Balance is the wrong word, and it quietly causes most of the damage. It frames every disagreement as two legitimate forces meeting in the middle, which turns the conversation into a negotiation between people’s opinions and hands the decision to whoever is most senior or most persistent in the room.
Most apparent conflicts between user needs and business goals are not conflicts at all. They are measurement problems, or they are arguments about time horizons wearing a costume. A smaller number are genuinely zero-sum, and those deserve to be named as such rather than balanced. Doing this well is mostly a matter of telling the three apart before anyone starts negotiating.
The short version
- Do not balance. Classify. Most conflicts are one of three kinds and each has a different resolution.
- False conflicts are measurement artefacts — find the metric both sides already share.
- Time-horizon conflicts are real but solvable with instrumentation, not argument.
- Genuine zero-sum conflicts exist. Name them honestly and escalate them rather than quietly designing them in.
- Whoever owns the shortest-horizon metric wins every unmeasured argument. That is a reporting problem, not a persuasion problem.
- The cancellation flow tells you more about a company’s real position than any value statement on its careers page.
Three kinds of conflict
1. The false conflict
These look opposed and are not. They almost always come from measuring a proxy instead of the thing you actually want. “Users want fewer emails, the business wants engagement” is the standard example: engagement was the proxy, revenue was the goal, and the extra emails were driving unsubscribes that cost more than the clicks earned.
The tell is that nobody can state the business goal without naming a metric. If the answer to “what are we trying to achieve” is a number on a dashboard rather than an outcome, you are probably looking at a false conflict and arguing about instrumentation.
The move: go up one level. Ask what the metric is a proxy for, then check whether the user need and that outcome actually diverge. Often they do not.2. The time-horizon conflict
These are real. The proposal genuinely does improve this quarter and genuinely does cost you later. A harder-to-find cancellation flow reduces churn this month and raises refund requests, chargebacks, support load and the number of people who will never recommend you. A trial that requires a card up front lifts conversion and lowers the quality of who converts.
Both sides are correct about their own horizon, which is why these arguments never resolve by debate. They resolve only when the later cost is measured and attributed, and most SaaS companies measure the gain in a dashboard and the cost nowhere at all.
The move: insist on the lagging measure before the change ships, not after. If nobody will commit to measuring the cost, that is the answer to whether the gain is believed.3. The genuine zero-sum conflict
Sometimes the business really does gain at the user’s expense, permanently and by design. Pricing that penalises the behaviour users most want. Lock-in through making data hard to export. Charging per seat in a product whose value grows with sharing. Using customer data in ways customers would object to if asked plainly.
These exist, they are not always wrong, and pretending otherwise is the least professional available response. A business may legitimately decide to extract. What it should not do is dress that up as a user benefit or leave the decision to a designer in a sprint.
The move: name it in writing, state who gains and who pays, and escalate it to whoever is actually accountable. Your job is to make the decision visible, not to make it for them or to disguise it.Whoever owns the shortest-horizon metric wins every argument that nobody is measuring. That is a reporting problem, not a persuasion problem.
Classify the argument you are currently having
Pick the decision on the table, the horizon you are personally measured on, and whether you can actually see the downstream cost. The third input changes the answer more than the first.
Conflict classifier
Three inputs. One of the outcomes is that you should stop arguing and go build instrumentation.
What is being proposed?
What are you measured on?
Can you see the downstream cost today?
Three inputs, one classification
Pick what is on the table, your horizon, and whether the later cost is visible.
What to do next
—
What evidence settles it
—
0 of 3 chosen
The instrumentation that ends these arguments
Design teams lose these conversations because they bring judgement to a numbers fight. The answer is not better advocacy. It is making the cost side visible on the same dashboard as the gain, so the comparison happens automatically rather than through persuasion.
| The metric being optimised | The lagging cost to measure alongside it |
|---|---|
| Trial-to-paid conversion | Retention of that cohort at 90 and 180 days, and refund rate. A conversion lift that does not survive to day 90 moved a number, not a business. |
| Churn reduction from a harder cancellation flow | Chargebacks, support contacts, review sentiment, and whether those saved accounts use the product at all in the following quarter. |
| Expansion revenue from upsell prompts | Interruption rate per session and task completion on the flow being interrupted. |
| Email or notification engagement | Unsubscribes, mutes and the share of users who have disabled everything. A mute is close to permanent. |
| Signups from a card-required trial | Share of trials that convert passively and never activate, plus the refund and dispute rate those produce. |
| Seats sold | Share of seats that are inactive, and the volume of shared logins — which is your pricing model telling you something. |
| Feature adoption after a forced prompt | Repeat usage in week two. First use after an interruption measures the interruption, not the feature. |
The practical ask is small and nearly always granted: before shipping the change, agree which lagging measure qualifies it and when you will look. Teams that refuse this are usually telling you they do not expect the gain to survive the look — which is itself the most useful information in the conversation.
The cancellation flow, as a worked example
Every SaaS product has one, and it is the clearest single test of how a company actually resolves this question.
The short-horizon case is strong and easy to measure. Add friction, offer a discount, require a reason, bury the button, and this month’s churn number improves. The effect is real and it shows up quickly.
The costs are real too, and arrive on a lag. Some of those retained accounts are now people who wanted to leave and could not, which converts a neutral ex-customer into an active detractor at exactly the moment they are most likely to be asked for a recommendation. Some go to their bank instead, and a chargeback costs more than the subscription. Support absorbs the rest. And in several markets, dark patterns in cancellation flows have moved from a reputational question to a regulatory one, which makes this a legal exposure as well as a design one.
The version that holds up is unglamorous. Make cancelling straightforward. Ask one optional question on the way out. Offer a pause, which is a genuine alternative rather than an obstacle. Confirm clearly and keep their data available for a stated period. You lose some saves and you keep the option of them returning, which in subscription businesses is a real and frequently underrated revenue line.
Note what the honest version is not: it is not the user’s preference winning over the business. It is the longer-horizon business case winning over the shorter one. That distinction is the whole argument of this piece.
How to run the conversation
Classify before you argue. Say out loud which of the three kinds this is. Half the time the act of classifying resolves it, because someone realises they have been defending a proxy metric.
Argue in the other side’s currency. “This harms users” is a weak move in a revenue meeting, not because it is wrong but because it is unfalsifiable in that room. “This lifts conversion four percent and historically costs us eleven percent of that cohort by day ninety” is the same argument in a form that can actually be checked.
Put a number on the user cost, even a rough one. Support contacts per thousand users, task time, the share of people who abandon. An imperfect number beats a well-phrased concern, because it can be tested and it stays in the record.
Escalate zero-sum decisions rather than absorbing them. If it is genuine extraction, your job is to make it visible to whoever is accountable, in writing, with who gains and who pays stated plainly. Then implement the decision they make. Quietly refusing to design it is not integrity; it is just a decision made at the wrong level and never discussed.
Know your own line in advance. Most designers have one — deception, dark patterns in cancellation, misuse of personal data. Decide where it is before you are asked, because deciding under deadline pressure with a job on the line produces answers people regret.
The honest cancellation flow is not the user winning over the business. It is the longer-horizon business case winning over the shorter one.
Where this goes wrong
| Failure | Looks like | Prevention |
|---|---|---|
| Balancing instead of classifying | Every conflict negotiated as a matter of opinion | Name which of the three kinds it is before discussing it. |
| Defending a proxy | The business goal is stated as a dashboard metric | Ask what the metric is a proxy for, then re-check the conflict. |
| Measuring the gain, not the cost | Conversion tracked weekly, 90-day retention tracked never | Pair every short-horizon metric with its lagging measure. |
| Arguing in the wrong currency | “This is bad for users” in a revenue review | Translate the user cost into a number that can be tested. |
| Absorbing zero-sum calls silently | A designer quietly implementing extraction | Escalate in writing, state who gains and who pays. |
| Treating all users as one | A change that helps buyers and punishes daily users | Say which segment gains and which pays. Averages hide this. |
| Design as the ethics department | One person holding the line against a system | Make the trade-off visible in reporting so it stops depending on a person. |
Where to start this month
Take the last three product decisions where someone invoked business goals, and classify each one. You will usually find one false conflict that could be settled with better instrumentation, one time-horizon conflict whose cost nobody is measuring, and perhaps one genuine zero-sum call that was never escalated because it never got named.
Then pick the single short-horizon metric your team is most driven by, and put its lagging measure on the same dashboard. One pairing, visible to everyone, changes more behaviour than a quarter of advocacy — because the comparison starts happening on its own rather than depending on whoever is willing to have the argument.
How we work at Adi Design. On SaaS engagements we classify conflicts before designing around them, pair each short-horizon metric with the lagging measure that qualifies it, and put genuine trade-offs in writing for whoever is accountable rather than resolving them quietly in a sprint. If a decision on your roadmap is stuck in this argument, get in touch.
