Measure emails to clients by replies, not opens. Reply rate (the share of delivered recipients who wrote back), time to first reply and replies per template tell you whether an email worked, while open rates have been distorted by Apple Mail Privacy Protection, Gmail’s image proxy and security scanners. Use opens and clicks only to decide when to follow up.
This is especially true for one-to-one business email: quotes, introductions, project updates, payment reminders. Newsletter marketers can live with fuzzy open rates across thousands of recipients. A sales or account manager writing to forty clients cannot.
Why opens stopped being a useful score
An open is recorded when a tracking image is requested. Since 2021, Apple Mail with Mail Privacy Protection can request that image through Apple’s servers whether or not the person opens the message, and Gmail serves images through Google’s proxy. Company security filters add opens and clicks of their own. The result: opens are inflated for some recipients, missing for others and timed by machines rather than people. We explain the mechanics in why Apple Mail and Gmail inflate open rates.
A reply has none of these problems. A proxy cannot write back. That is why the rest of this article is about replies.
Three metrics that work
1. Reply rate
Reply rate = recipients who replied ÷ recipients whose email was delivered
Three details make the difference between a useful number and a misleading one:
- Count people, not messages. If one client sends three replies, that is one replying recipient.
- Exclude automatic replies. Out-of-office and vacation messages are not answers. Well-behaved mail programs mark them with the Auto-Submitted header described in RFC 3834; others use older headers or a subject prefix such as “Out of office”.
- Use delivered emails as the base. Bounced messages never reached anyone. Counting them lowers your reply rate for reasons that have nothing to do with the content.
2. Time to first reply
Measure the time from when your email was accepted for delivery to the recipient’s first reply. Read it as a median where you can. An illustration: if four clients reply after 2, 3, 5 and 6 hours and a fifth after 300 hours, the median is 5 hours, while the average is over 63 hours. The average is dominated by one slow case and describes nobody.
Time to first reply is useful for two decisions: when a follow-up is reasonable, and whether something changed. If your median doubles over a quarter, look at what changed in your emails or your audience.
3. Replies per template
If you send similar emails often, such as a first message after a demo or a quote cover note, compare how many replies each version earns. Keep it fair:
- Send each version to a similar audience in the same period.
- Wait the same number of days before counting.
- Be careful with small samples: with ten emails per version, one extra reply moves the rate by ten percentage points.
| Template (illustrative example) | Delivered | Replied | Reply rate | Median time to first reply |
|---|---|---|---|---|
| Quote cover note A (long) | 40 | 9 | 22.5% | 26 h |
| Quote cover note B (short, one question) | 40 | 15 | 37.5% | 9 h |
In this invented example, version B looks better on both measures. Before switching, check that both went to comparable clients, and keep testing: forty emails per version is still a small sample.
What about opens and clicks?
They are still useful, just not as a score. Treat them as timing hints:
- A click with a human pattern (well after delivery, in a visible browser, on a home or mobile network) suggests interest. It can justify a follow-up a day earlier.
- An open at most suggests that the email arrived and may have been seen. It should never be quoted to the client (“I saw you opened my email”) and never used to judge the email.
- Nothing at all does not mean “not read”. Images may be blocked.
A simple monthly routine
You do not need a data team to measure this well. Once a month, for the emails that matter to you (quotes, first contacts, renewals), write down four numbers:
- Delivered emails in the month, after removing bounces.
- Recipients who replied, excluding automatic replies.
- Median time to first reply, from a sample of real conversations if your tool does not calculate it.
- Recipients with a human-pattern click, if you track links, as a secondary signal.
Then ask three questions. Did reply rate move compared with last month? Did the time to first reply change? Is one type of email clearly weaker than the others? Look at the actual emails behind any change before drawing conclusions: a holiday period, a new client segment or a price change can explain more than wording does.
Keep the method identical from month to month. A reply rate calculated one way in March and another way in April cannot be compared, however precise each figure looks.
Following up without guessing
A simple routine that works in most cultures:
- When you send, decide how many working days you will wait. For most business emails, two to five is fair; after a quote, about a week is common; before a deadline, a day or two earlier.
- If no reply has been recorded by then, reply in the same thread with one short message that restates the single thing you need.
- Make answering easy: a yes/no question or two possible call times.
- Stop after two or three polite follow-ups. The last one can say kindly that you will not write again about this and leave the door open.
Measuring replies in HawkOpen
HawkOpen, the email and proposal tracking app made by TriStiX S.L., treats a reply as the strongest signal it records. What it does today:
- Reply detection on the Pro and Business plans. HawkOpen checks each connected mailbox (IMAP/SMTP, or Gmail and Outlook connected by signing in) about once a minute and records a reply when it comes from the address you wrote to and belongs to the same conversation. It reads only the sender, subject and headers of unread messages, never the message text.
- Automatic replies and bounces are kept apart. Delivery failures are recorded as bounces. Out-of-office and similar messages are recognised from headers such as Auto-Submitted, X-Autoreply and X-Autorespond, a bulk or list Precedence, an empty return address, or a common out-of-office subject prefix, and they are not counted as replies.
- Reminders on Pro and Business tell you when the chosen number of days has passed without a reply, and close on their own when a reply arrives.
- Safe Follow-up on Business drafts a short follow-up at least 24 hours after the first email, in a professional, concise or warm tone. You edit and send it yourself; it never sends anything on its own and makes no draft if the person already replied, unsubscribed or bounced.
- Reports on Business, every Monday for the past week and on the first of each month for the past month, show replies and the average time to reply next to observed opens and clicks.
- Experiments on Business let you define variants and a measure, such as replies, and estimate the difference with its uncertainty.
One limit to be clear about: HawkOpen’s reports do not group replies by template. To compare templates, use experiments or keep a simple label in your own records. HawkOpen is currently in a controlled launch; the replies FAQ explains the details and limits of reply detection.