How to measure the ROI of design.

Design ROI feels slippery because design touches everything and owns almost nothing on the income statement. A new flow can lift sales, but so can a price change, a better ad, or a holiday weekend. So design gets credit when things go well and blame when they do not, and almost none of it is measured. That is a bad deal for design, and it is a bad deal for the business.

I have led and done design for 15 years, including an AI-first transformation inside a multi-billion-dollar eCommerce design org. The thing I have learned is simple. You can measure the value of design honestly if you decide what you are measuring before you start, and if you refuse to overclaim. This is a practical how-to for doing exactly that.

Why design ROI feels hard, and how to make it concrete

Most design ROI conversations fail for one reason. The work shipped, the number moved, and nobody set up the measurement ahead of time. Now you are arguing about a chart after the fact, and everyone has a story. Stories are not proof.

The fix is to make the work concrete before you touch a pixel. Pick one metric the business already tracks. Write down where it sits today. Decide how you will tell your change apart from everything else happening that week. Do this first, and ROI stops being a debate. It becomes a before and after with a clear cause.

Concrete also means small. A giant redesign is almost impossible to measure because you changed a hundred things at once. A focused change to one screen, one step, or one piece of copy is easy to measure because you can point at the thing that moved.

Pick the business metric before you start

This is the most important step, and it is the one people skip. Before you design anything, name the number. Not a design number. A business number that finance and your executives already care about and already report.

Here are the usual ones, in plain terms. Conversion is the share of people who buy or sign up. Activation is the share of new users who reach the first real win. Retention is the share who come back. Time to value is how long it takes someone to get the first good result. Support load is how many tickets or refunds you generate. Pick one of these as your primary number. Just one.

Then ask the obvious question. If this design works, which of these moves, and by when? If you cannot answer that, you are not ready to build yet. You are still guessing. Naming the number first also keeps you honest later, because you committed to the scoreboard before you knew the score.

When I rebuilt The Mysterious Package Company on Shopify, the number was conversion, and we ran weekly A/B tests against it. When I led design at Supply Drop, a predictive subscription business, we shipped monthly homepage revisions and let data pick the winners. In both cases the metric came first and the design came second. That order is the whole trick.

Attribute the change to design honestly

Honest attribution is where most ROI claims fall apart. You need a way to say this change caused this result, not this change happened near this result. There are three tools, and they stack.

First, a baseline. Write down the metric before you ship. A week or a month of clean before-data, depending on how much traffic you have. Without a baseline you have nothing to compare to, and every claim becomes a feeling.

Second, an A/B test when you can run one. Show the old version to half the people and the new version to the other half at the same time. Same week, same traffic, same ads. The only thing different is your design. That is as close to proof as most teams will ever get, because the test controls for the holiday weekend and the ad change and the price cut. They hit both groups equally.

Third, isolate the variable. Change one thing at a time, or as close to one as you can manage. If you redesign the page and rewrite the copy and add a new offer all at once and conversion jumps, you genuinely do not know which part did it. You feel great and you have learned nothing. Split the changes across tests so each result points at a single cause.

When you cannot run a clean A/B test, say so out loud. A before and after without a control group is still useful, but it is weaker, and you should label it that way. The honest version is, conversion rose after we shipped this, and no other big change went live that week, so we believe design drove most of it. That sentence is defensible. We tripled sales is not, unless you can show the test.

Value-mapped metrics versus vanity metrics

A value-mapped metric connects to money or to a thing the business already decided is money-adjacent. A vanity metric looks like progress and changes nothing on the income statement. Knowing the difference protects you from celebrating noise.

Time on page is usually vanity. People can linger because they are delighted or because they are lost, and the number cannot tell you which. Page views are vanity when they are not tied to anything downstream. A nicer satisfaction score is nice, but on its own it is soft.

Conversion, activation, retention, average order value, refund rate, and support tickets are value-mapped. They either are money or they push money directly. The test is simple. Ask, if this number goes up, does the business make or keep more money? If yes, it is value-mapped. If you have to tell a long story to connect it to money, it is probably vanity. Build your ROI case on the value-mapped ones and let the soft numbers be supporting color, never the headline.

Leading and lagging indicators

Some numbers move fast and some move slow, and you need both. Leading indicators move within days. Add-to-cart rate, step completion, sign-up rate, the click that starts a flow. They tell you early whether the design is working, while you can still react.

Lagging indicators move over weeks or months. Retention, repeat purchase rate, lifetime value, churn. These are the numbers executives trust most, because they are hard to fake and they tie straight to the business. The catch is you cannot wait a full quarter to find out if a button worked.

So use leading indicators to steer and lagging indicators to confirm. Ship the change, watch the leading number this week to decide whether to keep iterating, and watch the lagging number next month to confirm the value held. When I shipped monthly homepage revisions at a subscription business, the leading signals told us fast which version to keep, and the retention picture confirmed weeks later whether the win was real or just a sugar high. Report both, and label which is which, so nobody mistakes an early signal for a settled result.

Talk about ROI in the language of finance

Designers lose the room by talking about craft when the room wants money. Finance and executives think in revenue, cost, risk, and time. Translate your work into those terms and the conversation changes.

Do not say the new checkout is cleaner and more on-brand. Say the new checkout lifted completion, which at current traffic is worth this much more revenue a month, and we proved it with an A/B test. Do not say we improved onboarding. Say more new users now reach their first win, activation rose, and activated users retain better, so this protects revenue down the line.

Support load is an easy win here because it is pure cost. If a clearer design cuts confused tickets and refunds, that is money kept, and finance feels it immediately. Always pair the result with the method. We saw this lift, here is the test that proves it, here is what we did not change. Executives trust people who show the receipts and name the limits. They stop trusting the person who only ever brings good news with no math behind it. If you want the longer version of how I build that case end to end, I wrote it up in how I prove design moved the business.

When design value is real but not cleanly measurable

Some of the most important design work does not show up in a clean A/B test, and pretending it does will wreck your credibility. Brand, trust, and long-term retention are real, valuable, and slow. You cannot run a tidy two-week experiment on whether people trust you more.

The honest move is to say so. Split your work into two buckets and label them. One bucket is measurable, with a metric, a baseline, and a test. The other bucket is judgment-based, where you believe the work builds trust or brand strength over time, and you are not going to fake a number to prove it. Being clear about which bucket a project sits in is what makes people believe you on the measurable ones.

You can still bring evidence for the soft bucket. Track proxies over time, like repeat behavior, referrals, branded search, or how customers describe you in their own words. Treat those as directional, not as proof. The Mysterious Package Company lived on trust and atmosphere, and at Story Genie, where I let people see the entire finished book before they pay, the trust is the product. That kind of value is real even when no single test captures it. Say it plainly. Here is what I measured, here is what I cannot measure cleanly, and here is why I still believe it matters. That honesty is worth more than a confident number you made up.

One more honest limit. Small samples lie. If a test ran on a handful of conversions, do not crown a winner. Say the sample is too small to act on yet, and keep it running. Calling noise a result once will cost you trust on every result after.

Build the habit, not just the report

ROI is not a slide you make at review time. It is a habit you run every cycle. Name the number, write the baseline, ship the change, run the test, read the result, and say honestly what you proved and what you did not. Do that on a regular rhythm, weekly or monthly, and the proof accumulates on its own. You stop arguing about whether design pays off because you have a stack of clean before and afters that answer the question for you.

Most teams do not need a fancier framework. They need someone who installs this loop and holds the line on honest attribution. If you want help standing it up, that is exactly the job of a design leader who installs this habit.

The mistakes that wreck a design ROI case, and how to avoid them

Most design ROI stories do not fall apart because the design was weak. They fall apart because the case around it was sloppy. I have watched strong work lose a budget fight because the person presenting it overreached on one number, and a finance partner stopped believing the rest. So before you build your case, walk through the common traps. Each one is easy to make and easy to avoid once you name it.

The first trap is taking credit for a number you cannot attribute. Revenue went up, the redesign shipped that quarter, so the redesign did it. Maybe. But a price change, a new ad push, a seasonal swing, or a sales hire could have moved the same number. The fix is simple. Only claim what you can tie to the design with a clean line, and say out loud what else was running at the same time. Naming the other forces makes the part you do claim more believable, not less.

The second trap is changing ten things at once and then claiming the win. You redesign the page, rewrite the copy, add a new offer, and speed up the load time, all in one release. Conversion rises. Now you cannot say which change earned it, and neither can anyone else. When I rebuilt The Mysterious Package Company on Shopify, the discipline that saved us was running weekly A/B tests, one clear change at a time. The fix is to isolate changes when the stakes are high, so the result points at one cause you can defend.

The third trap is cherry-picking the one metric that went up while ignoring the ones that dropped. Sign-ups rose, so you lead with sign-ups. But refunds also rose, or support tickets climbed, or the people who signed up never came back. A number out of context is a half-truth, and finance smells it fast. The fix is to report the metric next to its guardrails. Show what you wanted to move and show the things you did not want to break, even when one of them moved the wrong way.

The fourth trap is calling a tiny sample a result. Forty visitors and a two percent lift is not a finding. It is noise wearing a result costume. I have seen confident decisions built on a handful of sessions, and they tend to reverse the next week. The fix is to wait for enough data that the pattern holds, and to say plainly when a number is early. An honest small sample is a hypothesis worth watching, not a win worth banking.

The fifth trap is measuring after the fact instead of before. You ship, you like how it looks, and then you go hunting for a number that makes the work look good. That is backward, and people can tell. At Supply Drop we picked the metric first, shipped monthly homepage revisions, and let the data choose the winner. The fix is to write down the business metric and your expected result before you build. A prediction you made in advance is worth ten you found afterward.

The last trap is overclaiming so badly that finance stops trusting you. One inflated number, and your whole case is suspect for a year. Trust is the real currency here, and it is slow to earn and fast to spend. The fix is to claim a little less than you could defend, and let the results catch up to your story rather than the other way around. Be the person whose numbers hold up under questions, and your next case gets the benefit of the doubt before you even open your mouth.

Pick the number before you build, prove the change with a real test, and tell the truth about what you cannot measure. That is the whole method, and it is the only version that survives contact with finance.

I have run this loop across The Mysterious Package Company, Supply Drop, Kinjo, Surf Studios, Enverus, and now Story Genie. The businesses and the metrics changed every time. The discipline did not. Decide what matters, measure it cleanly, and never claim more than the data earned. Do that, and design stops being the thing nobody can put a number on. It becomes the thing with the clearest receipts in the building.