▤Reading figures correctly

Four groups, and all four can be verified:

• Holdings. How many contacts, how they are distributed across categories, how many with an email address.
• Sending. What went out, what the receiving server accepted, what came back.
• Bounces. Hard (the address does not exist) and soft (mailbox full, temporary).
• Replies. What arrived in the inbox and which contact it has been assigned to.

What you do not have: open rates and click rates. That is intentional — there is a separate answer on this.

The most useful figure of all is not in the system: how many orders came out of it. You have to assign those yourself, and that is exactly what notes on the contact are for.

For two reasons, and both are good ones:

Data protection. Open tracking works via an invisible image that is fetched from your server when the email is displayed. This tells you when someone opened the email and roughly where they were — without them knowing. That is processing which requires consent.

The figure is no longer any good. Email providers now preload images on their own servers. An “open” then only means that a data centre fetched the image — not that a person saw the email.

What counts instead: replies. A reply is an unambiguous signal; an open is an assumption.

If you still need a comparative figure, use the delivery rate and the reply rate — both are hard figures.

How many of your emails were accepted by the receiving server. That is not the same as “landed in the inbox” — an accepted email can still end up in the spam folder.

As a guide:

• Over 97% accepted: normal.
• 90 to 97%: the list is outdated; there are dead addresses in it.
• Under 90%: a problem. Either the list is bad, or your sender address has a bad reputation.

What to do if the rate falls:

1. Look at the bounces — what kind of errors exactly?
2. Remove hard bounces. They do harm with every further sending run.
3. Check deliverability — SPF, DKIM, DMARC.

A falling delivery rate is always an early warning sign. It falls before you notice that nobody replies any more.

• Hard: the address does not exist. The receiving server says a clear no. Causes: typos, the person has left the business, the domain no longer exists.
• Soft: a temporary obstacle. Mailbox full, server unreachable, message too large.

What to do with them:

• Remove hard bounces immediately. Every further attempt damages your sender reputation. Email providers treat repeated deliveries to non-existent addresses as a sign of a purchased list.
• Keep an eye on soft ones. A single one is insignificant. The same address soft-bouncing three times in a row: remove it.

A high proportion of hard bounces on the first sending to a list tells you something about the list — immediately, and without you having to go through it.

That depends heavily on how cold the contact is:

• Cold first approach: one to five per cent. Anything above that is good.
• Warm contacts (have dealt with you before): ten to thirty per cent.
• Existing customers: considerably more.

More important than comparing with other people’s figures is comparing with yourself: how did it go last time, with the same kind of recipients?

And count refusals too. Ten refusals are a better result than zero replies: they show that the email was read and understood. Anyone who gets zero replies has either a delivery problem or a text that nobody reads to the end.

For small campaigns — under fifty recipients — do not calculate in percentages. Count.

By orders, not by emails. Everything in between is an interim result.

The chain you should keep an eye on:

1. Sent →
2. accepted →
3. answered →
4. conversation →
5. quote →
6. order.

Each step has its own cause when it stalls:

• 1 → 2 poor: list or deliverability.
• 2 → 3 poor: subject, text, audience.
• 3 → 4 poor: the closing question was too big, or you get back to people too late.
• 4 → 5 and 5 → 6: that is your sales work, not the system.

Aurora Hive Sales measures the first three steps itself. You record the rest as a note on the contact, and that is exactly why the note after the conversation is the most important habit.

With three rules, otherwise you are comparing chance:

1. Change one thing. Subject or text or audience. If you change all three, you know nothing afterwards.
2. Enough recipients per version. Under a hundred, every difference is noise.
3. The same conditions. Not one campaign in July and the other in October; not one to existing customers and the other to strangers.

What you compare: the reply rate, not the delivery rate — that depends on the list, not on the text.

And record what you changed. A comparison whose basis nobody remembers after four weeks is wasted work. Two lines in the campaign description are enough.

Longer than it feels. The usual mistakes:

• Too short. A campaign evaluated on Monday that runs until Friday.
• Measured too early. Replies come over weeks, not days — with business contacts, often only after the holidays.
• Unsuitable comparison periods. Comparing December with October says more about December than about your work.

As a guide:

• Delivery and bounces: meaningful after 24 hours.
• Replies: after two to three weeks.
• Orders: weeks to months, depending on the business.

And mind the period filter on the overview: it affects almost everything below it. A figure that looks wrong is usually a figure from the wrong period.

More than most people look for. Three questions that are worth asking:

• How many contacts have an email address? That is the number you can actually work with. A base of 4,000 with 900 addresses is a base of 900.
• How are the categories distributed? If 80% are in “Other”, the categories are not maintained, and then no audience selection is any good.
• How old is the last contact? Addresses without contact for three years are very likely dead.

That is the evaluation that comes before every campaign. Anyone who does it once a year spares themselves the surprise of a campaign finding only 200 recipients instead of 2,000.

From top to bottom, and with an eye on the period:

1. The three dots — system, licence, mailbox. If one is not green, everything below comes with a caveat.
2. The period. It affects almost everything below. Not the personal sections (“today”) and not status information.
3. What is due today — your follow-ups, tasks, appointments. Personal, not the team’s.
4. Delivery — what went out and what came back.
5. Usage — what AI and search services have cost this month.

Next to every figure there is an ⓘ that says what exactly it counts. Use it the first time: two figures with similar names rarely count the same thing.

In this order:

1. Period. By far the most common reason. Check which one is set.
2. Filter. A list with a filter set shows a different number from the total — both are correct; they just count different things.
3. What exactly is counted? “Written to” is not “delivered” is not “answered”. The ⓘ next to the figure tells you.
4. Visibility. What your role may not see is not counted either.
5. Time lag. Some figures are produced in the background and are a few minutes old.

If something is still not right after that: report it — stating which figure, which period, which value you expected. That can be traced; “the figures are wrong” cannot.

It shows what the chargeable services have used this month — broken down by type: AI evaluation, search service, web search.

What you should do with it:

• Set a budget. Without a limit, usage carries on even if someone accidentally starts twenty searches.
• Check outliers. A sudden rise always has a cause — usually a run that someone started several times.
• Put it in proportion. What did the usage achieve? Ten euros for an order is cheap; a hundred euros for zero results is expensive.

And the basic principle: these services cost per request. Anyone who wants to keep costs low searches more precisely, not less often — fewer terms, narrower location, smaller quantity.

For what you show someone else: a client, the management, a board.

The difference from the overview: the overview is your worksheet and changes constantly. A report is a recorded state for a period, as a PDF, with your sender details.

What this saves in everyday work: the monthly half hour in which someone copies figures from the system into a spreadsheet.

What you should watch out for:

• Check the sender details before sending for the first time.
• Choose the period deliberately and keep it. Reports that sometimes cover a month and sometimes six weeks are not comparable.
• Look at it first. A report goes to a customer — someone should have read it once.

Anything that rates people instead of processes:

• Who writes how many notes. Leads to notes being written in order to be counted — and those are worthless.
• Who replies how quickly. Leads to quick, poor replies.
• Calls per day. The classic key figure that leads to many short and few good conversations.

The reason is not restraint but effectiveness: every key figure by which someone is measured gets optimised — even when that harms the actual goal.

And legally: an evaluation by person is subject to co-determination where there is a works council. Anyone who uses the access log to monitor performance needs a works agreement.