the subject line. An unsubscribe click is a click. A click on your privacy policy is a click. A link preview in a text thread is a click.
To be fair to Treasure AI, I have no reason to think it plans to charge for any of that. Every serious email platform already flags bot clicks, and a vendor betting its reputation on “our intelligence is so good your emails will be clicked on” has every reason to count carefully. But it has not published how it will count yet. CMSWire’s own recommendation to buyers was the right one: ask for rates, the click definition and scanner-click handling before committing.
The definition is the price
If you want to see how much an outcome definition matters, look at the vendor that has already written one down.
HubSpot’s help center spells out what counts as a resolved conversation for its Customer Agent, which costs $0.50 each. The agent has to reply with a content source or take an action, and there must be no visitor-initiated handoff to a human within 72 hours of the visitor’s last response. Two details stand out. A conversation where a human rep is assigned manually, including through a workflow, can still count as resolved. And once the status is set, “subsequent actions, such as messages from a live agent, a transfer request, or negative feedback, will not change the resolution status.” For the lead qualification path, a lead marked “not qualified” counts as a resolution too.
None of this is hidden or unfair. HubSpot published it, and buyers can read it. But it shows the point. Two companies can both say “you only pay for outcomes” and mean very different things by the word, and the gap between those meanings is where your budget goes.
For email, the honest version of a billable click needs answers to a handful of questions. Does it exclude clicks the vendor’s own bot filter flags? Does it count one click per person per message, or every tap? Does it exclude unsubscribe, preference center, view-in-browser, logo and legal links? What happens to clicks inside the first seconds after delivery, when scanners do their work? Can you audit the raw events? And is there a ceiling, so a scanner storm like June 2023 cannot double your bill in a week?
The incentive moves too
There is a second, quieter change, and this one is my inference rather than anything Treasure AI has said.
When the vendor is paid per click, the vendor wants clicks. Most of the time that lines up with what you want. Better targeting, better timing and better content all produce more clicks, and you should be happy to pay for them.
But some clicks are cheap to produce and worth very little. A curiosity-gap subject line gets clicks. A discount in every send gets clicks. More links per email gets more clicks. An AI system tuned toward click-through will learn all of that faster than any human team. If your vendor’s optimizer is writing or ranking your email, and it is paid on clicks, ask what keeps it from teaching your list to expect a coupon.
Lindsay Marty, who runs an agency that markets law firms, named the gap in an email to CMSWire: “The open question is what happens when clicks go up but revenue doesn’t follow, since a click is still a proxy and not a sale.” Scott Brinker, who has tracked the martech market for years, called a click “a very reasonable outcome” to pay for, but only as an intermediate one. Both are right. A click is a fair thing to pay for as long as something further down the funnel is watching it.
What to do this week
You do not need to be a Treasure AI prospect for any of this to matter. If Flores is right that rivals will be pushed to respond, outcome pricing is coming to your email contract in some form, and probably at your next renewal.
Pull your human click baseline. Turn on your current platform’s bot filtering in reporting, or export click events with the bot flag, and compare the last three months of click rates with and without it. Split it by B2B and B2C domains if you send to both. That gap is roughly the share of your clicks that a careless definition would put on your bill.
Write your own definition of a billable click before a vendor writes it for you. One page is enough: human-filtered, unique per recipient per message, excluding unsubscribe, preference, legal, logo and view-in-browser links, with a short delay window after delivery, raw event access for audits, and a monthly cap. Send it to any vendor pitching engagement or outcome pricing and ask them to mark it up.
Pair the click with something a bot cannot fake. Revenue, orders, sign-ups or time on site, measured in your own analytics. If clicks rise and those do not, you want that conversation in the contract, not in a quarterly review.
Check who controls the optimizer. If the vendor’s AI picks subject lines, send times or offers, ask how it is tuned, and whether discount frequency or link count can creep up without your approval.
Use the announcement at renewal. If your current platform bills on contacts or volume and your contract is up in Q4, ask what an engagement-based option would look like. You may not want one. The question is still useful leverage.
And do not switch platforms on a press release. Treasure AI has made a bold, useful bet that pushes the market in the right direction. It has not yet published a price, a definition or a start date. When it does, the definition is the first page to read.
Paying for clicks instead of sends is a better deal for marketers in principle. It is only a better deal in practice if the click on the invoice belongs to a person.
