Year on year comparison that is actually fair

The report says sales were down a tenth on the same date last year, and the owner spends the morning worrying. Then someone points out that last year the date fell on the busiest night of the week, and this year it fell on one of the quietest. Compared with the matching night last year, sales were actually up. The figure was right; the comparison was wrong.

A year on year comparison is one of the most useful checks a restaurant has, because it sets each period against the same season a year before, with the same weather, the same holidays and the same rhythm of trade. But it is only fair if the periods really match. Dates move across the days of the week, holidays fall on different days, festivals move by weeks, and venues close, extend or change their hours.

This guide covers how to match weeks rather than dates, how to allow for closures, holidays and festivals that move, and how to read the result honestly. Every venue and figure in it is an example.

Comparing with last year removes most of the season from the figure. A spring month compared with the winter before will almost always look better, and a winter month compared with the summer will look worse, whatever the venue did. Compared with the same weeks a year earlier, the season is roughly the same, and what is left is mostly the venue's own progress.

It also shows the long trend. Month-to-month figures jump around; a comparison with last year, repeated each month, shows whether the venue is growing, holding steady or slipping.

The same date falls on a different day of the week each year, one day later in an ordinary year and two after a leap year. For a restaurant, where a busy weekend night may take several times what a quiet weeknight does, comparing the same date can swing the figure wildly.

The problem grows with the period. A single day compared by date is almost meaningless; a week compared by date may include one more busy night than last year; even a month can contain one extra weekend, which can move the comparison by several per cent.

The fix is to compare matching weeks rather than matching dates: this year's week 12 against last year's week 12, each running from the same day of the week. Each week then has the same mix of days, and the comparison is like for like.

Many businesses use the ISO week date system, which numbers the weeks of the year from one to 52 or 53, each starting on a Monday. Using it, or any consistent week numbering, means this year's week and last year's always contain the same days.

For a single day, compare with the same day of the week in the matching week last year, not with the same date. This Saturday against the Saturday of the same week last year is a fair comparison; this Saturday against the same date last year, which may have been a Thursday, is not.

The same rule applies to any period shorter than a year. A fortnight, a month or a quarter should be built from whole matching weeks, so each period contains the same number of each day.

Months are the hardest periods to compare, because they contain different numbers of each day from year to year. A month with five weekends will usually beat one with four, regardless of how well the venue is doing.

Either compare months built from whole weeks, such as four or five weeks each, or adjust for the extra days. Many retailers use a calendar of four-week and five-week months for exactly this reason, so each month always contains the same days as the year before.

Counting in weeks has one complication: some years have 53 weeks rather than 52. In those years, the extra week has nothing to match, and the annual total looks a week's trading better than it is.

Compare the 52 matching weeks and treat the extra one separately, or note it clearly on the report. A year that seems to have grown by two per cent may simply have had an extra week in it.

A venue that closed for a fortnight last year for refurbishment, or for a week this year because of a broken oven, cannot compare those weeks fairly. Compare only the weeks open in both years, and note the closed weeks separately.

The same applies to single days. If the venue closed on a day last year that it opened this year, leave that day out of both sides, or the comparison will show growth that is only an extra day's trading.

Holidays on fixed dates, such as Christmas Day or New Year's Eve, fall on the same date each year but on a different day of the week. When they fall near a weekend, the days around them trade very differently from a year when they fall midweek.

For these weeks, compare the holiday with the holiday, and the days around it with the days around it last year, rather than relying on the week numbers alone. A note on the report explaining the shift saves a great deal of confusion.

Some festivals move from year to year, and they are the biggest trap in any yearly comparison. Easter can fall in March or April. Lunar New Year moves between late January and February. Diwali moves between October and November. Each can lift or empty a restaurant for a week, and a week with the festival compared with a week without it shows a swing the venue had nothing to do with.

Mark the festival weeks in both years, and compare the festival week with last year's festival week, wherever each fell. The weeks that the festival left behind should be compared with each other, too.

Ramadan moves by about eleven days earlier each year, because it follows a lunar calendar. For venues in places where many guests fast, it changes the whole pattern of trade for a month: quieter days, busy evenings after sunset and a different menu, followed by the celebrations of Eid.

A comparison across those weeks is meaningless unless Ramadan is matched with Ramadan. Compare this year's Ramadan weeks with last year's, and the weeks either side with their own equivalents, even though they fall in different months.

School holidays shift by a week or two from year to year, and they change trade: busier lunches and family tables in a holiday area, quieter evenings near offices. A week that was in the holidays last year and is not this year will compare badly or well for reasons beyond the venue's control.

Note the school holiday weeks for both years, and read the comparison for those weeks with care. Where it matters, compare holiday weeks with holiday weeks.

Weather is the one thing a yearly comparison cannot remove. A wet summer week compared with a hot one last year, for a venue with a terrace, will look poor however well the room is run.

Note unusual weather on the report, especially for venues that depend on outdoor seating. Over a month or a quarter, the weather usually evens out; over a single week, it can dominate.

Local events change trade too: a festival in town, a big match, a concert nearby, roadworks outside the door, a new office opening down the street. Each can lift or depress a week in one year and not the other.

Keep a simple diary of the events that affected trade, for both years. Without it, nobody will remember twelve months later why a particular week was so busy.

If prices rose since last year, part of any growth in sales is the higher prices, not more guests or better trade. A venue whose sales rose by five per cent after raising prices by six has actually served fewer guests or sold less.

Separate the growth in sales from the growth in covers. If sales are up and covers are flat, the growth is mostly price. If covers are up too, the venue is genuinely busier.

Compare covers as well as sales, because covers show whether more guests came. The covers report, set against the matching weeks last year, is often the clearest sign of a venue's real progress, because it is not affected by price rises.

Our guide to the covers report, read on a Monday morning covers how to read covers by day, service and source, which is the detail behind a yearly comparison.

Spend per head, compared with the same weeks last year, shows whether guests are spending more or less on each visit. It rises with price increases and better recommendations, and falls with discounts, cheaper menus and guests who order less.

Read it together with covers. More covers at a lower spend may still mean more sales; fewer covers at a higher spend may mean the venue has become more expensive and is losing guests.

Sales by category, food, drinks, desserts and so on, compared with last year, shows where growth or decline is coming from. A venue whose total is flat may be selling more food and fewer drinks, which matters for margin and for what to do next.

Compare each category's share of sales as well as its total. A shift from drinks to food, or from wine to soft drinks, is often the first sign of a change in the guests who come.

If the venue opened for more hours this year, or fewer, the comparison is no longer like for like. A new breakfast service, a later close or a day added to the week each add sales that have no match last year.

Compare the hours that were open in both years, and report the new hours separately. Otherwise the comparison credits the new hours with growth that the rest of the business did not achieve.

The same applies to seats. A terrace added this year, a private room opened or a section closed for repairs changes how much the venue could sell. Sales up by a fifth after adding a fifth more seats is not the same as sales up by a fifth in the same room.

Where seats changed, compare revenue per seat as well as total sales, and note the change on the report.

Seasonality, the yearly rhythm of busy and quiet weeks, is what makes the comparison with last year so useful and so tricky. Knowing the venue's own seasons helps read each comparison in context: a fall in a usually quiet month means something different from a fall in a usually busy one.

Our guides to seasonal trade: mapping your own restaurant year and seasonality: planning for the quiet eight weeks cover how to map and plan for those seasons.

In a group, compare only the venues that were open for the whole of both periods. A new venue adds sales with nothing to compare against, and a venue that closed removes them. Including either makes the group's growth look better or worse than the venues that traded throughout.

Report the like-for-like figure, from venues open in both years, alongside the total. The total shows the size of the business; the like-for-like figure shows how well its venues are doing.

Restaurant benchmarks, figures from industry surveys and reports, can add context to a yearly comparison: whether a fall in covers is happening everywhere or only at your venue. But they measure other venues, in other places, in other ways.

Use them as background, not as a target. A venue that grew less than an industry average may still be doing well for its location and trade, and one that grew more may still have problems the average hides.

A simple check on any yearly comparison is the number of trading days on each side. If this year's period had more days open than last year's, because of a closure last year or an extra day this year, the comparison is not like for like even when the weeks match.

Divide each period's sales by its trading days to see the daily average, and compare that as well as the total. A small rise in the total with a fall in the daily average means the extra days, not better trade, made the difference.

A rolling comparison, such as the last four weeks against the same four weeks last year, smooths out the noise of single weeks. Updated each week, it shows the trend without being thrown off by one wet weekend or one big event.

It is often the most useful version for a manager: frequent enough to notice a change, long enough to trust.

Comparing with two or three years ago, as well as last year, shows whether the trend is real. A venue that is up on last year but still below two years ago is recovering, not growing; one up on both is genuinely ahead.

Apply the same rules to each year: matching weeks, festivals matched with festivals and closures left out. The further back you go, the more notes you will need, which is another reason to keep them.

Every fair comparison involves judgements: which weeks to leave out, how to treat a festival, what to note about the weather. Write them down on the report, in a few words each.

Twelve months later, the notes are what make next year's comparison possible. Without them, nobody will remember why a week was excluded or why a figure was adjusted.

When sharing a comparison with a partner, a landlord or the team, show both the raw figure and the fair one, with a sentence on the adjustments. A figure that has been adjusted without saying so looks like spin when someone finds out.

Most readers accept adjustments readily when they see the reasons: a festival that moved, a fortnight closed for repairs, an extra weekend. They are far less forgiving of a comparison that turns out to have been unfair.

A venue in its first year has nothing to compare with. Compare instead with the budget or forecast, week by week, and keep careful notes of closures, events, weather and festivals, so that the second year's comparison can be fair.

The first full year of trading becomes the baseline for every comparison after it, so it is worth recording well.

An invented brasserie compares its spring with last spring. By date, April looks down six per cent. Matched by week, from the same day of the week, the four comparable weeks are up three per cent. Easter fell in March last year and in April this year, so the Easter weeks are compared with each other, and the brasserie finds its Easter trade was up slightly.

Covers for the matched weeks are flat, and spend per head is up four per cent, matching the price rise in January. The owner's conclusion is that the room is holding its guests and the growth is mostly price, so the next quarter's focus is bringing in more guests on quiet weeknights.

servearound compares each range with the period of the same length just before it, not with last year, so the year on year comparison in this guide is set up by hand: run this year's weeks as a custom range, then last year's matching weeks, and set the two side by side. Reservation analytics show covers day by day, by shift, by source and by party size, with covers by day and hour; payment analytics show what was collected by day and revenue by hour; and the menu report shows each category's units, revenue and share of sales, though it has no comparison of its own with an earlier period.

Every report runs on the venue's own clock, and the dashboard gives revenue, covers and spend per cover for any range. A venue in its first year on servearound will have last year's figures in its old system or its accounts, so the first comparisons are made from those. See Analytics.

Because the same date falls on a different day of the week, which can swing a restaurant's figures widely.

Matching weeks, each starting on the same day of the week, and the same day of the week within them.

Compare the festival weeks with last year's festival weeks, wherever each fell.

Leave them out of both years and note them separately.

Not as growth in trade. Compare covers to see whether more guests came.

A comparison using only the venues, hours and seats that traded in both periods.

As background, not as a target.

Compare with your budget, and keep good notes for next year.

A year on year comparison is only fair when the periods match. Compare matching weeks rather than dates, the same day of the week rather than the same date, and months built from whole weeks. Leave out closed weeks, match fixed holidays and moving festivals such as Easter and Ramadan with their own equivalents, and note the weather, local events and school holidays that changed trade.

Separate growth from prices by comparing covers as well as sales, look at sales by category and spend per head, and report a like-for-like figure when hours, seats or venues have changed. Use the covers report and your own map of seasonality to read each result in context, treat restaurant benchmarks as background and write down every adjustment so next year's comparison can be fair too.