If someone in your business spends Monday morning copying numbers out of your CRM, ads accounts and accounting software into a slide or a spreadsheet, that report should be automated. Each number should have one definition and one source system. The report should assemble itself on a schedule and land in front of the team already built. Your people then spend the meeting deciding what to do, not arguing about whose figure is right.
Why the manual weekly report keeps surviving
Nobody defends it. It survives because it is nobody's main job, so nobody prices it.
The hidden cost is not only the hours. It is that the report arrives late, so decisions about last week get made halfway through this week. It is that two people pull "leads" differently and the meeting turns into reconciliation. It is that when the person who builds it is on holiday, the report simply does not happen.
There is a second cost that shows up later. When numbers live only in a weekly spreadsheet, you cannot look backwards properly. You cannot ask what your show rate was in March without opening eleven files.
Example calculation (example figures, not client data)
| Item | Example figure |
|---|---|
| Operations manager building the report | 2 hours per week |
| Two team members pulling their own numbers | 45 minutes each per week |
| Time spent in the meeting reconciling figures | 20 minutes per week |
| Total | about 3.8 hours per week |
| Weeks per year | 46 |
| Hours per year | about 175 |
| Blended cost per hour (example) | £35 |
| Annual cost | about £6,125 |
These are illustrative figures to show the shape of the maths, not results from a client. Put your own numbers in. The point is that a task which feels like "an hour on Monday" is usually a four figure annual line that also delays every decision it feeds.
What an automated weekly report should pull from each tool
An automated report is a set of scheduled reads from systems that already hold the data. You are not creating new data. You are stopping people from re-typing it.
| Source system | What it should supply | Common trap |
|---|---|---|
| CRM | New enquiries, source, stage changes, deals won and lost, value | Counting a duplicate record as a new enquiry |
| Calendar or booking tool | Consultations booked, attended, cancelled, no-shows | Counting bookings made rather than bookings held |
| Phone or call system | Calls received, answered, missed, after-hours | Missed calls that were returned still counted as lost |
| Ad accounts | Spend, clicks, cost per enquiry by campaign | Platform-reported conversions double counting the CRM |
| Accounting or payments | Invoiced, collected, overdue | Mixing invoiced revenue with cash received |
| Delivery system or portal | Jobs started, jobs completed, overdue tasks | Work marked complete in one place but not another |
| Support inbox | Open tickets, first response time, complaints | Internal emails counted as customer contacts |
Pull from the system where the work actually happens. If bookings are made in the calendar, the calendar is the source for bookings, even if the CRM also has a field for it.
For which numbers deserve a place at all, we set out the nine numbers every service business dashboard needs. Do not automate thirty metrics because they are available. A weekly report that nobody reads costs more than one that does not exist.
Define each number once, in writing
This is the part people skip, and it is the part that decides whether the automated report is trusted.
For every number in the report, write down four things: the exact definition, the source system, the time window, and the person who owns it. Keep that sheet next to the report, not in someone's head.
A worked definition looks like this. "New enquiry: a contact record created in the CRM between Monday 00:00 and Sunday 23:59 with a source tag, excluding records flagged as duplicate or internal test. Source: CRM. Owner: sales lead." That level of precision feels excessive until the first time two people disagree.
Expect to find disagreements during this exercise. Finance counts revenue when invoiced. Sales counts it when the deal is marked won. Both are reasonable. You just need to pick one for the report and label the column accordingly.
Where your definitions are inconsistent because the CRM was never set up properly, fix that first. Automating a report on top of unreliable records produces confident nonsense. That is a CRM setup and migration job, not a reporting job.
How the automation should actually run
Keep the mechanics boring. A scheduled job reads each source, applies the agreed definitions, writes the result to one place, and sends the summary.
Store the weekly snapshot. If you only ever show "last 7 days" from live systems, you lose history every time a record is edited retrospectively. A stored weekly row lets you compare this March with last March.
Build in failure handling. If the ads account returns nothing, the report should say "Meta spend unavailable" rather than printing zero. A silent zero is worse than a visible gap because people act on it.
Watch your automation costs if you are wiring this with off-the-shelf tools. Zapier counts each successful action step as a task, and at the plan limit new runs are held unless you pay for extra tasks. A report that pulls twelve sources every week can quietly eat a plan.
Keep a person on the send where the numbers leave the building. On MRM Group we built operational reporting with human approval at sensitive financial-promotion and AI decision points. Internal reports can go out automatically. Anything a client or investor sees should be checked by a named person first.
What to review each week once the report builds itself
The meeting changes. You are no longer presenting numbers, you are responding to them. Work through this list in order.
- Check the data integrity line first. Any source that failed, any count that looks impossible, flag it and move on.
- Compare this week against the four-week average, not against last week alone. One week is noise.
- Look at the top of the funnel: enquiries by source, and cost per enquiry where there is spend.
- Look at the conversion steps: enquiry to booked, booked to held, held to won. Find the step that moved.
- Look at delivery: jobs completed, overdue tasks, anything past its promised date.
- Look at cash: invoiced, collected, overdue over 30 days.
- Agree a maximum of three actions, each with a named owner and a date. Write them into the next report so they are reviewed.
If a number in the report never changes a decision, remove it next week. Reports grow by accretion and nobody prunes them.
When not to automate the weekly report
Do not build this if your processes are still changing weekly. Definitions have to be stable before you encode them.
Do not build it if the data is in three people's inboxes rather than in systems. Fix where the work is recorded first. What to automate first in a service business covers that sequencing.
And if you only need five numbers from one CRM, use the CRM's own reporting. Custom dashboards and reporting earn their cost when data sits in several systems, when definitions differ between teams, or when you need history you can trust a year from now.
Your next step this week
Open your last weekly report and write down, for each number, the exact definition, the source system and the owner. Do it before you automate anything. Most teams find at least two numbers nobody can define, and one that is pulled from the wrong place.
Bring that sheet to a conversation with us and we can tell you whether this is a configuration job in your existing tools or a build.
