Orita Review: 6 Weeks of Real Data From an 8-Figure Brand

We ran Orita on an eight-figure brand for six weeks. Click rate jumped 31%, and inside Klaviyo the strategy looked close to free.

We still turned it off.

This is the review I wish existed before we started: what Orita actually does, what the numbers looked like under three different attribution sources, and why "send only to your engaged subscribers" makes every dashboard look better while real revenue can suffer.

What Orita Is and What It Promises

Orita is an AI list-management tool that sits on top of Klaviyo. It builds engagement-scored audiences for your campaigns, suppresses profiles it considers dead weight, and unsuppresses profiles when it detects buying signals (they call this rescued revenue).

The pitch is compelling. Only a small slice of your list ever buys. So stop mailing the dead weight, send to the people who actually engage, and watch your open rates, click rates, and sender reputation climb. Better reputation means better inbox placement, and better placement means everything you send works harder.

In theory, great. In practice, it was a lot more complicated. And to be fair to Orita before the numbers start: their team was responsive and genuinely collaborative throughout, and one of their recommendations (more on it below) was correct and we're keeping it. This isn't a story about a bad vendor. It's a story about our experience and why it didn't make sense for this brand.

The Setup: A Real Before-and-After Test

The brand does eight figures a year through Shopify with a list in the high six figures. The old strategy: broader sends went to the majority of the list, minus a set of negative-engagement exclusions (recent bounces, spam complaints, chronic non-engagers), and our own automated suppression flow removed people entirely once they hit 12 months with no engagement and no purchase. Averaged across the 42 national sends in the comparison window, that worked out to about 166K recipients per send.

On July 1 we cut over to Orita's model. Everyday sends went to their engagement-scored audiences (about a third the size of the old sends). Only sale moments went broad. That gave us a clean comparison: six and a half weeks of the old strategy (May 15 to June 30, 42 national sends) against three weeks of the new one (July 1 to 22, 19 sends). All data pulled from Klaviyo via API, then checked against Google Analytics and Shopify.

First, the Part Orita Got Right: Your Open-Based Revenue Is Inflated

Orita's team recommended we change Klaviyo's attribution settings to ignore Apple privacy (MPP) auto-opens. Apple opens a large share of emails by machine, Klaviyo counts those as engagement, and any order that follows within the attribution window gets credited to the email. We made the switch, and Klaviyo recalculated history. Here's what that did to the same sends:

A typical campaign went from $15.1K to $9.7K. Nothing about the store changed. Shopify order totals held steady through the whole period. What changed is credit: about a third of the "email revenue" on full-list sends was attributed off Apple bot opens, from orders that would have happened anyway.

Notice the before period dropped twice as hard as the after period. That's the strongest single argument for Orita's worldview: full-list sending mails the segments dominated by machine opens, so its revenue numbers carry the most inflation. If your campaign reporting relies on open-based attribution, a chunk of it is fiction, and we've written about how email attribution behaves across brands before. This recommendation we kept. Every account should probably make this change.

The Scoreboard: What Improved and What Didn't

Here's the before-and-after on national sends, with both periods restated under the cleaned-up MPP-excluded attribution:

Start with what's uncomfortable for the thesis: clicks were the only clean engagement win. Open rates fell five points and bounce rates ticked up, on audiences that were supposed to be the most engaged people on the list.

Now the asterisk, because it's the most important thing in this review. That +19% revenue per recipient only exists under the restated attribution, which hit the full-list period twice as hard as the engaged period. Run the same comparison under the attribution that was actually live at the time and revenue per recipient went down, from $0.090 to $0.084. Same sends, same orders, opposite verdicts.

The rows that don't depend on attribution at all are the last two: each campaign reached 36% fewer people and made 24% less. You can always make rate metrics look better by cutting reach. Volume and revenue are correlated in a way that efficiency dashboards hide.

The Only Question That Matters: Is This Real Money or an Attribution Artifact?

Inside Klaviyo, the drop was defensible. Total store demand fell about 24% from the May/June run rate into July (summer slowdown plus a soft month), and campaign revenue fell 24.5%. Campaign revenue's share of store revenue was essentially unchanged: 12.4% before, 12.1% after. Measured against the business, the strategy looked close to free.

Then we checked the sources Klaviyo can't touch:

• Google Analytics: email-channel revenue down 38% month over month, against a 29% sitewide drop
• Shopify: Klaviyo-attributed revenue down roughly 50% month over month

GA and Shopify use their own attribution. Nothing we changed in Klaviyo could restate their history. Both said the same thing: email was falling meaningfully faster than the business. Shopify is usually the stingiest attributor of all, and it showed the biggest drop.

That was the tiebreaker. Not a feeling, not an open rate. Two independent measurement systems agreeing that engaged-only sending was leaving real dollars on the table.

Why the Deliverability Bet Never Paid: You Need a Venue to Cash It

The engaged-only thesis has a specific mechanism: higher click rates improve sender reputation, reputation improves inbox placement, and placement lifts revenue when you go broad for the moments that matter. The click rate did improve. But the payoff step never showed up.

Broad sale sends after the switch earned $0.042 per recipient at a 0.24% click rate. Historically, sale sends on this account did $0.065 to $0.070 per recipient. The sends that were supposed to benefit from all that accumulated reputation did worse, not better. Six weeks is short, but the arrow was pointing the wrong way.

And this exposes the structural problem: the strategy needs frequent big broad moments to convert reputation into money. A brand doing weekly sales has lots of them. This brand has a handful outside of Black Friday. Putting the whole year's thesis on BFCM performing better because of clicks accumulated in July is not a bet a sane operator takes. Meanwhile a 31% click rate lift, on an absolute click rate that starts under half a percent, is just not worth much on its own. If your deliverability is genuinely broken, this trade can make sense. Ours wasn't. Opens were in the mid-40s before Orita ever touched the account.

The fit problem showed up in the day-to-day too. Orita's answer to the lost volume was to expand total sending using their hyper-engaged audiences: keep your normal calendar, then layer extra sends to the people most likely to buy. If you send 10 campaigns a month, that play works. You add a handful of low-risk sends to your best group and claw back revenue without touching the broad list. This brand sends 20-25 days a month. There was no room to expand. The extra-send play assumes spare calendar, and a near-daily sender has none. A tool's default plays tell you who it's built for, and these were built for brands with headroom in their sending schedule.

The Tell That It Wasn't Working

On the call where we made the decision, the client said the thing that ends most tool experiments: "We're pumping time and energy into making it work, and we're not sure if it even is. The fact we're even having this conversation is probably enough."

He was right. Six weeks in, we had built comparison spreadsheets with 14 tabs, restated attribution twice, and run seasonality controls against last year, all to answer the question "is this thing helping?" A strategy that's working doesn't require forensic accounting to detect. And you definitely shouldn't be paying a monthly fee to lose revenue and get click rate back.

What We Kept From the Experiment

Reverting didn't mean going back to exactly what we had. Four things survived:

• The MPP attribution fix. Excluding Apple auto-opens from Klaviyo attribution stays. The inflated numbers were fooling us too.
• In-house automated suppression. Our suppression flow (12 months unengaged and never purchased gets one recapture email, then suppression) comes back on. Same list hygiene Orita provides, no per-month fee, and we control the thresholds.
• Purchase-triggered reactivation. Orita's "rescued revenue" idea, scaled down: if a suppressed profile buys, a flow reactivates them. Their version unsuppressed about 50,000 profiles at once, which is a deliverability grenade the moment you leave their engagement-only audiences. The signal is good; the blast radius wasn't.
• Occasional hyper-engaged sends. Layering a few sends per month to a genuinely hot audience (site visit, cart, or click in the last 180 days) on top of the broad cadence. Concentration as a seasoning, not the whole diet.

That last point is the real lesson. Engagement data is useful for excluding the truly dead and for occasionally concentrating. It's a bad basis for your default audience, which is a case we've made before: segment by purchase behavior, not engagement. Buyers don't open every email. Openers don't buy. A lot of the revenue in that "unengaged" base comes from past purchasers who ignore ten emails and then spend $500 on the eleventh.

So Is Orita Worth It?

First, to be clear: this is not a knock on Orita. Their team was incredibly helpful, genuinely wanted this to work, answered everything we threw at them, and pushed us toward an attribution fix we're keeping. We think they could be a great option for plenty of brands. This is one account, one sending strategy, and what happened when we measured it.

Based on our test: not for a brand like this one. If you have healthy deliverability, a purchase cycle with real gaps between orders, and only a few big sale moments a year, engaged-only sending trades real revenue for rate metrics you can't deposit. Two independent attribution sources priced that trade in real dollars, and all we got back was a click rate gain with no venue to cash it.

Where I'd consider it: a brand in a genuine deliverability hole (sub-20% opens, spam placement), a list that's never been cleaned, or a brand sending a few times a week with room to add the extra hyper-engaged sends the model counts on. The theory isn't wrong everywhere. It was wrong here, and I suspect it's wrong for more brands than the engagement-rate screenshots on social media suggest.

One test, one brand, six weeks. Take the sample size for what it is. But it's six more weeks of independently verified revenue data than most opinions about engaged-only sending are built on.