Year over year, sequential, or moving total: choosing a comparison
Every retail figure is reported against something. Against the same period last year, against the immediately preceding period, or against a moving total. The three answer different questions and can point in different directions at the same time without any of them being wrong.
Choosing deliberately is part of the analysis rather than a formatting decision.
What does year over year answer?
Whether the business is bigger than it was a year ago, with seasonality largely controlled because the comparison period has the same seasonal position. It is the default for good reasons and it has two weaknesses.
The first is the base: an unusual prior year distorts the comparison for a full year afterwards. The second is lag: it reacts slowly to a turning point, because half the comparison is old.
What does sequential answer?
What just changed. It is the most responsive comparison and the most contaminated by seasonality, which is why it is usually applied to seasonally adjusted figures rather than raw ones.
Sequential comparison on raw weekly retail data is rarely informative. Every December beats every November.
What does a moving total answer?
The level over a rolling window, which smooths noise and makes the direction visible. A twelve-month moving total removes seasonality arithmetically and shows the underlying level clearly.
Its weakness is lag and the way a single unusual period stays in the window for the full length of it, holding the line up or down long after the event.
How do they disagree?
Easily and informatively. A business can show year-over-year decline, sequential growth and a falling moving total all at once — which describes a business that fell, has started recovering, and has not yet made up the ground.
That combination is a clearer picture than any single measure. Reporting all three is more honest than choosing the flattering one.
What about the base-period problem?
A prior period distorted by an event distorts every comparison against it. The standard responses are to compare against two years earlier as well, to note the base explicitly, or to use a moving total that dilutes it.
What should not happen is quietly choosing the comparison that removes an inconvenient base. Any comparison chosen after the numbers arrive is a rationalisation.
How does store churn affect these?
Substantially, and it is the reason same-store measures exist. A total that grew because more stores reported is not growth in the market, and only a like-for-like basis separates the two.
Ask on what basis any comparison is computed. A total that includes new stores and a same-store figure are different claims.
What should a report show?
The measure, the basis, the comparison period, and the store basis. Four clauses again, and their absence is what makes a chart unfalsifiable.
Frequently asked questions
Which comparison should a monthly report lead with?
Year over year on a like-for-like basis, usually, with a moving total for direction. Sequential belongs in the detail rather than the headline unless the series is adjusted.
What if last year was unusual?
Say so in the same sentence as the number, and add a two-year comparison. A reader who knows the base can interpret the figure; one who does not cannot.
Is a four-week comparison better than monthly?
For weekly-organised businesses, often, because it avoids months with different numbers of weekends. Consistency matters more than which convention is chosen.
Does a moving total hide a turning point?
It delays it. That is the trade for smoothness, and it is why a moving total belongs alongside a more responsive measure rather than instead of one.
How do I compare across markets?
On a like-for-like basis with the same period convention in each, which is harder than it sounds when calendars differ. State the convention.
What is the most common presentation error?
Mixing bases within one chart — a total-market line against a same-store line — without labelling. The lines then diverge for a reason that is not about the market.
Should I ever use a two-year stack?
It is useful exactly when a base year is distorted, and it should be labelled as what it is. General retail series published by the Census Bureau show how a careful publisher handles comparison bases.
What settles a disagreement about which to use?
Asking what decision the number is for. Comparisons are tools for questions, and the question usually names the right one.
### Should the comparison basis ever change mid-year?
Only with the history restated on both bases and the change announced. Changing a basis quietly produces a step in the series that will be interpreted as a market event by everybody who did not make the change.
Where a change is unavoidable — a panel restructure, a definitional revision — publishing a period of overlap on both bases is the practice that preserves comparability. It costs one extra calculation and it saves a year of confusion.
### What should a standard monthly report include?
The period figure, the year-over-year comparison on a like-for-like basis, a moving total for direction, and the store count. Four elements on one page, consistently, every month.
Consistency matters more than sophistication. A report whose shape changes each month cannot be read as a series, which defeats most of the reason for producing it.
What about comparing against plan rather than against history?
It answers a different question — whether the business met its own expectation — and it should sit alongside the historical comparison rather than replacing it. A plan is a forecast, and measuring only against it measures the forecast as much as the market.
Working with retail data? See NRS Insights data solutions or contact NRS Insights.