NINE: Why We Didn’t Start by Polishing the Dashboard

While building NINE, the dashboard was the screen most likely to trigger our design instincts. It almost asked to be made beautiful: a cat portrait, a health score, weight trends, symptom alerts, vaccination status, recent activity, a restrained color palette, refined charts, and just enough motion to make the whole experience feel alive.

From a product-presentation standpoint, it was also the obvious place to invest. The dashboard is the first screen many users see. It photographs well, demos well, and quickly communicates a sense of completeness.

But as we continued reviewing the experience in PARDPro Lab, we stopped. We asked a less comfortable question: would a more beautiful health overview actually help an owner understand their cat’s condition?

If the answer was not yet clear, prioritizing visual polish would not necessarily improve the product. It could simply amplify a judgment we had not validated.

Health dashboards have unusual persuasive power. Once different kinds of information are arranged neatly on one screen, users can easily assume that the system has interpreted the data and reached an overall conclusion.

A green health score, a smooth weight curve, a card marked “Good,” several indicators without alerts, and a reassuring summary can collectively communicate something the product never says outright: your cat is healthy.

That implicit conclusion is exactly what concerned us. Visual design is not neutral. Color, position, scale, and motion all shape how reliable information feels. A number placed at the center of the screen, surrounded by a complete ring and colored green, will naturally appear more authoritative than a sentence of plain text.

If the reasoning behind that number is weak, a more refined design does not make it more accurate. It makes the unsupported certainty more convincing.

A health score is one of the most appealing dashboard elements. It is simple, immediate, and satisfies the desire for a quick answer. Weight, symptoms, Body Condition Score, vaccinations, and recent activity can all appear to collapse into a single result: 86—good condition.

But what does that number mean? It might reflect whether symptoms were recently logged, whether weight changed, whether BCS falls within a common range, whether a vaccination is approaching its due date, or whether the owner has recorded information consistently.

Those factors do not carry equal medical meaning. How many points should one mild vomiting episode remove? Does several weeks without a record mean the cat is stable, or that the system simply does not know what happened? Should a cat with stable but persistently unhealthy weight receive a high score? Can an upcoming vaccination and a recent loss of appetite responsibly sit in the same weighted model?

Without defensible answers, the score is not a summary of the cat’s health. It is the numerical output of a set of product rules. It may look precise without being accurate—and users rarely distinguish between those two qualities.

Dashboards can easily turn “no data” into “no issue.” In a health-record product, that distinction is critical.

If the system receives no symptom entries for seven days, it could display, “No symptoms this week.” But all it may actually know is, “No symptoms were recorded this week.” The first statement assesses health. The second only describes the state of the data.

An owner may have been busy and never opened the app. Another family member may have noticed something but failed to record it. A change may still be too subtle to recognize. The absence of an entry cannot prove that everything is normal.

The same principle applies elsewhere. No new weight measurement does not demonstrate stable weight. No BCS entry does not demonstrate normal body condition. Missing vaccination data does not mean vaccinations are current. No uploaded photo does not mean there is no visible abnormality.

A trustworthy dashboard must distinguish among confirmed facts, information the user has not provided, prompts derived from limited evidence, and questions that require professional judgment. Until those boundaries are established, visual completeness only conceals incomplete data.

Weight charts are another easy way to make a dashboard feel professional. A few measurements connected by a line, supported by dates, units, and a percentage change, can quickly suggest whether weight is rising or falling.

But a line is not automatically a meaningful trend. Measurements from different scales may not be directly comparable. A reading taken before a meal may differ from one taken afterward. A unit-conversion error can still produce a perfectly smooth chart. Two points collected six months apart can be connected visually even though the system knows nothing about what happened between them.

The chart may be technically correct while encouraging the user to over-interpret limited evidence.

Before asking whether the curve looks good, we need to ask how many valid data points exist, how much time they cover, whether the record contains long gaps, whether units are consistent, whether the change exceeds ordinary measurement variation, and whether the product should describe the observation or interpret it.

When the evidence is insufficient, an honest statement is more useful than a confident-looking curve: “Only one measurement is currently available, so a trend cannot yet be established.” It is less impressive and more trustworthy.

Dashboards often carry reminders. Vaccinations may be due soon, parasite prevention may need scheduling, weight may have changed, and symptoms may have been recorded recently. These prompts can be useful, but products often keep adding them in an effort to appear intelligent.

The owner can end up seeing a weight decrease, a suggestion to record a new BCS, an upcoming vaccination, a recent vomiting entry, a reminder that the health record has not been updated, an option to generate a veterinary report, and an invitation to upgrade for full history—all at once.

Each item can be justified individually. Together, they leave the user unable to tell what matters most.

A health dashboard needs more than a collection of notices. It needs a clear hierarchy among recording suggestions, schedule reminders, changes worth monitoring, situations that may justify contacting a veterinarian, and ordinary product-feature entry points.

If a commercial upgrade prompt carries the same visual weight as a health concern, the user cannot tell whether the product is caring for the cat or driving conversion. A restrained dashboard can still support commercial goals. It simply does not use health anxiety to achieve them.

We are not opposed to making the dashboard beautiful. Clear hierarchy, comfortable typography, purposeful spacing, and a stable color system all reduce the effort required to understand information.

The issue is that visual design is not decoration applied after the product decisions are complete. Enlarging a number declares that it matters. Coloring a status green suggests that it is safe. Using a red alert tells the user that something is urgent. Animating a module actively competes for attention.

Visual design must therefore follow the judgment model. When the judgment itself has not been validated, mature design makes the wrong information more persuasive.

That is why we did not begin by making NINE’s dashboard more elaborate. We first needed to decide what it should help the owner accomplish. The goal is not to make users think, “This app knows everything.” It is to help them understand what has been recorded, what recently changed, which information is missing, which items are only prompts, and what the most reasonable next action may be.

We initially treated the dashboard as a comprehensive health overview. We now think of it as a next-decision interface.

When owners open the app, they may not need to see every available data point. They are more likely to ask whether anything needs attention today, whether a recent change deserves monitoring, which record has gone too long without an update, whether they should log weight or symptoms next, and whether they have enough information for an upcoming veterinary visit.

This means the first screen should not attempt to contain every feature. It should establish a deliberate order.

The first layer is the most relevant current change, such as sustained weight loss, recurring symptoms, or an overdue vaccination—but only when enough evidence supports that description. The second is data completeness: if recent weight, BCS, or symptom information is missing, the interface should say that the picture is incomplete instead of claiming that the status is good.

The third layer is the next action: record information, review a change, complete the profile, or prepare a report. That action should reflect the user’s current situation rather than exposing every feature by default. Long-term charts and history remain valuable as a fourth layer, but they should not overpower what needs attention now.

Green, yellow, and red are efficient status signals. In health contexts, they can also flatten complex evidence too aggressively. Green usually means safe, yellow means caution, and red means danger—but many records do not justify any of those conclusions.

When information is missing, the state is not green or red. It is unknown. When weight changes but the sample is too small, the result may not be abnormal; it may require continued observation. When an owner records a symptom, the product can preserve its timing and severity without claiming to understand the cat’s overall health.

We believe the dashboard needs room for at least four kinds of state: confirmed normal, worth monitoring, action recommended, and insufficient data to assess.

“Unable to assess” is not a product failure. In a health product, it is part of professional communication.

Dashboards invite animation: numbers count upward, rings fill, curves draw themselves, and cards move into place. These techniques can make a screen feel active, but health interfaces require more restraint.

If a health score animates from zero to 86, users may infer that the system is performing a sophisticated calculation. A pulsing red notice can intensify anxiety. If every card shifts and scales on entry, visual rhythm competes with the information that actually matters.

We prefer motion only when it explains a state transition: a new record joining the timeline, a chart updating after a unit or date-range change, a reminder being acknowledged, or data moving from loading to available.

Those transitions must also become stable, immediate states when the user enables reduced motion. Animation should not make a health condition feel more serious or an algorithm feel more capable than it is.

This may be the most important conclusion from the exercise. Product teams want interfaces to provide answers because answers look valuable and uncertainty can look like limited capability.

But uncertainty is normal in longitudinal health records. Owners may not record consistently. Data may be incomplete. Photos may become unavailable. Measurement conditions may differ. Some changes require professional examination before they can be explained.

If the product removes those unknowns to keep the page looking complete, it will inevitably manufacture false certainty.

We want NINE’s dashboard to say, “Based on the available records, these are the changes we can see.” When necessary, it should also say, “There is not enough information to determine a trend,” and, “These records may be useful to a veterinarian, but NINE does not provide a medical diagnosis.”

This language will not make the product appear all-knowing. It will make the product more worthy of trust.

PARDPro Lab still wants the NINE dashboard to feel bright, restrained, and carefully made. It should have clear hierarchy, comfortable reading rhythm, accurate state colors, and charts that genuinely help people understand their records.

But those qualities must rest on clear data provenance, visible gaps, adequately supported trends, prioritized reminders, commercial entry points that do not masquerade as health warnings, system prompts that are not presented as medical conclusions, and a reason for every form of visual emphasis.

So we did not begin by asking how beautiful the dashboard could become. We first asked whether every judgment on the screen deserved to be designed with that much confidence.

Only after that question is answered does visual refinement become meaningful. Otherwise, a more beautiful dashboard may simply make an immature judgment look more like an answer.