The 5 Metrics I Check Weekly and the Ones I Ignore

My weekly review takes under ten minutes. I check five numbers, write two lines in a notebook, and close the laptop. That short habit changed what I build more than any tool I bought this year.

For a long time I opened six dashboards every morning and felt busy. Busy is not the same as informed. Most of what I stared at could not tell me what to make next. Once I cut the list down to five metrics that actually move a decision, the studio got quieter and the output got sharper. Here is the exact list, the ones I dropped, and how I keep the whole thing fast.

The five numbers that actually change what I build

I only track a number if it can change a decision within the next two weeks. If a number is interesting but I would do nothing differently after seeing it, it is entertainment, not data. That filter alone killed most of my old dashboard.

Here is the working list.

First, repeat buyers. How many people bought something from me who had bought before. This is the single strongest signal that a product is good and not just well marketed. Last month 31 of 118 orders were repeat buyers. When that ratio climbs, I make more of the same category. When it drops, I look at the product, not the ads.

Second, email replies. Not opens, not clicks. Actual replies from people who read what I sent and typed a sentence back. I got 22 replies to my last newsletter. Nine of them mentioned the same missing feature. That told me exactly what to build next, and I built it.

Third, orders per active product. I divide total orders by the number of listings pulling weight. If one product does 60 orders and eight products do 3 each, I stop pretending I have nine products. I have one product and eight distractions. This number tells me where to cut.

Fourth, help requests logged as support tickets. I track how many people ask for help after buying, and what they ask. High help volume on a product usually means the description promised something the file did not deliver. I fix the description or the file.

Fifth, cash in the account at week end versus week start. Simple. Did the number go up or down, and by how much. No fancy dashboard needed. A spreadsheet cell and a bank tab.

Five numbers. Each one points at an action. If a metric cannot pass that test, it does not make the list. I covered the broader operating setup in Claude Blueprint, which is where the review habit started.

The vanity metrics I stopped opening

I used to check follower counts every morning. I had 4,200 followers on one platform and I refreshed the number like it paid rent. It did not. Followers are a lagging, noisy signal that rarely maps to anyone buying anything. I have posts with 40,000 views that sold zero units and posts with 900 views that sold eleven. Views told me nothing useful about that gap.

So I stopped opening the follower tab. I did not delete the accounts. I just removed the number from my daily loop. Nothing bad happened. The accounts kept growing at the same slow pace whether I watched them or not.

Same story with impressions. Impressions measure how many times a thing appeared on a screen. A screen is not a person and appearing is not reading. I once celebrated a 200,000 impression week and then noticed I had made 14 EUR that week. The two numbers were not related. Impressions are the classic feel-good metric that produces zero decisions.

I also dropped time-on-page for the blog. It sounds smart. In practice a high number could mean someone loved the article or that they walked away with the tab open while making coffee. I cannot tell the difference, so I cannot act on it. Dropped.

Bounce rate went too. Someone lands, reads the one paragraph they needed, and leaves happy. That counts as a bounce and looks like failure. The metric punished me for answering a question well. Gone.

The pattern here is simple. Any number that measures attention without measuring intent is noise for a one-person studio. Attention is cheap and easy to fake. Intent (a reply, a repeat purchase, a support question) is expensive and honest. I want the honest numbers even when they are smaller and less flattering. A quiet 900-view post that sells is worth more than a viral one that does not, and my metric list now reflects that instead of fighting it.

How I keep the weekly review under ten minutes

The review is short because I refuse to gather data during the review. All five numbers live in one place before I sit down. If I have to open five tabs and do math while reviewing, it turns into an hour and I skip it the next week.

Here is the actual flow. Sunday evening, one spreadsheet tab. Five cells at the top, filled by me copying numbers from four sources: the store backend, the email tool, the support inbox, and the bank. Copying takes four minutes because I know exactly where each number sits. I do not explore. I grab and paste.

Then I look at the five numbers next to last week's five numbers in the row below. Up, down, flat. I write two sentences in a paper notebook. Example from last week: "Repeat buyers up, email replies flagged three requests for the same thing, build it." That is the whole output. Two sentences that point at next week's work.

I do not make charts. Charts are where ten minutes becomes forty. A number next to last week's number tells me direction, and direction is all I need to decide. If I want a trend, I scroll up the spreadsheet and read the column. The column is the chart.

For the social side, I batch everything into Buffer once a week so I am not logging into platforms daily and getting pulled into follower counts. Scheduling ahead means the only time I see a social dashboard is when I choose to, not every time I want to post. That single change removed most of my daily vanity-metric temptation.

The store itself runs on Shopify, and I pull the two numbers I need (repeat buyers, orders per product) from its built-in reports rather than a third-party analytics stack. Fewer tools means fewer numbers begging for attention. I deliberately keep the reporting shallow. If a feature would let me slice data forty ways, I do not turn it on, because forty slices is thirty-nine ways to waste a Sunday evening.

What I do with the numbers once I have them

A metric is only useful if it changes the calendar. So each Sunday's two sentences become Monday's task list. That handoff is the entire point of the review.

Concrete example. Three weeks ago the orders-per-product number showed one product doing 58 orders and six others doing under 4 each. I killed the six. Not paused, killed. Removed them from the store. My catalog dropped from seven products to one plus two I was actively improving. The next week was calmer and the strong product sold more because the store stopped scattering attention.

Another example. Email replies flagged that people wanted a voiced version of a written guide. Nine separate replies over two weeks. That is a loud signal from a small list, so I acted. I recorded it using ElevenLabs for the narration and shipped it as a bonus file. Sales on that product went up the following week and support questions went down, because people who prefer listening got what they wanted.

The help-request metric works the same way. When one product generated four support tickets in a week, all asking where a promised template was, I did not argue. The description overpromised. I rewrote it in twenty minutes and the tickets stopped. The metric did not just measure a problem, it located it precisely enough to fix in one sitting.

Cash at week end is the metric I act on least often but check most seriously. It is the reality check. If the four action-oriented numbers all look good but cash is falling, something in my head is wrong about the business and I stop to figure out what. It has caught me twice, both times because I was counting orders that had not actually cleared.

The discipline is treating every number as a question that ends in an action. Repeat buyers ask "make more or fix this?" Replies ask "what do they want?" If you want the wider system this sits inside, Claude Blueprint walks through how the studio decides what to build in the first place.

Bottom Line

Five numbers, ten minutes, one notebook. That is the whole system, and it beats the six-dashboard habit I ran for a year. The trick is not finding better metrics. It is throwing out every number that cannot change what you do next week.

Repeat buyers, email replies, orders per product, support requests, and cash at week end each point at a decision. Follower counts, impressions, time on page, and bounce rate do not, so I stopped opening them and lost nothing. If a number only makes you feel good or bad, it is not data, it is mood.

If you run a one-person operation, try cutting your review down to the numbers that end in a verb. Copy them into one tab, compare to last week, write two sentences, close the laptop. Do it for a month and you will notice your task list getting clearer while your dashboard gets quieter. The studio does not need more information. It needs less, chosen better, and read fast.

This article contains affiliate links. If you sign up through them, I may earn a small commission at no extra cost to you. (Ad)