Where Monday morning goes
The weekly report is one of the tasks a team names last, because no single week of it feels expensive. Somebody opens the first ad platform on Monday, sets the date range by hand, exports the spend, then opens the second platform and does it again. In the CRM they filter enquiries to the last seven days and count them by source. The counts go into a spreadsheet with a tab for every week since the report began, and quotes sent and won come from another CRM view or from whoever sends the quotes, asked over a message. A summary goes to a channel by late morning.
Nobody times it. Asked how long it takes, the person gives the figure for a week where nothing went wrong. When we reconstruct the last five working days with them, the morning includes an export that failed and a CRM count that didn't match the week before. What the people it goes to do with it is harder to tell, because a channel records that a report arrived and not whether a figure in it was read.
The morning is the smaller cost. The larger one is that the report depends on a person. When somebody covers a holiday they pick a different date range or leave out a campaign the regular person includes, and the figure moves for a reason that has nothing to do with the campaigns. A number produced a different way each week can't be compared with the week before, which was the whole reason for producing it.
The three figures and where each comes from
Our report carries three figures and nothing else to start with: enquiries by source, spend per enquiry, and quotes sent and won. We hold it there on purpose. Each is taken from the system that already holds it, with no person in between. Everything else a team asks for in the first meeting waits until those three have held up for a month.
Enquiries by source come from the CRM, and only from the CRM, which works where every enquiry became a record when it arrived with its source stamped on it. That is why the follow-up build comes first in most of our engagements and the report comes after it. The report counts the records in the last seven days and groups them by that field. It does no modelling of its own. Where the record carries no source, the report doesn't guess.
Spend comes from each ad platform's own reporting, pulled for the same seven days, and spend per enquiry is arithmetic done inside the report and not by a person. We set the division once. Which campaigns count and which day the week starts on are decided then and written into the job, because a person who does the sum by hand makes those choices again every week without noticing.
Quotes sent and quotes won come from the CRM's deal stages, which only works where the deal moves stage on its own as a quote is accepted or paid. Where the stage is still changed by hand on a Friday, the report reads the Friday and not the week. So the quote figure is the one we trust least in the first month. We say so on the page beside it until the stage changes have been automated too.
The build is mostly definitions
The build itself is small. A scheduled job runs early on Monday morning. It pulls the previous seven days of spend by campaign from each ad platform, and the enquiries and deals for the same window from the CRM. The result is written into a page that has looked the same every week since it started: one figure for each enquiry source, one spend per enquiry for each ad platform, quotes sent and quotes won, and last week's figure beside every one of them. Every cell on the page is one of the three figures broken out, and how many cells there are depends on how many sources and platforms the client runs. A short message with the page attached goes to the team's channel at the same minute every week.
Most of the time goes into the definitions, which are the ones written into the baseline document before the build. The week runs Monday to Sunday. An enquiry is a record created in the window, so a spam submission deleted on Tuesday still counts unless the rule for removing it is written into the job. Each definition gets a line of small type under the table, so the person reading a figure can see what it is a count of.
For the first two weeks the job runs beside the manual report and we put the two side by side. They rarely agree in week one. The gap is nearly always a definition, a campaign the person had been quietly leaving out or a date range that ran Sunday to Saturday, and once it has been named the manual version stops.
A new campaign is the usual way the job drifts. It launches midweek, the rule for which campaigns count never gets the line it needs, and the report runs short of that spend for as long as the campaign does. We agree, before the first run, which person in the client's team owns that rule. What they get is a handover note written for the tools they already open, with the schedule, each connection and the place the campaign list sits, so the line can be added without calling us.
Why the plain version is worth more
Teams usually want the new report to look better than the one they were building by hand, and we push back on that for the first month. A table with the same cells in the same places every Monday lets a reader go straight to the one that moved, because the eye goes to the cell and not to the layout. A chart that is redesigned whenever somebody has a better idea gets read for its design. The plain version is the one that gets read.
The prettier report has a person inside it. Somebody keeps the template current when a platform changes its export, and the same somebody has to be at a desk on Monday for it to go out. A weekly figure that arrives most weeks is worth much less than one that arrives every week, because the comparison with last week is the only thing the number is for.
There is a version of this argument that ends in a dashboard, and we resist it for the same reason. A dashboard is open all the time, which in practice means nobody has a reason to open it on any particular day. The scheduled report interrupts the week once, at a fixed time. That interruption is the point. When a client wants charts we add them in month two, after the numbers have sat in a table for a few weeks without anyone querying them.
