Transformation · Workbench
Your process ran without trouble last month. That tells you very little about next year. What matters is what happens to the work when 100 requests a month become 1,000. Software engineers describe this with Big O notation, and the same idea applies to any process a business runs. Answer three questions before you read on.
Guess first
Question 1 of 3 · Tap an answer
A finance team matches each incoming payment to its invoice by reading down the list of open invoices until the team finds the right one. With 100 payments and 100 open invoices a month, matching takes about 2 hours. Next year both numbers grow to 1,000. How long will matching take?
Question 2 of 3 · Tap an answer
A company onboards new suppliers by email. A coordinator spends 3 hours on each supplier, and nothing needs to be built. A self-serve supplier portal would take 120 hours to build and then 20 minutes of review per supplier. How many suppliers does the company need to onboard before the portal saves time?
Question 3 of 3 · Tap an answer
The finance team from the first question buys software that makes each check ten times faster. The team keeps the same method of reading down the invoice list. Volume then grows tenfold again, to 10,000 payments a month. Compared with the 200 hours at 1,000 payments, how long does matching take now?
The model
When the volume of a process grows, the work in that process grows in one of five shapes. Engineers name the shapes with Big O notation, shown under each name below for anyone who wants to look the terms up.
The shape matters more than the speed. A slow process with a gentle shape will eventually cost less than a fast process with a steep shape. Big O notation ignores the speed on purpose: a method that takes 3 steps per item and a method that takes 100 steps per item are both O(n), because both double when the volume doubles.
The examples marked “From the reel” are the figures from the Instagram reel this page is based on, where the creator ran each algorithm and counted the steps.
The work stays the same whatever the volume. A published returns policy answers the same question for every customer, and a rule set once in a system applies to every order.
From the reelAsking a computer for item 50 in a list takes one step, whether the list holds 100 items or 100 million.
The work grows by one step each time the volume doubles. A triage form where each question halves the remaining options works this way, and so does a filing system sorted by customer number.
From the reelBinary search found one name among a million in 20 comparisons.
Double the volume and the work doubles. A person who keys in every invoice by hand, or an agent who answers every support ticket from scratch, does straight-line work.
From the reelSearching a million usernames one at a time took 987,413 checks, and doubling the list doubled the checks.
Each item is compared or coordinated with every other item, so ten times the volume means a hundred times the work. The payment matching in the first question is every-with-every work, and so is a decision that needs everyone in a group to agree with everyone else.
From the reelComparing every user with every other user took 10,000 checks for 100 users and 1,000,000 checks for 1,000 users.
Each new option doubles the number of combinations. A product where customers can switch 20 features on or off has 1,048,576 possible versions to price, support and test.
From the reelAt a billion steps a second, an explosive method with 100 items would take about 40 trillion years. The universe is 13.8 billion years old.
Engineers use a sixth shape, O(n log n), for sorting. Sorting grows a little faster than straight-line work and far slower than every-with-every work. That gap is why sorting or indexing a list first is often the cheapest way out of every-with-every work.
Try it
Each bar shows how long the work takes if one step takes one minute of someone’s time. The bars use a log scale, because on an ordinary scale the flat and slow-growing bars would be too short to see.
Where this bites
Each example names the shape, what happens as volume grows, and a redesign that changes the shape.
A group of 5 people has 10 pairs who may need to talk, and a group of 10 has 45. Brooks (1975) counted these pairs for software teams: a group of n people has n(n−1)/2 of them. A decision that needs every pair to agree gets slower with every person added.
Redesign: name one person who decides, and consult the rest.
When every ticket needs an agent, the support team grows in step with the customer base.
Redesign: a help article written once is flat for every customer whose question the article answers, so agents are left with the questions the articles do not cover.
A spreadsheet where each row searches the other rows for a match runs about a million comparisons at 1,000 rows and about 10 billion at 100,000 rows. A file that opened instantly last year can take minutes this year.
Redesign: sort the data or join the tables on a key, so each row finds its match directly.
Every optional feature doubles the number of versions that sales must quote, support must understand and quality assurance must test.
Redesign: offer three or four standard bundles and price everything else as a custom project.
Suppose each new hire spends an hour with every existing team member. On a team of 10, each new hire costs 10 hours of the team’s time, and on a team of 50, each new hire costs 50 hours. Growing the team from 10 to 50 people costs about 1,200 hours of onboarding.
Redesign: give each new hire one buddy and a recorded walkthrough of the team’s work.
Your turn
Work through the six steps for one process you are responsible for. Your answers stay in this browser, and you can print the plan at the end or save the plan as a PDF.
Pick a process that you expect to handle more volume in the next three years, such as invoice matching, onboarding or customer requests.
An item is whatever the process handles one at a time: an invoice, a ticket, a new hire or a request. Estimate the number of items a month today and the number you expect in three years.
When one more item arrives, what does someone have to do?
Choose a shape and enter both volumes to see how the work will grow.
The moves that change the shape appear here once you choose a shape in step 3.
A redesign usually costs time to set up. Enter the minutes each item takes today, the hours the new way takes to set up, and the minutes each item will take afterwards.
Enter the three numbers to see when the new way pays back its setup time.
Write one action small enough to start this month, the person who owns the action and the date the action will happen.
Use it this week
Pick the process your team expects to grow fastest, and fill in the worksheet with the people who do the work. If the shape is every-with-every or explosive, plan the redesign now. A faster tool will buy a few months, but the work will catch up as volume grows. If the redesign removes steps, use the Simplification Playbook to check where the work of each removed step goes.
A faster tool changes how long each step takes, but only a new method changes how the work grows.
Judging how a process will behave at ten times its size is one part of thinking clearly about digital work. The Digital Intelligence course builds the rest of that skill one day at a time.