Was the message accepted?
Delivery generally concerns whether a message avoided an immediate failure such as a bounce. Acceptance by the receiving system does not necessarily tell you where the message appeared afterward.
Evidence Education
Email deliverability claims: what do the percentages actually prove?
A vendor says 97%. Another test says 91%. A third platform promotes a different benchmark entirely. Those numbers may look comparable even when the tests behind them are not. Before declaring a deliverability winner, inspect what was measured, how it was measured, and whether the comparison is genuinely matched.
Start with the vocabulary
Delivery, deliverability, and inbox placement are not interchangeable.
A percentage becomes meaningful only after the outcome being counted is clearly defined.
Where did the message land?
Inbox-placement testing attempts to distinguish messages that appear in an inbox from messages routed elsewhere, such as a spam or junk area.
What happened after delivery?
Opens, clicks, replies, conversions, and other recipient actions are separate outcomes. They can be influenced by far more than sending infrastructure alone.
A “98% delivery rate” is not the same claim as “98% inbox placement.”
Before comparing numbers, verify that the numerator, denominator, and
measured outcome mean the same thing.
The comparison problem
Two accurate percentages can still create an invalid comparison.
If the underlying tests differ, ranking the products from the final percentages alone can create false precision.
The mistake is not using percentages.
The mistake is treating percentages produced under different
conditions as though they came from one controlled experiment.
A stronger comparison
Better evidence holds important variables as constant as practical.
No real-world email test is perfectly controlled, but comparison quality improves when competing products are evaluated under similar conditions.
Even matched testing answers a narrower question than “Which ESP is best?”
It tells you how those systems performed under the particular tested
conditions. Generalization beyond those conditions still requires care.
Example
Why 96% versus 92% may not mean what it appears to mean.
Hypothetical comparison
Platform A reports 96%. Platform B reports 92%.
If Platform A's number came from a recent vendor-controlled test using one recipient mix, while Platform B's number came from a third-party study using another list, different content, a different time period, and a different definition of success, subtracting 92 from 96 does not produce a defensible four-point platform advantage.
Both percentages may accurately describe their own tests. The comparison becomes weak when we pretend the tests were equivalent.
What a deliverability claim can prove
A percentage can be useful evidence when its boundaries are preserved.
What it does not prove
One percentage should not carry more weight than the methodology can support.
What affects real-world results
The ESP is only one part of the sending system.
Deliverability should be evaluated as an interaction between platform infrastructure and the sender's own practices and audience.
Controlled tests
Small tests are useful — but only for the questions they can answer.
StackEvidence uses controlled accounts and test contacts to verify product behavior where practical. That evidence has real value, but it should not be stretched into claims the test was never designed to prove.
StackEvidence evidence boundary:
our controlled email-platform tests verify specific workflows and
observed results. We do not convert a few test sends into a claim that
one platform has universally better deliverability.
Evaluate a deliverability claim
Use five questions before allowing the number to influence a buying decision.
1
Define
What exactly does the percentage measure?
2
Source
Who ran, funded, or published the test?
3
Method
How were audience, message, timing, and placement measured?
4
Match
Are competing numbers actually comparable?
5
Scope
How far beyond this test can the conclusion reasonably extend?
Should deliverability decide the winner?
Deliverability matters, but weak comparative evidence should not dominate the entire product decision.
If a platform cannot support responsible production sending, that is obviously important. But once products meet the basic requirements, the final choice still needs to account for the wider operating system.
Better research language
Strong conclusions describe the evidence without pretending uncertainty disappeared.
“Platform A has the best deliverability.”
This implies a broad winner even when the evidence may come from one unmatched test or vendor-controlled statistic.
“Platform A scored higher in this specific test.”
This keeps the conclusion attached to the evidence that actually produced it.
Describe method, result, and limitation.
State what was tested, what happened, and why the result should or should not be generalized beyond that experiment.
What buyers should do
Use deliverability evidence as one input, then test your own setup responsibly.
1
Shortlist
Choose products that satisfy the actual business requirements.
2
Configure
Set up sender identity and required authentication correctly.
3
Test
Verify messages, links, workflows, and controlled inbox behavior.
4
Observe
Monitor the real production program rather than relying only on marketing benchmarks.
5
Improve
Correct list, configuration, content, or workflow problems as evidence appears.
Do not choose an ESP solely from a leaderboard.
A benchmark can inform the decision, but the system still has to work
for your audience, content, policy requirements, budget, integrations,
and operating process.
StackEvidence standard
Deliverability claims receive the same evidence treatment as pricing and policy claims.
We separate what was directly observed, what a vendor reports, what a third party tested, and what can reasonably be concluded from that evidence. We do not create a universal platform ranking from unmatched percentages.
Define the metric
Delivery, inbox placement, engagement, and conversion are different outcomes.
Identify the source
Vendor-controlled and independent evidence can both be useful, but provenance matters.
Inspect the method
Audience, content, authentication, timing, mailbox mix, and sample size affect interpretation.
Match before ranking
Percentages from different methodologies should not be treated as one shared experiment.
Preserve limitations
Controlled testing can verify observed behavior without promising future production results.
Commercial transparency
Affiliate economics should never turn weak evidence into a deliverability winner.
This page contains no affiliate links. If StackEvidence later has commercial relationships with email software providers, commission differences will not change the evidentiary standard used for deliverability claims or comparisons.
Our position:
when the available evidence cannot support a matched ranking, the
correct conclusion is that a universal deliverability winner has not
been established.