Unit Economics: The Discipline of Close, Every Day, Every Unit

Who this is for. You should already recognize the P&L (income statement), balance sheet, cash flow statement, accounting equation, and double-entry bookkeeping. You do not need to be an accounting specialist. For a refresher, start with James Hearle’s Accounting Stuff accounting-equation refresher, then return to the worked example.

Public sources and fictional examples. Company-specific facts come only from cited public materials. Every scenario, workflow, cutoff, rate, volume, allocation, and profitability amount is invented for teaching and must not be read as a description of any company’s internal practice. The interpretations are my own and do not represent my employer or any other organization. No confidential, proprietary, or nonpublic information is used.

Why we need these numbers in the first place

In Why Unit Economics? The Next Decision Needs More Than an Average, we followed the purpose: connect the books to decisions about price, customer promises, processes, and capacity. A profitable company total can hide very different economics underneath. The useful unit view explains those differences while reconciling to the larger result.

This series takes the next step. How do we compute that view accurately enough to guide daily decisions, while keeping its definitions, estimates, and revisions tied to accounting? At scale, that requirement adds substantial engineering complexity to an accounting problem that was already difficult.

If you are looking for the reason to do all this work, begin with that companion essay. Here, we work through the discipline needed to trust the answer.

Month-end close was already hard

“Month-end close is easy,” said no accountant ever.

Closing a large organization means coordinating records, cutoffs, estimates, reconciliations, consolidation, and review across people and systems. Even a reviewed result contains judgments. Credibility comes from making those judgments supportable and the differences explainable.

Monthly close is covered in Month-End Close: Five Questions Before You Trust the Profit. This series carries that discipline into daily unit economics.

Now shrink the reporting boundary from an organization to a ride, reservation, retail order, completed task, or day of subscription service. Ask for a useful result tomorrow morning. Repeat across a very large population while invoices, refunds, claims, and corrections continue arriving.

Computing trustworthy unit economics is, in this sense, a daily unit-level close. It adds substantial engineering complexity to an accounting problem that was already difficult.

This is an analogy about discipline, not a claim that every transaction needs a statutory close. We need relevant recognition rules, completeness checks, supported estimates, reconciliation, and review at a much finer grain. We do not create a legal entity, full balance sheet, or audit for every click.

Five things to remember

Keep five prompts in working memory: Unit. Timing. Missing amounts. Shared costs. Revisions.

Remember	Ask before trusting the margin	What makes it hard
Unit	What exactly are we measuring, and from whose perspective?	An order can have several shipments; a viewing event is not a separately sold subscription.
Timing	Which service date and reporting period own each amount?	Booking, payment, delivery, and settlement happen at different times.
Missing amounts	What happened but has not reached our records?	Supplier bills, refunds, claims, and support work can arrive later.
Shared costs	Which resources did the unit consume, and what remains shared?	Activity drivers are imperfect; idle capacity and corporate costs do not disappear.
Revisions	What changed since yesterday’s answer, and why?	New evidence, corrected errors, and changed models must remain distinguishable.

These questions connect directly to close’s period, completeness, treatment, evidence, and review. They are a small mental map, not a claim to cover every accounting rule or business model.

How this follows allocation and activity-based costing

Allocation Is Life begins with choices about finite resources. Allocation Becomes a System follows those choices into organizational rules and software. Abundance Is an Allocation Problem asks what better choices make possible.

Activity-Based Costing asks which work consumes those resources. Month-end close asks whether the reported result is supported. Daily unit economics combines both questions and adds a demanding engineering requirement: preserve the meaning of the calculation across many units, every day, as evidence changes.

Company profit divided by transaction count is an average. It does not tell us which customer promise created the work or which decision would actually change the cost.

An event is evidence about a unit

A purchase can create a payment authorization, two shipments, dozens of scans, a delivery confirmation, and a refund. Those events describe one commercial relationship. They are not dozens of independent sales.

A system must connect the records, recognize repeats, and distinguish a reversal from another charge. It must also detect missing records: a perfectly deduplicated dataset can still be incomplete.

A useful daily run has a defined service-day boundary and an information cutoff. For example, it might report services completed by local midnight using evidence received by 06:00 the following morning. The timezone, cutoff, source completeness, and rule version belong with the result. These are design assumptions in our examples, not disclosed company routines.

If records arrive after that cutoff, the system needs a controlled update. It should preserve both what managers knew when they decided and the later, better-supported view.

What a daily run actually does

A practical run establishes the eligible units, attaches known amounts and estimates, reconciles source totals and cost pools, reviews exceptions, and publishes a version that can later be revised. Operations owns evidence about service completion; finance owns recognition and estimation policies; engineering owns reliable execution and traceability. Review concentrates on exceptions and consequential model changes, rather than a human signature for every unit.

If a material cost feed is absent, the affected segment must either be withheld or published with a visible provisional label under an approved estimation policy. Record the missing coverage and its owner. A successful software job does not by itself mean a successful close.

A margin earns credibility in stages

Consider a fictional unit with $8 of company revenue and $3 of immediately recorded cost. Its first visible contribution is $5.

The daily review adds $1.20 of expected risk costs, $0.40 of support activity, and $0.20 of infrastructure cost. The supported daily estimate becomes $3.20. Later evidence raises total costs by $0.50; the updated contribution is $2.70.

None of these is company net profit. Each is a contribution measure after its stated cost boundary. Corporate expenses, financing, tax, and other omitted amounts need their own treatment and reconciliation.

The sequence matters: $5 visible → $3.20 estimated with missing costs → $2.70 after later evidence. A dashboard that shows only the latest answer conceals how the organization reached it.

Scale changes the control problem

Imagine one billion relevant records over a reporting period. A one-cent error repeated across them creates a $10 million distortion. That is illustrative arithmetic, not an assertion about any company’s activity volume.

More often, errors concentrate in a particular path: split shipments, cancellations, a promotion, a currency, or a failed payment retry. A total can reconcile while costs sit on the wrong units.

Engineering must therefore establish that the population is complete, each amount is counted once, records join to the right unit, rules are reproducible, and unresolved differences are visible. Running the same input again should not create another sale or another expense. Recomputing a million records should not erase the version that informed yesterday’s decision.

At scale, people review exceptions, model changes, reconciliations, and patterns. They cannot approve every event by hand. A system needs thresholds for both aggregate amounts and decision-sensitive segments: a small corporate difference can still reverse the apparent margin of a small service line.

Five businesses, five different difficulties

Example	Unit in the teaching model	The difficulty we follow
Uber	Completed ride	Insurance and service costs can remain estimated after the ride ends.
Airbnb	Short-stay reservation	Booking, cash, check-in, and payout belong to different clocks.
Walmart	Owned-inventory retail order, built from item deliveries	Split fulfillment and returns change both revenue and inventory economics.
Taskrabbit	Completed task	Customer collections contain provider entitlements as well as platform fees.
Netflix	Subscriber-day within a subscription period	Revenue follows service availability; viewing and shared content consumption are different measures.

Each essay follows a fictional case through the initial view, daily close, later adjustment, reconciliation, and a decision. The examples are independent, not a comparison of the companies’ profitability. None uses confidential employer information.

What this helps you decide

A business leader can ask whether a delivery promise, promotion, or service category earns enough to support the work it creates. An engineer can see why event identity, late data, and version history are financial controls. An accountant can see why a daily analytical estimate needs an explicit bridge to the ledger. An investor can ask whether apparently attractive growth relies on missing costs or fragile allocation assumptions.

There is no universal margin for every choice. An incremental decision needs the money and capacity that actually change. A service redesign also needs shared resources and long-term capacity. A customer acquisition decision needs a cohort over time.

Better unit economics does not automatically make better decisions. It gives those decisions evidence that can be questioned, reconciled, and improved.

Start with the ride. The passenger has arrived. Its economic story is still unfolding.

---

Uber · the completed ride →