Your finance team asks what the spam folder costs. The honest answer is a worksheet, not an industry statistic with a dollar sign painted on it.
You can estimate the exposure. You cannot pull a single “ecommerce loses $X” number out of a placement study. This page is the estimate, the inputs it needs, and the ways it lies if you skip them.
The only placement fact this page borrows
InboxEagle’s Q1 2026 ecommerce benchmark covered 721,351 emails and 3,474 sender domains from January through March 2026. In that set, 21.6% of messages landed in spam, 59.7% in Promotions, and 9.9% in Primary. Combined placement in any inbox tab was 78.2%. The median domain’s inbox rate was 96.1%, which is a different cut than the message-weighted 78.2%. A few large senders can move the message-weighted number. Your domain might sit at either end.
That paragraph is context. It is not your revenue loss. It is not a full-year 2026 result. It is not “all ecommerce email.” The methodology limits are on that article. If you paste 21.6% into a board slide as your spam rate, you are using someone else’s dataset.
Delivery versus inbox placement is why the ESP’s 99% delivered figure never belonged in the revenue line by itself.
The email deliverability ROI estimate, with your numbers
Use one period. A month is enough.
- Attributed email revenue for the period. Call it R. Use the attribution your finance team already argues about. Do not switch models mid-estimate.
- Messages accepted in that period. Call it M. Accepted, not “attempted,” if those differ.
- Revenue per accepted message: R / M. This is an average. It hides that a cart email and a newsletter are not worth the same.
- Share of accepted mail that went to spam, from a placement test or from monitored sends in the same period. Call it S. This is a fraction, not your open rate.
- Illustration of revenue sitting on mail that went to spam: (R / M) × (M × S), which is R × S.
Work a blank example so the shape is obvious and the dollars are not. If R is revenue you type in, and S is 0.10 because a seed test said so, the illustration is 10% of attributed email revenue. If S is 0.02, the illustration is 2%. The revenue calculator will do the arithmetic from inputs you type. The output is only as good as those inputs. It is not a measurement of your account until you supply them.
Do not multiply by the Q1 study’s 21.6% unless that is actually your spam share. Do not multiply Promotions into S. In the Q1 set, Promotions was the majority of messages and is still a folder people shop from.
What the multiplication leaves out
- Not every unseen email was a sale. The formula treats average revenue as if it applied to the spam slice. The people filtered to spam may be the least engaged, so the true lost revenue can be lower than R × S. It can also be higher if your best customers are on a provider that is filtering you. You do not know which until you split placement by provider.
- A seed test is a sample. InboxEagle’s seeds are Gmail, Yahoo, AOL, and Outlook. They are not your full list. An Outlook
550never enters S, because those messages were not accepted, so they never reach a folder. Count bounces as failed delivery, not as spam placement. Delivered but not received separates the cases. - Attribution double-counts. Email revenue that ads or the site also claim will inflate R. Say which model you used.
- Fixing placement does not pay R × S back. Some of that mail, once visible, still will not convert. Judge the project by the spam share and the complaint rate after the change.
Where the money usually leaks, in store terms
The streams worth splitting, because their revenue per message is not the average:
- Abandoned cart and browse, which are close to a purchase and painful to lose to spam.
- Welcome, which is the first reputation impression and a high-intent message. Welcome series risk is the deliverability side.
- Campaigns to a cold segment, which move S up and are a poor place to “make the number back.”
- Order and shipping mail, which may not be in your attributed marketing revenue at all, and still cost support time when they disappear.
If only one provider is in spam, the dollar figure should use that provider’s share of the list, not the blended S. Gmail troubleshooting and Yahoo are the splits.
What to do with the number
If R × S is small against the work, monitor so it does not grow, and do not fund a domain migration to chase it. If it is large, the next page is the diagnosis, why mail is in spam, not a new ESP. Recalculate S after the fix, same way you calculated it the first time.
The first 24 hours is the right sequence when S moved suddenly. ROI language does not change the order: stop the harmful send, then measure.
InboxEagle
Put your revenue and your placement in the same month
The calculator runs the multiplication. A domain check and a seed test are how you get a spam share that belongs to you, instead of a benchmark from another dataset.
Sources
- Ecommerce email benchmarks, Q1 2026, InboxEagle. 721,351 emails, 3,474 sender domains, January through March 2026. Placement shares only.
- Email sender guidelines, Google. Cited because spam rate and spam placement are different metrics. Do not put Google’s 0.10% complaint guideline into this formula as if it were a revenue rate.