Email open rates look impossibly high
A 68% open rate is usually true and useless at the same time: mail apps and company security software open messages before a person does, and every one of those counts. The Email tab is built around that, which is why it judges sends on clicks and marks the open rate as a rough guide on the tile itself. If your sends are missing from the tab altogether rather than reading too high, read MarqIQ says my tool cannot provide this instead.
Before you start
Open Analysis, find the tool that sends your email, and click Email under it. That opens the tool’s own page with the Analysis tab showing, and the Email analysis selected. The figures cover the last 90 days until you pick a range under Created at the top of the page.
What the tab counts
Totals are the period’s one-off campaigns. On HubSpot and Mailchimp, automated and triggered emails are listed separately under Always-on emails, because those two report their results since the email was switched on rather than for your period. Marketo is read for the period itself, so its always-on numbers do cover it, and the note above the table says which of the two you are reading. Marketing Cloud is the exception: MarqIQ marks every send it reads there as a campaign, so that table never appears and your automated sends sit inside the totals.
Read the two tiles together
The In short row is where the answer starts. Opened carries the words “inflated by mail apps; a rough guide only” under it, and Clicked carries “of people who received it; the measure that counts”. That is not a disclaimer, it is the method: MarqIQ judges one send against another on clicks whenever there are at least 30 clicks across the period, and falls back to opens only when there are not.
- Every rate counts people, not events. One person opening an email five times is one open.
- Open, click and unsubscribe rates are out of emails delivered. The bounce rate is out of emails sent.
- A single send is called above or below average only when at least 100 people received it, and only when the gap is beyond what chance alone would produce.
- The comparison is against your own average for the period, never an industry figure.
The rate you see is already lower than a raw export
On HubSpot and Marketo the open count MarqIQ shows leaves out the opens the tool itself flagged as machine activity. So the tab’s open rate is normally below the one you get by exporting every open event and counting the rows. Compare like with like before you decide which is wrong.
Find out how much of it was a machine
Where your tool flags the activity, MarqIQ counts it and shows it. This is the only figure on the page that tells you the size of the problem rather than working around it.
Scroll to Clicks that were never a person on the Email tab.
The card is absent when your tool flags nothing. Which tools flag what is in the table further down.
Read People who clicked, the first of the three tiles.
This is the number every click rate on the page counts, after the flagged activity is taken out. It is the top half of the rate; the bottom half is the people the email was delivered to. Quote both.
Read the middle tile.
On a tool that reports events it reads Clicks by security software as “N of M recorded clicks”. Otherwise it reads Addresses only a scanner clicked, with the share of addresses with a click beside it. Both are left out of the rates.
Read Opens by security software only, the third tile.
Addresses where no person opened at all — the only open came from software. These are already left out of the open rate above.
Check the card’s own sentence for the coverage.
On HubSpot it counts only campaigns sent in the last 120 days, because that is how far back MarqIQ reads the event stream. A campaign older than that shows nothing here even though it has the same problem.
Take care
MarqIQ does not estimate how much of a typical open rate is machine activity, and no figure on the page claims to. It counts what your tool flagged, on the sends it could read, and says how many. Treat that count as a floor rather than a total — a mail app that loads an image without announcing itself is flagged by nobody.
The causes, most common first
Work down these. The first that matches is almost certainly yours.
The open rate is high across every send, including ones nobody replied to
Why: Mail apps with privacy protection fetch every image in every message as it arrives, whether or not the person opens it. An open is a request for a tracking image, so the fetch is recorded as an open. The share of your list on those apps is what sets your floor, and it does not move between sends.
Fix: Stop reporting on opens. Move your reporting to clicks and to what a click led to. The tab already does this: the verdict column Against your average is computed on clicks whenever the period holds at least 30 of them.
Clicks are high too, and the clicks arrive within seconds of the send
Why: Security software at the recipient’s company opens every link in an email before it reaches anyone, to check where it goes. Each of those is recorded as a click on the address it was sent to.
Fix: Open Clicks that were never a person. HubSpot and Marketo both flag this activity and MarqIQ leaves it out of every rate on the page. Then check anything that starts when somebody clicks — a lead score, a workflow, a sales alert — because those usually count the raw click and are still inflated by it.
One send has a far higher open rate than the rest
Why: A small delivered count. An open rate out of 60 delivered moves ten points when six more people open, and out of 6,000 it does not.
Fix: Check the Delivered column for that send. Under 100 delivered MarqIQ refuses to give it a verdict at all and marks it Too few recipients to judge, which is the signal to stop reading it as a result.
MarqIQ’s open rate is lower than the one on your tool’s own dashboard
Why: The two are counting different things. HubSpot leaves the activity it flags out of the ratios it reports, and MarqIQ takes HubSpot’s own ratio where it gives one; a raw event export includes everything.
Fix: Decide which question you are answering, then use one number for it. For “how many people looked”, use the tab’s figure and the machine count beside it. For “how many events were recorded”, use the export, and never put the two in the same slide.
Your always-on emails show enormous rates next to your campaigns
Why: On HubSpot and Mailchimp an automated email reports its results since it was switched on. A welcome email running for two years is being compared with a campaign sent last Tuesday.
Fix: Read Always-on emails as its own table and never add it to the campaign totals. On HubSpot, Mailchimp and Marketo MarqIQ keeps that table out of the totals, the day analysis and the trend for exactly this reason, and the note above the table states which behaviour your tool has.
Marketing Cloud has no always-on table at all
MarqIQ reads Marketing Cloud from the Send object and marks every row it finds there as a campaign, so the Always-on emails table never appears for that tool. An automated or triggered send whose send date falls inside your period is counted in the totals, in every rate, and in the day, month and trend analysis, next to your one-off campaigns. One whose send date falls outside the period is dropped from the tab rather than listed on its own. Quote a Marketing Cloud figure as all sends in the period, not as campaigns, even though the tab labels them campaigns. The How this was worked out list still carries a line about always-on emails being held back, and on Marketing Cloud there are none to hold back.
What to report instead
Every one of these is on the tab already, and every one states what it rests on.
| Instead of | Report | What it is measured on |
|---|---|---|
| Open rate | Click rate | People who clicked, out of people the email was delivered to, with machine-flagged activity removed |
| “Our best subject line” | The subject-line patterns card | Opens, deliberately — a subject line is all a reader sees before opening, and inflated opens affect both sides of the comparison alike |
| A single send’s open rate | The Against your average verdict | That send’s clicks against your own average for the period, on at least 100 delivered |
| “Engagement is up” | Unsubscribe and spam-complaint rates by month | Unsubscribes out of delivered; spam complaints out of delivered, where the tool reports them |
Why subject lines are still judged on opens
A subject line can only ever earn the open, so it is compared on opens on purpose. The inflation sits on both sides of the comparison in the same proportion, so the gap between them survives it. A pattern is only called out when at least three sends used it and three did not, each side reached at least 500 people, the typical send on one side beats the other by 15% or more, and two sends in three agree.
Which tools show the machine share, and which cannot
| Tool | What it flags | What you get |
|---|---|---|
| HubSpot | Marks an open or click it believes came from security software and leaves it out of the counters it reports | Clicks that were never a person, covering campaigns sent in the last 120 days. Where HubSpot recorded more events than MarqIQ reads in one go, the card says the count is a minimum |
| Marketo Engage | Marks the activity with its own “Is Bot Activity” attribute | The full card. Somebody who also opened or clicked for real counts as a person, not as a machine |
| Mailchimp | Reports no such flag | No card. The open and click rates are Mailchimp’s unique counts as they stand |
| Salesforce Marketing Cloud | Reports no such flag | No card. Rates come from the Send object’s unique opens and clicks |
The reading limits, before you quote a period
Each tool is read up to a stated point, and the tab prints the point it reached. HubSpot lists the 2,000 most recent emails. Marketo’s activity stream is read up to 45,000 email activities in one go and for at most 40 seconds; when it stops short the tab says so and tells you to choose a shorter period. Marketing Cloud is read for its most recent 25,000 sends. Mailchimp says how many of its campaign reports for the period were covered. Read that sentence before you quote any total on the tab.
Before this reaches a board pack
Take care
Quote the denominator with the figure. “A 68% open rate” cannot be checked by anyone; “68% of the 4,812 people it was delivered to, of whom 214 clicked” can. Every tile on this tab prints that second half already. Keep it, and say which of the four tools the figures came from, because only two of them can tell you how much of it was a machine.
If it did not work
The Email tab is greyed out and says “Not something this tool can do.”
Why: MarqIQ reads send results from HubSpot, Mailchimp, Marketo Engage and Salesforce Marketing Cloud. The tool you are looking at sends no marketing email, or has no reader yet.
Fix: Open Analysis and pick one of those four from the list of your connected tools. Three tools do send marketing email and still read grey: Pardot (MCAE), whose v5 API does not total a period, Adobe Campaign and Adobe Journey Optimizer. The sentence on the card says “does not send marketing email” for all of them, which is the wording of the tab rather than the truth about those three.
“There is not enough activity to judge sends yet”
Why: The period holds fewer than 30 clicks and fewer than 30 opens across every campaign, so no send can be compared with another.
Fix: Widen the period. The totals and the tables still fill; only the verdicts are held back.
Clicks that were never a person is missing
Why: Your tool flags nothing, or the flag could not be read. Mailchimp and Marketing Cloud never flag; on HubSpot the card also skips campaigns sent more than 120 days ago.
Fix: Check your tool against the table above. Where it is HubSpot and the card is still missing, click Read the latest results from HubSpot and check the sentence under the heading for a reason.
The figures have not changed since you fixed something in your tool
Why: Results are held for six hours per tool, and the computed view for five minutes on top of that.
Fix: Click Read the latest results from your tool above the tab. The line beside it says when your tool last reported these figures.
The card says security software clicked, but the number of people who clicked is unchanged
Why: Every flagged click came from an address where a person also clicked, so no person was removed from the count.
Fix: Nothing to fix in the rates. Check your lead scoring instead: anything adding points per click is still counting those events.
Next steps
Related articles
Was this article helpful?