The numbers a small business owner should see every week

I ran multi-unit restaurants for sixteen years before starting SEOD, across more than $54M in annual P&L. The weekly packets I inherited were twelve pages. Nobody read page four. The general managers scanned the top of page one, made a decision, and went back to the floor.

That is not laziness. That is a correct response to a report built for the person who assembled it rather than the person who has to act on it.

Why five numbers and not twenty?

Because a number you look at without acting on is a number you have trained yourself to ignore.

Every metric on a weekly report should have an owner and a possible response. If your website sessions went up nine percent this week, what do you do on Monday? Nothing. So sessions do not belong on the weekly. They belong on a monthly trend line where the movement is large enough to mean something.

A weekly metric earns its place only if a bad week changes what you do next week. Revenue does. Leads do. Labor hours do. Impressions do not.

The second reason is attention. When twenty numbers are in front of you, the two that matter get the same weight as the eighteen that do not. You end up reacting to whatever moved the most in percentage terms, which is almost always the smallest number on the page.

The third reason is that most of what fills a long report is not decision-grade anyway. Kohavi, Deng, Longbotham and Xu, writing in the KDD 2014 proceedings from thousands of controlled experiments at Amazon, Booking.com, LinkedIn and Microsoft, put the problem in one line they call Twyman's law: "Any figure that looks interesting or different is usually wrong!" Their own tests ran across hundreds of thousands to millions of users, and even at that scale the dramatic movements were usually instrumentation. On a weekly page built from forty leads, the odds are worse.

Which five should I pick for my business?

Start with the one your business dies without. For a restaurant that is covers. For a home service company it is booked jobs. For a clinic it is scheduled appointments. Write it down first.

Then add the number that feeds it. Leads, inquiries, calls, whatever converts into the first number. If bookings fall next month, this is the number that told you three weeks early.

Third, add your largest controllable cost. Labor for most service businesses. Food and labor together if you are running a kitchen. This is the line with the most published evidence behind it. The National Restaurant Association's 2025 Operations Data Abstract, built from financial data supplied by more than 900 operators, puts full-service salaries and wages including benefits at a median of 36.5% of sales, and splits that by outcome: operators who reported a pre-tax profit ran 34.2%, and operators who reported a loss ran 42.9%. The gap between a profitable full-service restaurant and a losing one is 8.7 points of labor, and most of that gap is made of scheduling decisions taken on a Tuesday, which is why the damage usually comes from the schedule rather than the wage rate.

Fourth, your top marketing channel's spend and its cost per lead. One channel. The biggest one.

Fifth, cash. Not profit. Cash in the account, plus what is coming, minus what is due. Profitable businesses close every year because nobody put this on a weekly page.

What does a real weekly page look like with numbers on it?

Five rows. Here is a full week for a single-unit full-service restaurant, with the arithmetic shown so you can rerun it with your own figures.

Row one, revenue. $23,400 this week against $22,100 the same week last year. That is $1,300 more, or +5.9%derived. Same week last year, not last week, because last week had a holiday in it and this one does not.

Row two, new bookings. 61 covers booked this week against 58 in the same week last year. Three bookings. Write it as three, not as a percentage, because three is what it is.

Row three, channel spend and cost per lead. $520 on Google Ads producing 17 tracked leads. $520 divided by 17 is $30.59 per lead (derived). That number is only usable against your own trailing weeks, for reasons the next section covers.

Row four, labor. $8,915 in wages against $23,400 in sales is 38.1%derived. Set that beside the Association's medians and the read is immediate: you are 1.6 points above the all-respondent full-service median of 36.5% and 3.9 points above the 34.2% that profitable operators ran. 3.9 points of $23,400 is $913 in a single week (derived). Hold that for a year and it is roughly $47,000, which is larger than most small marketing budgets.

Row five, cash. $41,000 in the account, $18,600 due inside fourteen days, $12,400 expected in. That is the row nobody wants and the row that ends businesses.

Now the part that makes it a weekly page rather than a report. Only row four triggers an action this week, and the action is specific: pull the schedule for the two shifts that ran heaviest and cut where the cover count did not justify the cover. Rows one, two, three and five get noted and nothing else. Four of five numbers producing no action is a healthy week, not a wasted page.

Should I compare these numbers to an industry benchmark?

Mostly no, and this is where weekly reporting goes wrong most often.

Take the cost per lead above. LocaliQ, which sells campaign management, publishes search advertising benchmarks by industry, and its 2026 edition puts Restaurants & Food at an average cost per lead of $30.57 against $66.69 across all industries. Our worked example landed at $30.59, which looks like a perfect match and means almost nothing.

Here is why. Those figures are campaign-level medians across thousands of US accounts nationally, not page-level or week-level benchmarks. LocaliQ disclosed the method for the 2025 edition of the same series: 16,446 US search campaigns, a minimum of 64 active campaigns per category, and a note that its stated averages are technically medians to control for outliers. The 2026 edition does not publish its sample size at all. A national median across a whole category tells you nothing about your city, your ticket size, your landing page or your week.

And you should be careful about lifting numbers off a page wholesale, including that one. The same LocaliQ article carries the claim that 88% of searches for a local business result in a call or visit. That figure has no primary source in that form. The underlying research is a May 2014 deck by Google with Ipsos MediaCT and Purchased, built from an Ipsos online survey of 4,500 respondents plus a mobile diary study in which 3,431 smartphone searches and 2,262 store visits were logged. What that deck actually says, verbatim, is that "50% of consumers who conducted a local search on their smartphone visited a store within a day, and 34% who searched on computer/tablet did the same." Neither 88% nor the widely repeated 76% appears anywhere in it. A page can carry a sound benchmark table and a corrupted statistic in the same column, so take the table and check the sentence.

The operating rule is the same one that applies to Google Business Profile reporting, where there is no current published benchmark worth using either: compare a number to its own trailing weeks and to your own named local competitors, never to a published average. Your history is the only benchmark that shares your conditions.

How do I know whether a weekly number actually moved?

Mostly you do not, and that is the honest answer nobody selling dashboards will give you.

Week to week, a small business generates small samples. Forty leads becomes thirty-four and you have not learned anything. Weather, a holiday, one large customer, a staff absence. All of it moves the number more than any decision you made.

There is a formal version of this and it is worth knowing. In the KDD 2014 paper above, the authors state that the minimum sample size for a test is a function of "the metric's variance and sensitivity (the amount of change one wants to detect)," not a fixed traffic number. Their own worked table shows the answer moving by a factor of more than twenty across metrics on the same site: one Bing metric needed 114k users per variant to detect a 4.4% change, while a less volatile one needed 4.70k for 5.4%. The noisier the metric, the more of it you need before a move means anything. Weekly leads for a local business is a very noisy metric with a very small base.

So use the weekly numbers as an alarm, not as evidence. The weekly report tells you where to look. It does not tell you what happened. Compare against the same week last year rather than last week wherever your business has any seasonality at all. Comparing a rainy February week to a sunny one tells you about rain.

Where do these numbers come from without building a project?

From four places, and you probably already have three of them.

Your point of sale or booking system holds revenue and volume. Your payroll or scheduling tool holds labor. Your bank holds cash. Your ad accounts hold spend.

For website and lead numbers, a small number of Google Analytics 4 reports will cover everything a weekly page needs without you learning the rest of the tool. Set them up once.

The assembly is the part people overthink. A spreadsheet with five rows, filled in Monday morning, beats a connected dashboard that takes six weeks to build and gets abandoned. Automate later, once you know the five numbers are the right five, which usually takes a quarter.

One warning worth knowing before you start: your ad platform and your website analytics will report different numbers for the same week, and neither of them is lying to you. Pick one source per metric and stay with it.

What to do this week

Open a blank spreadsheet. Five rows, one column per week, twelve columns.

Fill in the last four weeks from records you already have. That takes about an hour and it gives you something the report alone cannot: a starting position. Capturing where you actually stand before you change anything is the difference between knowing a change worked and believing it did.

Write the trailing range beside each row rather than a national benchmark. Lowest and highest of the last twelve weeks is enough.

Then put fifteen minutes on your Monday calendar, recurring, and do not let it move. The habit matters more than the spreadsheet.

If your lead number is the one that looks weakest, resist the urge to buy traffic. Check what the page receiving that traffic is asking people to do first, because a homepage and a landing page have different jobs and sending campaign traffic to the wrong one wastes the spend before analytics ever sees it.

Questions about your weekly numbers?

Email me at eric@seod.com with the list of what is currently on your weekly report, however messy. I will tell you which lines to cut, which one is missing, and which two are probably contradicting each other. Paste it in the body of the email, no attachment needed.

I answer these myself and it takes me about ten minutes. There is no sequence attached, and if the honest answer is that your current report is fine, I will say so.

Or keep reading the rest of the analytics library.