PredictAble

PredictAble helps people with mobility issues see when their day will require a lot of movement and plan ahead with simple adjustments.

Comment

Inspiration

PredictAble was inspired by a simple but common problem: many people do not struggle with their entire day, they struggle with specific windows of the day. Back-to-back classes across campus, long hallways before meetings, crowded transitions, and low-energy periods can make a schedule that looks manageable on paper feel overwhelming in practice.

We wanted to build something that helps users plan before those moments hit. Instead of generic productivity advice, PredictAble focuses on practical, personalized support for movement-heavy transitions and high-friction time blocks.

What it does

PredictAble forecasts difficult parts of a user’s day and turns that forecast into clear, low-effort actions.

  • Syncs and summarizes daily context (events, activity trends, patterns, places, and rough schedule flow)
  • Detects likely high-strain windows and transition bottlenecks
  • Lets users confirm/refine what the model inferred (e.g., “yes, this time window is harder”)
  • Captures personal barriers and support preferences
  • Produces actionable suggestions such as:
    • leave earlier
    • add buffer time
    • simplify route complexity
    • insert short recovery gaps

In short, it helps users move from “reacting in the moment” to “planning ahead with confidence.”

How we built it

We built PredictAble as a web app with a guided onboarding-first experience.

  • Frontend: Next.js + React for fast iteration and route-based flows
  • Design system: Reusable UI primitives (cards, buttons, chips/toggles) for consistency and accessibility
  • Onboarding engine: Multi-step flow that:
    1. introduces the experience
    2. confirms synced context
    3. refines inferred patterns
    4. captures user constraints and priorities
    5. shows a forecast preview
  • Forecast logic (current phase): Rule-based heuristics that combine schedule timing, transition density, known difficult windows, and user-selected barriers/preferences to rank recommendations
  • Architecture approach: Built around modular components so we can later swap rule-based scoring with learned models without rebuilding the UX

A rough framing we used internally: [ \text{Friction Score} \approx w_1(\text{time pressure}) + w_2(\text{route difficulty}) + w_3(\text{transition clustering}) + w_4(\text{user barriers}) ]

Challenges we ran into

  • Interpretable personalization: Balancing “smart” suggestions with explanations users can trust
  • Signal quality vs. user burden: Asking enough questions to personalize recommendations without making onboarding exhausting
  • State complexity: Managing many onboarding steps and conditional branches while keeping UX smooth
  • Demo realism: Creating representative sample data that feels believable and useful without requiring full production integrations
  • Build/runtime stability: Debugging cache/chunk issues in development while actively iterating on route and component structure

Accomplishments that we're proud of

  • Built an end-to-end onboarding journey that feels focused and human-centered
  • Converted abstract “difficulty forecasting” into practical, concrete actions users can take immediately
  • Created a clear narrative from data sync -> validation -> personalization -> forecast
  • Established reusable UI and flow patterns that make the product easy to extend
  • Shipped a usable prototype that communicates the core value of PredictAble quickly

What we learned

  • Personalization needs transparency. Users trust recommendations more when they can see why a suggestion appears.
  • Micro-decisions matter. Small choices (5-10 minute buffers, route simplification) can have outsized impact on daily strain.
  • Good onboarding is product logic. The onboarding sequence is not just setup; it is where confidence and data quality are created.
  • Constraints are features. Designing for real-world friction (not ideal schedules) makes the product more relevant and empathetic.
  • Shipping early reveals truth. Prototype feedback surfaced UX and logic gaps faster than planning alone.

What's next for PredictAble

  • Add live integrations (calendar/activity/location) with robust consent and privacy controls
  • Improve forecasting from rule-based heuristics to hybrid ML + rules while keeping explanations user-friendly
  • Introduce adaptive recommendations that learn from user follow-through and feedback
  • Expand accessibility features and personalization depth (fatigue patterns, environment context)
  • Build longitudinal outcomes: weekly trends, progress tracking, and recommendation effectiveness
  • Run pilot testing with target users to validate impact on stress, punctuality, and perceived effort

If you want, I can also rewrite this into a tighter Devpost-style submission version (more concise, more punchy) or a judge-facing version (impact + technical depth + metrics placeholders).

Built With

Share this project:

Updates