Guide · What to measure
Almost every company that moves goods measures the same things. So do the ones that sell to other businesses, among themselves. Not because they copy each other: when two businesses work the same way, the same things break, and they end up watching the same numbers.
Here are those numbers, grouped by how your business works rather than by your industry. Each one with what it measures, how it is calculated and the exact question you would put to Sapioverse. And at the end, the awkward part: what you need to have stored for the question to be answerable.
It is not a hunch, it is a consequence. If you hold stock, you have money sitting on a shelf and you need to know how many times a year it empties. If you sell on credit, you need to know how long you take to get paid. If you live off the customer coming back next month, you need to know how many do not. The company does not pick the indicator: the mechanics of the model impose it.
And this is not us saying it. The supply chain reference model half the industry uses describes its processes as industry neutral, and states in writing that every company will have to go one level deeper to add its own. The only international standard that defines operations indicators declares that it uses "the most generic terms possible, rather than terms specific to one sector". And in Spain, the FMCG association publishes a recommendation whose stated aim is to fix a common set of indicators shared by suppliers and distributors.
The name matches, the formula does not. That is the real failure, and the most expensive one. You do not have to imagine it: two retail chains filed their annual accounts the same year and both published "sales per square metre".
Sales per square metre
Chain A
Average floor area at year-end. Stockroom and back office included. Week 53 excluded.
Chain B
Weighted daily average. Selling floor only. Week 53 counts.
Both warn in writing, in that same document, that their number may not be comparable with anyone else’s.
The same goes for the rest. Days sales outstanding has six named calculation methods, and the advice from whoever compiles them is to pick one and never change it. The Spanish service-level recommendation allows full deliveries to be measured by units or by order lines, which give different numbers. The European and US regulators have given up: instead of defining these indicators, they require every company to publish its own definition.
There are three more limits, and it pays to keep them in view. The standard moves too: the supply chain model has rewritten its headline indicators three times in twenty years, and of the ones it had in 2003 only three are still on today’s list. Measuring something deforms it: the moment an indicator becomes somebody’s target, lead time improves by pushing out the date you commit to, and turnover improves by buying less and running out of stock. And two companies that look alike from outside can need different things: if you compete on price your number is cost per unit shipped, and if you compete on service it is the percentage of complete orders.
The fifth limit is the one you feel most day to day: an indicator tells you whether you are doing well, it does not tell you what to do. Low turnover does not tell you whether you bought too much or sold too little. You find that out by asking again, and again. It is exactly why a fixed dashboard falls short: it answers the first question and not the second.
The honest conclusion: what is universal is what to measure. What belongs to each company is how it is calculated. That is why a list like this saves you the easy part and not the hard one, and why the number a tool gives you is only worth something if that tool knows the criteria you calculate it with.
The four in the common block apply to everyone. The rest depends on this.
Common block
These four do not depend on how you sell. They come out of the arithmetic of any business that invoices.
What is left of each sale after paying for what you sold. It is how you discover that what bills the most is not what earns the most, which happens more often than you would think.
(revenue − cost of goods sold) ÷ revenue
Which product families earned the most margin last year, and in which ones has margin fallen even though revenue went up?
How much of your revenue rides on very few customers. It is what decides which call you take yourself instead of the sales rep.
revenue from the top N ÷ total revenue
What percentage of my revenue depends on my ten biggest customers, and how has that percentage changed over three years?
How many days pass on average between invoicing and the money being in the account. It is the difference between selling and getting paid, and it explains why a good year can leave you short of cash.
(accounts receivable ÷ credit sales for the period) × days in the period
How many days do I take on average to get paid, and which customers go past the terms they agreed?
Whether the movement is yours or the calendar’s. It tells a real drop apart from the August one.
each month against the same month last year, never against the previous month
Compare this year’s revenue month by month with last year’s.
If you move goods
The percentage of orders that went out complete and on the committed date, both conditions at once. It is the indicator your customer notices: if it drops, you find out through a complaint.
orders delivered on time AND in full ÷ total orders
What percentage of orders went out complete and on the committed date last month, and which customers account for the ones that failed?
A precision that heads off an argument: this is not the same as the "perfect order" in the supply chain standard, which also requires correct paperwork and undamaged goods. Two companies that say they measure the same thing may be measuring two.
The same thing, but line by line. An order with twenty lines and one unfilled counts as a failure in the indicator above, and that sometimes overstates it. It tells you whether the problem is general or comes down to three specific items.
lines shipped in full ÷ lines ordered
Which products can I least often ship in full when they are ordered?
How many times a year the warehouse empties and refills. It is about idle money: each extra turn is cash released without selling a euro more.
cost of goods sold for the year ÷ average inventory value at cost
How many times did I turn the warehouse last year, and which families turn below the average?
The same idea as turnover, said in days, which is how a purchase actually gets decided. It is the number that opens the conversation with the supplier.
current stock ÷ average daily sales over the last N days
How many days of sales do I have left in each family at the pace of the last three months?
The opposite of the above: what you are holding that does not sell. It is money you have already paid, and it unlocks the clearance, the promotion or dropping the item.
items with zero movements in the last N months and stock greater than zero
Which items have gone more than six months without a single movement, and how much money do I have tied up in them?
The times there was demand and no product. It is the sale that shows up in no sales report, because it never happened.
lines unfilled for lack of stock ÷ total lines
Which products hit zero while orders were pending, and how many times did it happen this year?
How long from the order coming in to it going out the door. It separates "we are slow" from "the carrier is slow".
average of (dispatch date − order entry date)
How long do I take on average to prepare an order once it comes in, and has it got worse than last year?
What proportion of what goes out ends up coming back. It nearly always points to a catalogue or picking problem, not a customer one.
units or amount returned ÷ units or amount shipped
What percentage of what I ship comes back, and which products account for the returns?
If you sell to other businesses
How many buy from you today and how many used to and have stopped. The customer going cold gives no warning: they stop ordering, and you find out at year-end.
customers with an order in the period; lost ones are those who had one before and do not now
Which customers bought from me last year and have not placed a single order this year?
The days since each customer’s last order, set against what is normal for that customer. A customer who orders every three weeks and has gone seven is a call for this week, not for the quarter-end.
days since the last order, against that customer’s historical average frequency
Which customers have gone longer than usual for them without placing an order?
How many different families each one buys from you. That is where the cheapest sale there is lives: the second family to a customer who already buys from you.
number of distinct families per customer and period
Which high-volume customers buy only one product family from me?
What each customer leaves you after the discount you give them and the cost of what you ship them. It is how you find the big customer who is not profitable, usually the one making the most noise.
(amount invoiced − cost of goods sold) ÷ amount invoiced, per customer
Which customers are in my top twenty by revenue and below average on margin?
How much is being given away, by sales rep, by channel and by product. There is no possible conversation with a rep without this number in front of you.
1 − (average price achieved ÷ list price)
What average discount is each sales rep applying, and who falls outside the usual range?
How many of your customers someone has seen, and how many not. Crossed with recency there stop being two problems: a customer with no order and no visit is a single one.
customers with at least one visit in the period ÷ assigned customers
Which customers has nobody visited this quarter, starting with the biggest by revenue?
The average amount of each order. Small orders cost the same to prepare as big ones, and that is where the operations margin goes.
revenue ÷ number of orders
What is the average order value by channel, and what percentage of my orders falls below the threshold it costs me to prepare them?
Whether a territory sells what it should for the number of customers it has. It tells "that province is small" apart from "that province is being worked badly".
territory revenue ÷ active customers in the territory, against the average
Which provinces bill less per customer than the national average?
If you sell to the end consumer
How much each shopper spends and how many things they take. It is the way to grow without bringing in more people, which is the expensive part.
revenue ÷ number of transactions, and units ÷ number of transactions
How much does a customer spend on average per purchase, and how has it changed over the year?
What percentage of the people who come in end up buying. If you buy traffic or pay rent on a good street, this is the number that says whether you are making it count.
purchases ÷ visits, or transactions ÷ entries if you count store footfall
What percentage of visits ends in a purchase, and how does it change by day of the week?
How many of those who bought from you have bought again. It is the indicator that separates a shop with regulars from a shop with visitors.
customers with two or more purchases ÷ customers with at least one
What percentage of those who bought from me last year have bought again this year?
How much of what you brought in you have already sold, and how fast. It is the decision to mark down, and above all, when.
units sold ÷ units received, per item and season
What percentage have I already sold of what I bought for this season, by family?
How much margin every euro tied up in stock gives back. It puts margin and turnover into a single number, which is why it is the one that decides the assortment: the family with good margin that does not turn stops looking good.
gross margin for the period ÷ average inventory value at cost
Which families give me back the most margin for every euro I have invested in stock?
What the space yields. It lets you compare stores of different sizes without fooling yourself, and decide where each category goes.
revenue for the period ÷ square metres of selling floor
Which stores sell less per square metre than the average, and since when?
It is the indicator on this list whose definition varies most between companies, as seen above. It works perfectly to compare you with yourself. To compare yourself with someone else, first ask what they put inside it.
What each first-time buyer costs you, by acquisition channel. It is what lets you stop splitting the budget on instinct.
channel spend for the period ÷ new customers attributed to that channel
How much did a new customer cost me on average through each channel last quarter?
An honest warning: this can only be answered if the order records where the customer came from. If it does not, the data is not in your database, it is in the ad platform.
What proportion of what is sold comes back, and of what. In online sales, a high rate concentrated in a few items is nearly always a product page problem, not a customer one.
amount returned ÷ amount sold
What percentage of what I sell is returned, and which ten products account for the returns?
If you bill every month
How many customers leave in a period, and how much they take with them. In this model retaining is worth more than acquiring, and it is what gets looked at least.
customers lost in the period ÷ customers at the start. And the version that hurts: the same sum in euros
How many customers cancelled last quarter, and how much did they bill between them?
What you have committed for next month, and where the change comes from. A flat month can hide plenty of sign-ups and plenty of cancellations: the net figure misleads.
active recurring amount at period end, broken down into new, upgrades, downgrades and churn
How much recurring revenue do I have today, and how much came in from new customers and how much left through churn this quarter?
How much margin a customer leaves, on average, before going. It is the ceiling on what you can spend to acquire one: without this number, any acquisition budget is a bet.
average margin per customer and period × expected average lifetime, which comes out of the churn rate
How much margin does a customer leave me on average from joining to leaving?
How long you take to recover what it cost to bring the customer in. It is the number that decides whether you can step on the gas: if it takes eighteen months, growing leaves you short of cash.
acquisition cost ÷ monthly margin that customer leaves
In how many months do I recover on average what it costs me to acquire a new customer?
How much of your growth comes from new customers and how much from the long-standing ones spending more. Growing on new customers alone is expensive and fragile.
change in revenue from the customers already there at the start of the period
Of what I have grown this year, how much comes from new customers and how much from the ones I already had buying more?
This is the part almost no list of indicators tells you. Sapioverse does not make the numbers up: it takes them from your database. If the data is not there, the honest answer is that it is not there, and that beats an invented number.
| You need | With which fields | Without it, there is no |
|---|---|---|
| A sales line table | date, customer, product, quantity, amount | Nothing. It is the basis for everything |
| The cost of what you sell | unit cost or cost of goods sold | Margin. Revenue only |
| A customer master | channel, territory, sign-up date | Breakdowns by channel, territory or tenure |
| A product master | family or category | Any grouping by assortment |
The second one is the one missing most often. Without cost you can answer "how much have I sold" and you cannot answer "how much have I earned", which is nearly always the real question.
It pays to keep them apart, because they are fixed in different ways.
The data is missing
You do not store the committed date
It gets fixed in the system that generates it. Until it is, that indicator does not exist for anyone, with or without Sapioverse.
The rule is missing
Nobody has written down what counts as an active customer
The data is there, but the criteria live in the head of whoever has been doing the numbers for years, and nowhere else. Whether a credit note comes off the month it is issued or the month of the original sale. Which family groups which.
That second part is what you load into Sapioverse as business context, and it is what makes the answer come out with your criteria and not a textbook’s. What the tool cannot do is invent a rule nobody has decided yet: if the rule does not exist, it has to be decided first. Where each of those layers lives is covered in the data architecture guide.
You do not need everything in place to start asking. With a sales table and the cost of what you sell, three of the four already come out: margin, customer concentration and the change against last year. The fourth, days sales outstanding, also needs your accounts receivable.
Connect your database and ask in your own language. The query that runs is always in plain sight, so you can check where every number comes from.