The dashboard chose the comparison
Most reporting starts with a comparison nobody picked. The CRM dashboard and the ad platform's overview page both open on this month beside last month, with a percentage and an arrow, because that is the comparison a tool can compute on the day a business starts using it. A year-earlier figure needs a year of history in the same system, and a rolling average needs somebody to decide how many months it rolls over. The month before is always there, so the month before is what gets shown.
We have watched that arrow travel a long way. It goes from the dashboard into the Monday report and from the Monday report into the meeting where spend gets decided, and at no point does anyone go back to the setting that chose the comparison. The number is treated as a reading of the business when it's a reading of the tool's default. Where enquiries arrive evenly through the year that does little harm. A heating firm or a garden centre is a different case, because for either of them the month before sat on a different part of the year.
The reason it hides more than it shows is arithmetic. A month-over-month figure is the current month minus the month before, and whatever the business did in the current month is in that number, and so is whatever the calendar did between the two months. The arrow doesn't separate the two. For a seasonal business the calendar's share is usually the larger one, and the dashboard reports it under the heading of performance.
A quiet month is not a decline
A business whose enquiries fall away every January is the plain case. December closes and the phones go quiet, and the first report of the new year shows enquiries down by whatever the drop happens to be, in red, with an arrow beside it. Nothing about the business changed between the two months. The people, the offer, the website and the ad spend are the same as they were in the second week of December, and the comparison reports a decline all the same.
The reaction to that arrow is where the cost sits. A red January gets read in a meeting as a problem, and the fixes on offer are the ones that can start that week, which means more budget on the ads or a promotion nobody had planned. The spend goes up in the quietest month of the year, and in March the enquiries come back because March is when they come back, and the extra spend is credited with the recovery. Both readings were wrong. There was no decline to fix and no recovery to credit, and the business now holds a belief about what January spend does that it will act on again next year.
In the busy season the same misread runs the other way. A campaign that goes live in the first week of the busy period gets the whole of the seasonal rise credited to it, because the month-over-month figure can't tell the campaign from the weather. The team that ran it is sure it worked and the report agrees with them, so the following year the same campaign runs again at the same time on the strength of a number that measured the calendar. Nobody in that room is being careless. They are reading the number the tool gave them.
An automated report makes this worse until the comparison inside it is fixed. A report that assembles itself on the first Monday of the month, with the comparison it inherited from the dashboard, sends the misreading on time every month with nobody in the chain who might have paused over it. That is why the comparison inside a report is settled during the build and written into the note underneath it, and never left to whatever the tool showed first.
The same month a year earlier
The comparison we put on the report first is the same month a year earlier. January against last January sits on the same part of the year, so the calendar's share of the difference is mostly gone and what is left is closer to what the business did. Enquiries down against last January are worth a meeting, which enquiries down against December never were.
Our reasoning is the one behind the baseline we agree before any build. A comparison is only as good as the two things it puts side by side, and the second thing has to have come out of the same tool under the same definition. It also has to come from the same part of the year, which is where the month before fails in any business with a season. The year-earlier month passes, as long as the definition of an enquiry hasn't changed in between, and that is the condition we check before the figure goes anywhere.
The year-earlier figure also shows the size of the swing itself. The month before never could, because it sat inside the swing. Once the report puts the same month for two years side by side, the team can see that the January drop is the January drop, at about the size it was last time, and stop treating it as news. Knowing the swing lets them plan for the quiet month, which is a better use of January than spending against it.
A rolling baseline moves slowly on purpose
The second comparison is a rolling one, and we use it where the year-earlier month was itself unusual or where one month's count is too small to read on its own. The simplest version is the total of the last twelve months, recalculated every month, so that each reading contains one of every month and the season cancels out. A number built that way barely moves from one reading to the next. That is the point. A real decline shows up as a slope over several months, and a quiet January doesn't show up at all.
Slowness is the trade. A rolling twelve months will not tell the team that something went wrong in the last three weeks, because three weeks is a small part of twelve months. It tells them whether the business is larger or smaller than it was, which is the question the management meeting is usually asking when it points at the monthly arrow. For the shorter question we use a shorter window, the last thirteen weeks against the same thirteen weeks a year earlier, and we write down which question each window is there to answer.
Which of the two goes first depends on the shape of the business's year. A trade business booked out through the summer reads better against the year-earlier month, because the swing is large and its shape repeats. Where the season is real but mild, one busy quarter and three ordinary ones, the swing is small enough that a single month's comparison is mostly noise, and the rolling figure is the one worth reading. We settle that in the strategy session with two years of the count on the table, and the choice goes into the baseline document.
What goes on the report
The report we build shows the current month's count with the same month a year earlier beside it. Where the counts are small the difference is written as a count, because a percentage of a small number makes every change look large. The rolling twelve-month total sits below as a single figure, with the previous reading next to it. The month-over-month arrow is either taken off the page or moved to the bottom with a label saying what it compares. Either way it has stopped being the first thing anyone sees.
Underneath the figures there is a note, and the note is the part of the report we spend longest on. It says which comparison the report uses, why that comparison was chosen for this business, what an enquiry meant when both figures were taken and the date that definition was last checked. It is written for the person who will read the report in eighteen months without us there, and for whoever joins after that and asks why the page doesn't look like the dashboard. Neither of them will have been in the room.
Ownership goes with it, the same way it does with every automation we hand over. The person named on the report is the one who knows that last March had a price change, or a week when the website form was down. They can say so before the year-earlier figure gets read as a clean one. The report carries the figure, and the memory of what made the figure odd stays with a person.
