Year on year

Year on year compares a period against the same period twelve months earlier. September against last September, the fourth quarter against the fourth quarter before it. It is the plainest tool in reporting and the one most often left out of a monthly update, which is how a business ends up alarmed about a fall that happens every year.

In short

• It compares a period against the same period one year earlier.
• It removes seasonality, which month on month cannot do.
• It needs twelve months of comparable data, not just twelve months of data.
• Publish the underlying numbers beside the percentage or it misleads.

What it protects you from

Almost every business has a shape to its year, and that shape is usually larger than anything marketing does.

In the Philippines the shape is pronounced. Christmas spending begins in September and builds for four months. Holy Week empties some businesses and fills others. Payday cycles put a rhythm into the middle and the end of every month. Enrolment, harvest and the rainy season move whole categories on their own.

Against that, comparing one month to the month before it tells you almost nothing. A January that falls thirty per cent below December is not a January that went wrong. It is January. The only comparison that answers whether the business is improving is the one that holds the season still, and that is the same month one year earlier.

The condition people forget

Twelve months of data is not the same as twelve months of comparable data.

Between then and now the website may have been rebuilt, the tracking may have been reinstalled, the analytics platform may have changed, or a consent banner may have started suppressing measurement that used to be automatic. Any of these changes what the earlier number counted, so the comparison is now between two different definitions rather than between two periods.

None of this makes the comparison useless. It makes it something to annotate. A single line in the report saying what changed and when means a reader can judge the figure instead of trusting it or dismissing it. The alternative is a tracking change from eight months ago quietly explaining a trend that everybody attributed to something else.

The cheapest version of this discipline is a dated list kept beside the report. One line per change, with the date it took effect: new website, tracking reinstalled, consent banner added, a service dropped, a branch opened. It takes minutes to maintain and it answers most of the awkward questions a year on year comparison will eventually raise.

Percentages, and the small numbers problem

A percentage change is a compression, and compressions lose the information you need most.

Two inquiries becoming three is a fifty per cent increase. Two hundred becoming three hundred is also a fifty per cent increase. Presented as percentages they are identical, and one of them is noise while the other is a different business. Small bases produce enormous percentages from ordinary variation, and a report built on percentages alone will send somebody chasing a change that was never there.

The fix is to publish both figures every time: last year, this year, and the percentage between them. It takes one more column and it removes the entire class of error.

Where the calendar refuses to cooperate

Three things stop the two periods lining up cleanly, and all three are worth naming in the report rather than hiding.

• Weekdays. The same date falls on a different day of the week each year, so a month can contain five Saturdays one year and four the next. For anything weekday driven, align the days rather than the dates.
• Moving holidays. Holy Week moves between March and April. Any comparison that crosses it needs the two months read together, or the weeks around the holiday compared instead.
• Leap years. A small effect and a real one for daily totals in February.

None of these are difficult. They are simply invisible until somebody is asked to explain a twenty per cent swing that turns out to be a calendar.

Reading it beside the other numbers

Year on year is a direction, not a diagnosis. It says whether this period is better than the same period last year. It does not say why, and on its own it invites the wrong explanation.

Put it beside the numbers that describe the mechanism. Conversion rate separates a change in how many people arrived from a change in how many of them acted. Average order value separates more orders from bigger ones. Sessions, inquiries and orders each moving differently is the useful finding, and a single percentage at the top of the report hides it.

Words you will hear

• YoY. The abbreviation, used interchangeably with year on year.
• Like for like. A comparison adjusted so the two periods describe the same thing, for example excluding a branch that opened during the year.
• Rolling twelve months. The last twelve months as a single block, compared against the twelve before it. Smoother than a single month and slower to react.
• Seasonality. The repeating annual pattern that year on year is designed to hold still.
• Base effect. A change that looks dramatic only because the earlier figure was unusually high or low.

A useful habit is to write the comparison you used into the report itself. Year on year, month on month and rolling twelve months answer three different questions, and a number with no stated comparison is an invitation for the reader to assume the one that suits them.