Automating reports: from manual work to a real-time dashboard

In most of the companies we talk to, "the sales report" exists. The problem is not its absence but the way it gets produced: from memory, from three files, and from a full day of somebody's work. This article is about how you get from that report to one that generates itself — step by step, with what has to be decided beforehand and what automation does not solve.

How the manual report actually gets made

The typical scenario looks like this. On Friday afternoon the sales manager writes in the team chat: "Send me this week's numbers." Each salesperson recalls what they did — how many customers they called, how many quotes they sent — and types a message. The manager copies it into their Excel. Accounting has another Excel, with the invoices issued, which matches neither the period nor the definition of a "sale". Marketing has a third one, with the Facebook leads, which says nothing about what happened to them after they were passed on.

On Monday morning somebody spends several hours reconciling the three files. That day shows up in no report, but it happens every month. The result is a table that is broadly right, impossible to verify in detail, and ready a few days after the moment it would have been useful.

The important observation: the report is not hard to produce because the formulas are complicated. It is hard because the data does not exist in one place, in the same format, with the same definitions.

The four layers of an automated report

A report that makes itself has four parts. If the first one is missing, the other three look good and lie.

1. Data enters by itself. Every website enquiry, every Meta lead, every call through the PBX, every Telegram message becomes a record in the CRM, with source and time, without anyone copying it. Every stage change is logged automatically. This is where most of the implementation effort goes — and most of the value.
2. Aggregation rules are defined once. What "lead" means (a new enquiry? a qualified contact?), what "quote sent" means, what "sale" means (signed contract? deposit paid? invoice issued?). Which period, which groupings: by source, by employee, by product. We agree on them together, write them down and stop changing them every month — otherwise you cannot compare two months.
3. Generation is scheduled. Daily at 8:30 the list of unanswered leads; Monday at 9:00 the previous week; on the 1st of the month, the previous month. Or on demand, from a button.
4. Delivery reaches the right person. The manager gets the summary in Telegram or by email; each employee sees their own report; the dashboard in Web Admin shows everything in real time, with filters. Not one file sent "to everyone" that nobody opens.

The first weekly sales report, step by step

Do not start with the complete dashboard. Start with a single report, the Monday-morning one, and build it like this:

Step 1. Define three words. Put sales, marketing and accounting at the same table and agree in writing: "lead" = any new enquiry with a valid contact, regardless of source; "quote" = a price proposal sent to the customer, in any form; "sale" = signed contract or deposit received — pick one of the two. It sounds trivial; it is the step most often missing.

Step 2. Pick five indicators. Not fifteen. For example: new leads by source; leads contacted within the first hour; quotes sent; sales closed and their value; deals lost and the reason. If the manager would not change a single decision based on an indicator, it does not go into the report.

Step 3. Connect the two biggest sources first. If most enquiries come from the website and Facebook, those flow into the CRM automatically from the first week. Telephony, Telegram and secondary forms come afterwards. A report that correctly covers the bulk of your leads is more useful than a "complete" one that arrives two months late.

Step 4. Make the loss reason mandatory. When a deal moves to "Lost", the CRM asks for a reason from a short list: price, response time, wrong product, competition, no response. Without this field, the most valuable line in the report stays empty.

Step 5. Schedule it for Monday at 9:00. The report is generated from last week's data and reaches the manager in Telegram, with a link to the dashboard for the details.

Step 6. Review after four weeks. Which indicator changed no decision? Remove it. Which question came up in the meeting and was not in the report? Add it. Then leave it fixed for a quarter.

What goes into the manager's Monday message

An example, with fictional numbers, to make the format clear — not any client's results:

Week of 10–16 August
New leads: 37 (website 16, Meta 14, phone 7)
Contacted within the first hour: 28 of 37
Quotes sent: 9 · Sales: 3 (total value 84,000 lei)
Lost: 4 — price (2), no response (2)
Attention: 9 leads contacted after the first hour; 6 with the same agent
[Open dashboard] [See delayed leads]

Three observations about this format. It is short — it reads in thirty seconds on a phone. It has an "attention" line that points to where the problem is, not just the number. And it has buttons to the details, for whoever wants to dig. The Monday meeting starts from it.

Common mistakes

• The indicators change every month. The director asks for a new column, another one disappears, the definition of "sale" drifts. The result: reports that cannot be compared. The fix: written rules, reviewed quarterly, not at every meeting.
• Reports nobody reads. An eight-page PDF, sent to twelve people, Monday at 9:00. If you cannot say who makes which decision based on the report, do not send it. A summary for the manager, an individual report for each employee, a dashboard for whoever wants detail.
• Nobody owns the report. Someone has to answer "why does this number look like that?". Usually it is the head of sales or the person who used to keep the Excel — they no longer compile, but they check and explain.
• The display gets automated, not the collection. A handsome dashboard on top of data typed in by hand, on Friday evening, from memory. It looks good and is exactly as approximate as before.

When AI summaries help — and where they stop

On top of the numbers you can add a plain-language summary: "This week fewer leads came in from Meta than the average of the last four weeks; response time improved for all agents except one; two of the four losses share the same reason — price — on the same product." For a manager reading on a phone between two meetings, that is useful: it says where to look.

The limit is simple and must be respected strictly: the AI explains the numbers in the CRM, it does not produce them. It does not invent an indicator that is not in the data, does not "estimate" what was never recorded, and every statement must be verifiable in the dashboard within two clicks. If a summary says something you cannot find in the numbers, it is configured wrong.

What report automation does not solve

• Data that never enters the system. If calls are not connected to the CRM, the report cannot see them. If a salesperson keeps customers in a notebook, they do not exist. That is why we start with the integrations, not the charts.
• Missing definitions. An automated report on undefined indicators is accurate and useless.
• Accounting. Operational reports — leads, conversions, response time, workload — do not replace the books. We export data to accounting; we do not replace it.

More on which reports can be automated and where they are delivered: report automation · the rest of the CRM automations: process automation · the system that collects the data: CRM Moldova · reports and alerts straight in Telegram: Telegram CRM.