Which line on the P&L you can actually move

The last lesson ended on an uncomfortable claim: your proposal is not competing against doing nothing, it is competing against every other way of growing earnings. That claim is useless until you can say where a design change enters the earnings calculation at all. This lesson makes that concrete, using a real income statement, and it ends somewhere that will feel like a loss before it feels like a gain: most UI work touches the earnings line only through a long causal chain, and the honest move is to say which link you are claiming rather than gesturing at the end of it.

What the buyer is buying

At exit, a software company is generally valued as some measure of earnings multiplied by some multiple. The operating group controls one of those two terms and merely hopes about the other. Bain’s 2026 framing of the industry is that this has become sharply less forgiving: deals that once “required only 5% annual growth in… EBITDA” now “require around a 10% to 12% average annual EBITDA growth” for comparable returns — Bain being a consultancy that sells advisory work to the firms it is describing.

Which makes it worth knowing what EBITDA actually is, precisely, because the word gets used as though it were a fact about a company rather than a construction. It is not an accounting standard. The SEC’s interpretive guidance treats it as a non-GAAP measure and pins the base: under Question 103.01, “earnings” in EBIT and EBITDA means net income as presented in the statement of operations under GAAP, and a measure calculated any other way must be labelled Adjusted EBITDA rather than EBITDA.

That distinction is not pedantry, and it is the first place this course asks you to be suspicious of your own future sentences. If the headline number the whole structure runs on is itself a construction with publicly-policed rules about honest presentation, then a design claim attached to it is at least two constructions deep. Precision is the entry fee.

A real income statement, with the lines named

Portfolio companies do not publish their statements, so use a public one with the same shape. Snowflake reported the following for the twelve months ended 31 January 2026 (thousands of dollars, from its fourth-quarter and full-year results):

Product revenue                          4,472,317
Professional services and other            211,629
  Total revenue                          4,683,946

Cost of product revenue                  1,260,324
Cost of professional services and other    277,481
  Total cost of revenue                  1,537,805

Gross profit                             3,146,141

Sales and marketing                      2,062,137
Research and development                 1,969,472
General and administrative                 549,697
  Total operating expenses               4,581,306

Operating loss                          (1,435,165)

Three things are worth reading off that before going anywhere near a design argument.

Revenue has two kinds. Product revenue and professional services are separated because they behave differently — one recurs and scales, the other is people-hours sold once. That separation is the seed of the ARR-versus-revenue distinction the metric-set module picks up, and it is already visible here as a structural choice about how the business is reported.

Gross margin is the ceiling on everything. 3,146,141 divided by 4,683,946 is about 67%. Every dollar of new revenue arrives carrying only about sixty-seven cents of gross profit, which is why the gross-margin-adjusted version of any efficiency metric differs so much from the unadjusted one.

Operating expense is where the arguments happen. Sales and marketing at 2,062,137 is roughly 44% of total revenue; research and development at 1,969,472 is about 42%; general and administrative at 549,697 is about 12%. Your team’s salary sits inside one of those lines. So does the budget for the thing your proposal would displace.

The four honest routes from a UI change to a line item

There are not many, which is good news — a short list is checkable. A UI change can reach the income statement through:

• Cost of revenue, by reducing what it costs to serve an existing customer. Fewer support contacts, less manual remediation, less human intervention per transaction. This is the route the review gate most plausibly takes.
• Sales and marketing, by making a deal easier to win or an existing customer easier to expand — a demo that does not need a specialist to survive, an onboarding path that does not need a services engagement.
• Research and development, by reducing the cost of building the next surface. Shared components are the classic case, and the portfolio-standardisation lesson is entirely about this route.
• Revenue retention, by removing a reason an existing customer leaves. This is the most valuable route and by far the most dangerous to claim, because retention has many causes and your change is one small candidate among them.

Notice that new-customer revenue is not on the list. It is not impossible — a genuinely differentiating surface can win deals — but a design change claiming new revenue is claiming the longest, most contested causal chain available, and it will be read that way by anyone who has heard the claim before.

There is also an important warning about where support costs actually land: companies classify them differently, and this lesson deliberately does not tell you which line customer support sits on, because the answer is company-specific and this course could not open a filing that stated it plainly for the example above. Ask, in the specific company, before asserting it in a case. Getting that wrong in a room where the finance lead knows the answer costs more than not knowing.

The study you will be tempted to cite, and why not to

Sooner or later somebody will hand you a widely-quoted statistic showing that design-led companies outperform their peers, and it will be tempting, because it appears to settle the argument this whole course says is hard. Do not lead with it, for three reasons.

The first is who publishes such work. The best-known design-ROI index is published by a consultancy’s own design practice — a party that sells design services, producing a finding that design services are valuable. That is not disqualifying, but it is exactly the conflict you would name instantly if a vendor produced a benchmark showing its own category wins.

The second is methodology. A published critique by Charles L. Mauro and Paul W. Thurman argues that the report was never put through peer review and that its findings are, in their words, “both inaccurate and unsupportable” as research. Their central objection is the one that matters for your purposes: an index built on correlation between design practices and financial performance cannot carry a causal claim, however large the correlation. And note the second conflict honestly — Mauro’s firm sells human factors engineering and usability research, so it is also an interested party, arguing about whose evidence for design’s value should count.

The third reason is the one that should actually stop you: even if the finding were solid, it is about a category, not about your change. No portfolio-level correlation tells anyone whether your review gate reduced this company’s cost to serve. Citing it in place of a mechanism swaps a specific, checkable claim for a general, uncheckable one — which is a downgrade, dressed as support.

What this does not cover

This lesson stayed on the income statement and deliberately avoided the metric vocabulary the room actually speaks in — ARR, gross and net retention, CAC payback, the Rule of 40. Those are operating metrics layered on top of these lines, each defined differently by different companies, and they get their own module. Before that comes the portfolio question: what changes when the surface you are proposing is not for one product but for ninety, which is the standardisation lesson and the last one in this module.