Sales activity tracking and conversation intelligence

Sales activity tracking counts what people and agents do: messages sent, replies, meetings booked and held. Conversation intelligence reads what was said, in calls and messages, to find patterns. Track few numbers and read many conversations. On our own account the reply rate held near 40% all year while meetings per person messaged drifted below 1%, and only reading the threads explained why.

The two ideas are often sold together, and they answer different questions. Tracking tells you how much happened. Intelligence tells you what happened. A team that has only the first knows its numbers and not why they move.

What is sales activity tracking?

A count of actions and outcomes, usually per person, per list and per week. For outbound, a short list is enough:

• Messages sent, by channel.
• Replies, and the share that are more than a pleasantry.
• Positive replies, meaning someone raised a hand.
• Meetings booked and meetings held.
• Time to first reply.

Each number answers one question. Sent: did the work happen? Replies: did the message land? Positive: did the audience fit? Held: did the person show? Time: did we answer while the person was still there?

Which numbers mislead?

Three, and our own case studies show each.

Raw reply rate. In the ShortVibe study 42% replied (395 of 935), and about half were courtesies such as thanks for connecting. In the wealth advisor network study, 44.8% replied, but only 33.7% replied with more than a pleasantry, and the agent turned 1,035 of 1,751 thanks for connecting replies into real conversations. The headline reply rate hides how much of it is conversation.

A steady reply rate. On our own account replies stayed between 35% and 45% in each of twelve monthly cohorts, while meetings per person messaged peaked at 2.4% in November 2025 and drifted below 1% from June 2026. A stable reply rate hid a changing audience.

Meetings booked. 351 showcase meetings were booked through the agents and 33 are confirmed held. Count held, not booked.

What is conversation intelligence?

Software that records or reads sales conversations, transcribes them, and analyses what was said: topics, objections, questions, talk time. In the world of calls, it usually means a recorder and a transcript, with a summary and a search across all of them. In the world of messages it can mean reading every thread for patterns.

Sellinger is not a note taker, and it does not record meetings. It keeps the full conversation history across channels for every lead, in its CRM, and a Fireflies integration is live, so a call recorded in a note taker can sit beside the conversations the agents held. If you are choosing a note taker, the comparison of AI note takers with CRM sync dates and sources each cell and names no winner.

What do you learn by reading?

More than a dashboard shows. Examples from our own case studies, all found by reading threads:

• Capability questions. In the Lamothermic study, 26 replied threads asked what the foundry could cast. The agent could answer precisely only after the president wrote down what the company does and does not pour. No dashboard showed this.
• The follow up that brings a new angle. In the Uberwood Agency study, follow ups added 43% more replies when they brought a new angle, not a plain bump. That came from reading which follow ups worked.
• Where the link goes. On our own account the booking link typically went out after three prospect messages, and 56% of bookings confirmed in the chat needed at least one more nudge after it.
• What the second reply says. In the Hotel Pirin study, the last note in a sequence produced 4 inquiries, one of which, forwarded inside a company, became a confirmed booking.

A pattern like these is invisible in counts. It shows up when someone reads fifty threads with one question in mind.

What do you do when a number moves?

A number is only useful if it points to an action. This is a rule of thumb for outbound, not a measured result.

The order matters: look at a few real threads first, then change one thing. If the cause is the audience, rewriting the copy will not fix it.

How should a small team run it?

Keep it light.

1. One weekly table. Messages, replies, positive replies, meetings booked and held, time to first reply, per list. The same columns every week. See reading your analytics.
2. Twenty threads a week. Ten that went well and ten that did not. Ask one question per week, such as: what did people object to?
3. A note of what you changed. One change a week, with the date. When a number moves, you can see what you did.
4. A monthly pass over the held meetings. Which lists produced them, and what the call revealed that the thread did not.

Start with the threads the agent stopped on. They show where its rules are missing.

What about privacy?

Recording and analysing conversations are processing of personal data. In calls, many jurisdictions require telling the other party. In messages, the people wrote to you and your agent, but data protection law still applies to how you store and use it. Keep what you need, and let people have their data erased. Sellinger has an erasure function that anonymizes matching leads and adds the identifiers to the suppression list. See the help page on data requests.

What should you not conclude?

Do not conclude causes from a single week or a single thread. A number moves for many reasons, and a vivid thread is not a trend. Where we cite a pattern in this post, it comes from a published case study with its denominator, and each is one account's data.

What would a one page weekly report contain?

A small team can run on one page. At the top, five numbers for the week with last week beside them: messages sent, replies, positive replies, meetings booked, meetings held. Below, the same numbers by list, so you see which source worked. Then three short quotes from threads, one that went well, one that went badly and one surprise. At the bottom, the one change you will make this week and why.

The quotes matter more than they look. A number says that replies fell. A quote says that three people asked whether the message was written by a bot, and that is an action. This is the habit behind most findings in our case studies, and it costs twenty minutes. Our weekly client report template shows the same idea for agencies reporting to clients.

A checklist

• [ ] I track sent, replies, positive replies, booked, held and time to first reply.
• [ ] I look at replies with more than a pleasantry, not only the raw rate.
• [ ] I count meetings held per person messaged.
• [ ] I read twenty threads a week with one question.
• [ ] I log one change a week.
• [ ] I know how I handle recorded calls and data requests.