The Baseline

Measure before you change anything: door counts, sales per head, same-weekday comparisons, and an eight-week normal.

Here is a question that sounds simple and isn’t. Was last Thursday a good night? You’ll have an opinion. The room felt full, the bar was busy, the band was tight. But “good” compared with what? Last Thursday, or the Thursday before? A Saturday? The Thursday of a county fair? Without a normal to compare against, every number is a mood. The baseline answers that question. It is the normal level of a number, measured before you change anything, with a record of how much it usually moves.

A baseline is the normal level of whatever you’re measuring, worked out before you change anything. It’s the most skipped step in small-business PR and the one that makes every other step honest. Chapter 21 introduced it and Chapter 22 told you which numbers to track. This chapter shows how to build the normal for them: how many weeks to gather, how to compare like with like, what to do about weather and holidays, and how to find the normal swing that tells a real change from an ordinary wobble. It also covers the patterns that eight weeks can’t see, such as seasons and the yearly calendar, and how to compare this year with last year without fooling yourself.

The method is deliberately plain. You need a notebook or a spreadsheet, a few minutes after each night, and the discipline to write the same things down every time. The reward is something most rooms never have: a real answer to the question “is this working?” At the end you’ll fill in the Baseline Worksheet, and Appendix B carries a blank copy with an eight-week example you can follow line by line. If you only ever build one thing from this book, build this. Every other measurement depends on it.

PR TACTIC: The Eight-Week Notebook

What it is: After every night for eight weeks, write down the same five things: headcount, sales, weather, anything unusual, and one word for how the room felt.

Why it works: Eight comparable nights are enough to show the normal level and the normal swing, and the notes explain the nights that don’t fit.

Try this: Start tonight. Write the five things on one line in a notebook and keep it by the register for eight weeks before changing anything.

23.1 What a Baseline Is, and Why Eight Weeks

A baseline has two parts. The first is the typical level: add up the figures for several comparable nights and divide by the number of nights. The second is the normal swing: the lowest and the highest of those nights. Together they say, “a normal Thursday here is about 60, and it ranges from 52 to 69.” With that in hand, a night of 62 is nothing, a night of 50 is a mild dip, and a night of 78 is the first number that deserves a real question. Without it, each night is a story that proves whatever you already believe. The swing is the part most people skip, and it is the part that does the most work. An average alone tells you what a typical night looks like. The swing tells you how different a normal night can be, which is what you need to judge whether a change is real. Two rooms with the same average of sixty can have very different swings, one from fifty-five to sixty-five and one from thirty to ninety, and a night of seventy-five means something very different in each.

Why eight weeks? Because fewer than that is too easy to bend. Three Thursdays can be three lucky or unlucky nights, while eight start to show the pattern and give you a swing wide enough to be useful. If you can’t wait that long, use what you have, six at the minimum, and say so. If you already have records, such as door counts or point-of-sale reports, use them to build a baseline at once. The eight-week number isn’t a law. It’s a practical compromise between waiting too long and drawing a conclusion too soon. More weeks make the baseline more reliable, as long as nothing big changed in between. Weekly nights fit eight weeks well. Monthly events need more patience: eight months is a long time to wait, so use what history you have and treat the first year as the baseline year. Classes in eight-week series can use past series as the comparison, with the same week of the series compared to the same week.

The baseline also has an enemy: change. If you alter the cover, the format, the start time, and the posting schedule during the eight weeks, you no longer have a baseline of the original night. Make one rule for the period: change nothing you intend to test. That doesn’t mean freezing the business. It means holding still on the thing you’re about to measure, and writing down anything else that happens. A baseline that’s been disturbed is still useful as a history, but it can’t be used as a yardstick. If you can’t avoid a change during the baseline weeks, write the date and the change on the sheet and split the baseline in two, before and after. Two short baselines with a known change between them are more useful than one long one that hides it.

QUESTIONS PEOPLE ASK

“I just opened. I don’t have eight weeks.” Start counting tonight. Use the first four weeks as a rough baseline, mark it as rough, and replace it at eight. Don’t test big changes until you have it.

“My nights are all different.” Then build a baseline for each kind of night, even if each has fewer weeks. A rough baseline for the right kind of night beats a tidy one for the wrong kind.

23.2 Same Night, Same Kind

The comparison has to be fair. Compare Thursdays with Thursdays, never with a Saturday. Compare a music night with a music night of the same kind, and a beginner lesson with a beginner lesson. The same goes for time of day: the first hour of a night isn’t comparable to the last. If you run two sets, count them separately. People who work in rooms know this by instinct. You can usually tell by five o’clock how the night is going to run. The point of writing it down is to stop leaving that knowledge in your head, where it only gets used to protect your pride. Fairness also means comparing the same part of the year when you can. A Thursday in March and a Thursday in August may both be Thursdays, but in much of Texas the heat, school calendars, and harvest or hunting seasons can change who goes out. The seasonal box in this chapter shows how to handle that.

Count the same way each time. The door count should be taken by the same person, at the same point, with the same method: a clicker, a tally, a wristband count, or the cover receipts. The most common baseline error isn’t arithmetic, it’s a changing method. If last month’s count came from the register and this month’s from a clicker, the two aren’t comparable. Decide the method once, write it at the top of the sheet, and don’t change it without drawing a line across the page. Staff who know why will help keep it consistent. If you have a point-of-sale system, the number of tabs or transactions is a useful check on the door count, as Chapter 21’s case study showed. A clicker counts every entry, including smoke breaks and the patio. Tabs count paying parties. Neither is perfect, but when the two disagree, you learn something about the count.

Sales per head is the second number worth building, and it’s simple: the night’s sales divided by the headcount. It tells you something the door count can’t. A busy night with low spend per head may be full of people who came for a deal. A quiet night with high spend may be a room of regulars who linger. Use the point-of-sale report for the night’s sales, and take food and drink separately if you can. Margin is better than sales, as Chapter 32 explains, but sales per head is a good place to start because almost every room already has the data. Record sales per head separately for nights with and without a cover, and with and without food, since those change the figure in ways that have nothing to do with how good the night was.

COMMON MISTAKES WHEN BUILDING A BASELINE

Too few weeks. Three nights are a story, not a normal.

Mixed nights. A Thursday lesson and a Saturday dance averaged together describe neither.

Flags added afterward. Excuses invented once the number is known aren’t flags.

Changing the count method midway. A clicker one month and tabs the next isn’t a series.

Forgetting the swing. An average without its range can’t tell you whether a night is unusual.

Never refreshing. A baseline from last spring can’t judge this fall.

23.3 Weather, Holidays, and the Nights That Don’t Count

Some nights shouldn’t go into the baseline, or should go in with a flag. A county fair weekend, a rodeo, a big football game, a holiday, the week of payday, a thunderstorm, a heat wave, a cold snap, a road closure, the night the creek came up and half the county stayed home. Each changes who comes and how many. Write the reason next to the number the same night, before you know whether the night was good. A note made in the moment is evidence. A note made afterward, when you’re looking for an excuse, is a story. A short list of standard flags helps. Write them at the top of the sheet: H for holiday, W for weather, E for a nearby event, P for a private event, S for a staffing problem, and O for anything else, with a word of explanation. Anyone counting can mark a flag on the night, and the baseline stays honest.

There are two honest ways to handle a flagged night. One is to leave it out of the average and say so. The other is to keep it and show the average both ways. If your eight Thursdays include a rodeo week, compute the average with all eight, then with the seven that remain, and note the difference. If the two are close, the flagged night didn’t matter much. If they’re far apart, it did, and you should be wary of any conclusion that depends on it. Showing both is the more candid approach, and it protects you from deleting inconvenient data. Showing both also makes a good conversation starter at the monthly meeting. When the two averages differ a lot, someone usually asks why, and the answer often points to something worth planning around, such as a festival weekend every year that the calendar should prepare for.

A caution about overcorrecting. If every low night gets a reason, the baseline becomes whatever you wish it were. A fair rule is to flag only events you wrote down at the time and that plausibly change attendance for the whole town, such as weather, holidays, or major events. A personal reason, like “we didn’t post this week,” isn’t a flag. It’s a variable you might want to test. Keep the list of flags short and written. A baseline with five flags in eight weeks is telling you the nights are too unpredictable to compare and that you need more weeks. When in doubt, keep the night in and add a note. It’s better to have a baseline that’s a little noisier than one that’s been trimmed to look tidy.

SEASONS, THE YEARLY CALENDAR, AND YEAR-OVER-YEAR

Know your seasons. Most rooms have predictable busy and slow stretches: summer heat, school starting, holidays, rodeo and fair season, football season, the weeks after the new year. A baseline built in one season can’t judge a night in another.

Build a baseline per season. Once you have a year of counts, make four quarterly baselines, or as many as your seasons need. Compare an August Thursday with the summer baseline, not the spring one.

Compare year over year. The fairest comparison for a seasonal room is often the same weeks last year. Line up the same weekday in the same weeks, such as the four Thursdays after Labor Day each year, and compare the averages. Note anything that moved, such as a holiday that fell on a different day.

Watch for growth hiding inside a season. If this summer’s Thursdays average 52 and last summer’s averaged 44, the room grew even if 52 is below the spring baseline.

Keep the calendar. Mark next year’s known events on this year’s baseline sheet as soon as you learn the dates. A room that plans around its seasons wastes less money promoting nights that would have been slow no matter what.

23.4 Building It: The Arithmetic, Step by Step

The arithmetic is addition and division, and it’s worth doing by hand once. List the headcounts for your comparable nights in a column. Add them. Divide by the number of nights. That’s the average. Find the lowest and highest, which is the swing. If you like, find the middle value too: sort the numbers and take the one in the center, or the average of the two in the center if there’s an even number. The middle value is called the median, and it’s useful because one odd night pulls the average but not the median. When the two are close, the data is steady. When they’re far apart, a night or two is distorting the average. The median is easy to find by hand. Sort the numbers from smallest to largest. With an odd number of nights, the middle one is the median. With an even number, it’s halfway between the two middle ones. For 52, 55, 58, 60, 61, 64, 68, 70, the two middle numbers are 60 and 61, so the median is 60.5.

Then add the second layer: write the same things for sales per head. Divide each night’s sales by its headcount, then average those figures across the eight weeks. Notice that this isn’t the same as dividing total sales by total heads, though the two are usually close. Pick one method and stick to it. Finally, write one line at the bottom: “Baseline for Thursday, eight weeks to June 30: average 61, swing 52 to 70, median 60, flagged nights: 1 (rodeo). Sales per head: $24, swing $21 to $27.” That line is the yardstick you’ll hold every later change against. If you track first-timers and return rates, do the same for them. Their swings are usually wider than the door count’s, because the counts are smaller. A first-timer count that ranges from six to fourteen on a normal Thursday means a night of twelve is not news.

Keep the baseline alive. A baseline is a snapshot, and rooms change: seasons, staff, neighborhoods. Refresh it every quarter, or whenever you make a lasting change you want to track. Keep the old ones, because trends show up over several. If you change the format permanently, start a new baseline for the new night and mark the date. And store it where you’ll find it. A baseline in a drawer is a record. A baseline on the wall behind the register is a tool. The worksheet below and Appendix B are built for the wall. When you refresh, keep the old baseline line and write the new one beneath it. The sequence of baselines, quarter after quarter, is itself a useful record. It shows whether the room’s normal is rising, falling, or holding, which no single baseline can tell you. Put the date of each refresh at the top, and keep the old sheets in the same folder, oldest at the back.

23.5 Using the Baseline: Real Change or Ordinary Wobble?

Once you have it, using it is simple. When a number comes in, hold it against the swing. Inside the swing, it’s ordinary. A night below the bottom or above the top gets a question. One such night means little, since swings aren’t perfect walls. Two or three in a row, or a steady drift in one direction for several weeks, means more. It’s the same instinct you’d use with the weather: one cold day isn’t a change in the season, but a week of them is worth a coat. Chapter 31 shows how to test a specific change using the baseline as the yardstick. Two nights outside the swing in the same direction are worth more attention than one. Three in a row are a pattern. Use that simple rule in the weekly check, and you’ll avoid both overreacting to a single night and ignoring a real change.

There are limits. A baseline of eight weeks can’t see slow trends, and it can’t be used across seasons, because summer and winter nights differ. Compare a summer baseline with a summer night. Its swing is only a rough guide, and a night just outside it isn’t a revolution. If your numbers are small, such as fifteen people, a swing of five is a third of the night, and a change that big could be luck. Use the baseline to ask better questions, not to produce certainty. And never use it to excuse a bad trend because “it’s within the swing.” A slow drift inside the swing is still a drift. To see slow trends, compare baselines, not nights. If this quarter’s Thursday average is 58 and last quarter’s was 63, and the swings overlap, the change may be noise. If it’s the third quarter in a row of falling averages, it’s a trend, even though no single night was outside its swing.

Finally, write the baseline into the plan. The Chapter 4 page asks for a baseline for the measure. Now you have one. Add it to the metric tree from Chapter 22. Put the date next to it, and the dates of any flagged nights. Then, and only then, start the thing you wanted to test. Everything that follows in Part Three, attribution, funnels, lists, content, partnerships, experiments, assumes you’ve done this first. It’s dull work, and it’s the reason the rest of the numbers can be trusted. Keep a copy with the date somewhere safe as well. When someone asks next year whether the new lesson night helped, this is the page that will answer.

THREE RULES FOR READING A NUMBER AGAINST ITS BASELINE

1. Inside the swing: ordinary. Note it and move on.

2. Outside the swing once: worth a question. What else happened? Was anything flagged?

3. Outside the swing twice in a row, or three times in a quarter, in the same direction: a pattern. Look for the cause and decide what to do.

These are this book’s working rules for small rooms, not statistical laws. They’re built to keep you from overreacting and from ignoring a real change.

TRY THIS WEEK

Pull the last eight weeks of door counts for one night. Mark any flagged nights. Work out the average, the median, and the swing by hand. Write the one-line baseline on a card and tape it beside the register.

CASE STUDY

Eight Thursdays and a Rodeo

Thea manages a dancehall’s Thursday night. She records headcounts and sales for eight weeks. Headcounts: 58, 62, 55, 71, 60, 64, 91, 59. The 91 was the Thursday of the county rodeo, which she flagged that night. Sales that night were $1,380, $1,500, $1,320, $1,780, $1,440, $1,600, $2,520, and $1,475, in the same order.

Step 1 — Average with all eight. 58 + 62 = 120, plus 55 is 175, plus 71 is 246, plus 60 is 306, plus 64 is 370, plus 91 is 461, plus 59 is 520. 520 ÷ 8 = 65. The swing is 55 to 91.

Step 2 — Average without the rodeo. 520 − 91 = 429. 429 ÷ 7 = 61.3. The swing is 55 to 71. The rodeo night pulled the average up by 65 − 61.3 = 3.7 and stretched the swing from 16 to 36. She uses 61 as her normal and notes the rodeo.

Step 3 — The median. Without the rodeo, sorted: 55, 58, 59, 60, 62, 64, 71. The middle value is 60. It’s close to the average of 61, so the data is steady.

Step 4 — Sales per head. Dividing each night: 1,380 ÷ 58 = $23.79; 1,500 ÷ 62 = $24.19; 1,320 ÷ 55 = $24.00; 1,780 ÷ 71 = $25.07; 1,440 ÷ 60 = $24.00; 1,600 ÷ 64 = $25.00; 1,475 ÷ 59 = $25.00 (not counting the rodeo, which was 2,520 ÷ 91 = $27.69). The seven ordinary nights add to $171.05, so the average is $171.05 ÷ 7 = $24.44. Her baseline line reads: average 61, swing 55 to 71, median 60, sales per head about $24, rodeo flagged. When a Thursday later comes in at 66, she knows it isn’t news. All figures are illustrative.

REAL TALK

From behind the bar:

For years I told the owner Thursdays were “getting better.” When we finally wrote down eight weeks, they were exactly the same as the year before. I was embarrassed for about a day. Then I was glad, because now when something moves, we’ll know.

HEADS UP

Counting heads is harmless, but tying counts to individuals is not. If your baseline starts to include names, phone numbers, or tabs linked to people, privacy and consent rules may apply, and so may rules about how long you keep records. Sales records also matter for sales tax and your books. This is general information, not legal or tax advice. Keep baseline records anonymous and aggregate where you can, ask a CPA about retention of sales records, and see Chapter 35 and Appendix J.

BASELINE WORKSHEET, EIGHT COMPARABLE NIGHTS

Use one sheet per night type, such as Thursday. Write the counting method here before you start, and don’t change it.

Fill in: Night, Counting method, Start date.

Columns to fill in: Week, Date, Headcount, Sales, Sales per head, Weather / event, Flag?. Copy as many rows as you need.

Average = total headcount ÷ number of nights. Swing = lowest to highest. Sales per head = sales ÷ headcount. Show the average with and without flagged nights.

MEASURE IT

The number: Your baseline line: the typical level, the normal swing, the median, and the number of flagged nights, for one comparable night over eight weeks.

By hand: For eight comparable nights, add the headcounts and divide by eight. Find the lowest and highest. Sort the figures and take the middle. Flag nights you marked at the time. Example: average 61, swing 55 to 71, median 60, one flag. Do the same for sales per head.

Watch out: Don’t change the cover, format, or counting method during the eight weeks. Don’t add a flag after the fact to explain a bad night. Seasons differ, so refresh the baseline each quarter.

Go deeper: Chapter 31 (running experiments in a real room) and Appendix B (baseline worksheet with an eight-week example).

ROLE BOX

• Owner: Pick your key night, hold the format still, and write headcount, sales, weather, and one word about the room after every night for eight weeks.
• Booker: Build a baseline for each night type, and one for acts: average draw per act over their last several dates.
• Musician: Keep a draw record for each room: door count or tickets per date. Your normal draw is what you can prove.
• Bartender: Baseline your average views, sends per reach, and visits you can trace to content over eight posts of the same kind.
• Dance Teacher: Baseline your beginner class: average new students, average returners, and swing, over eight weeks of the same lesson.

TALK ABOUT IT

• What is a normal night for you? Can you state it as an average and a range, or only as a feeling?
• Which nights should you flag, and how do you keep that list honest?
• What change are you tempted to make before you have eight weeks of baseline?