Cairn

What is the problem you are trying to solve?

It’s the second week of university. Everyone keeps saying you need an internship, but you don’t know what one involves, when to apply or who to ask. You open a dozen tabs, read three contradictory pieces of advice, and close them all. You haven’t done anything wrong. You just can’t see the first step.

That feeling is the problem. The hardest part of becoming someone new is not the work, it’s the first step, and most people stall before they take it because they can’t see the path. Close to one in two university students don’t feel prepared for getting a job, and lack of work experience is a key obstacle (The Conversation). We surveyed 50+ students and professionals. 82% said they feel overwhelmed by the number of goals or directions in front of them, 69% said they find it hard to know what their next step should be, and 69% said that after completing certain goals there is no visible reminder of what they achieved. People with ability and ambition still stay stuck, because nothing shows them where to begin.

Who does it affect?

It affects anyone starting something new without someone ahead of them to follow: students beginning university who want an internship, people switching degrees or careers, and anyone whose goal feels too big to start. Our survey reached people at every stage, and the answers were consistent across them.

What is your idea?

Cairn is a game-style roadmap that turns a goal that feels out of reach into a road you can see and walk, one step at a time.

It starts with you. Users first create and customise their own character, so the journey feels like theirs before any goals are added. They then choose the category that fits them right now: Student, Career or Personal. Within it, they choose their specific goal and add a deadline. A student might choose “secure a summer internship”. Each goal becomes its own road, laid out as a flowchart of levels, each made of small steps. The next step is always highlighted, and some steps are optional. The character moves along the road as steps are completed. Someone working on several goals has one road for each and can toggle between them. A hub lets them tick off steps and see a progress bar for each level. When someone finishes a road, they can share it as a template, and anyone stuck can ask someone a step ahead for advice, so the next person starts further along. That is how we pass it on.

How does it fix the problem?

Research on the intention-behaviour gap explains why people stall before they start. Good intentions alone rarely become action: across 422 studies, intentions explained only about 28% of the variance in behaviour (Sheeran, 2002). The Fogg Behavior Model adds that behaviour happens when motivation, ability and a prompt come together at the same moment, and when it doesn’t happen, at least one of them is missing. Each feature below supplies one or more of these.

Create your character

Before anything else, users create and customise a character that travels along their roads. Customising a player character increased identification with it, and that identification was linked to feeling more competent and having more fun (PMC study). The tool feels like theirs from the first screen.

Choose who you are, then your goal

Users pick the category that describes who they are (Student, Career or Personal), then their specific goal within it. Student covers school and university, Career covers working or changing career, and Personal covers goals outside study or work. Goals that fit a person’s own interests and values lead to more persistent effort (Sheldon), so every road begins from what the user actually wants. This also cuts through the overwhelm of endless directions: 82% of our respondents said they feel overwhelmed by the number of goals or directions in front of them.

A road of levels and steps

Each goal becomes a road of levels made of small steps, with the next step always highlighted and some steps optional. The highlighted step is the clear, simple next action and the prompt to start now. Each step gets a deadline and a “when and where” prompt, because plans that spell out when and where had a medium-to-large effect on goal attainment (Gollwitzer and Sheeran). In our survey, 93% said breaking a goal into smaller deadlines helps them stay on track, and 94% said levels or clear progress stages help them stay motivated.

Progress bars and a road that starts already begun

Every road starts with the first step already complete, and each level has a progress bar as the character moves along. In a car wash experiment, 34% of customers given a pre-stamped card redeemed the reward versus 19% with an empty card (Nunes and Drèze). Progress shown visually also energised people more than numbers (Cheema and Bagchi). In our survey, 69% said that after completing certain goals there is no visible reminder of the achievement, and 86% said seeing their progress laid out visually helps them stay motivated.

Switch between your roads

Someone working towards several goals has one road for each, and can toggle between them from the hub, so each goal keeps its own clear path instead of blurring together. 86% of respondents said that when working towards several goals at once, having them all organised in one place helps them stay on track.

Share your road and ask someone a step ahead

When someone finishes a road, they can share it as a template, and anyone stuck can ask someone a step ahead or ask an AI. People changing careers move forward through small wins and by finding role models (Ibarra). As 56% of our respondents said they have found it hard to find someone who has already taken the path they are trying to follow.

Built for everyone

Progress is never shown by colour alone, each road has a text description for screen readers, there is no CAPTCHA at sign-up, and the character can be skipped. About one in 12 UK men and one in 200 women have a colour vision deficiency, and CAPTCHA is the barrier screen reader users report most often.

What it means for the person using it

Imagine that second-week student again. They create their character, choose “Student”, and set the goal of securing a summer internship. Their first step is already done, and the next one is highlighted: update your CV by Friday. Within weeks, they can see how far they’ve come along the road, and a goal that once felt out of reach has a clear route. When they finish, their road helps the next student start further ahead. We built the tool we wished we’d had.

References:
The Conversation, work placements: https://www.theconversation.com/university-is-expensive-how-a-mid-course-work-placement-can-help-with-costs-and-careers-203709
Sheeran (2002), intention–behaviour gap: https://cancercontrol.cancer.gov/brp/constructs/implementation_intentions/goal_intent_attain.pdf
Fogg Behavior Model, Stanford: https://behaviordesign.stanford.edu/resources/fogg-behavior-model
Avatar customisation study (PMC): https://pmc.ncbi.nlm.nih.gov/articles/PMC8765231
Sheldon (2014), goal selection: https://www.x-mol.com/paper/1212947096468791296
Gollwitzer and Sheeran (2006): https://kops.uni-konstanz.de/entities/publication/2e749bfb-8533-437c-8203-7e788c910c5f
Nunes and Drèze (2006): https://siliconcanals.com/t-car-wash-loyalty-cards-endowed-progress/
Cheema and Bagchi: https://news.vt.edu/content/news_vt_edu/en/articles/2011/08/082311-research-goalvisualization.html
Ibarra, Working Identity: https://flora.insead.edu/fichiersti_wp/inseadwp2004/2004-21.pdf

How do all the pieces fit together? Does your frontend make requests to your backend? Where does your database fit in?

The frontend was built using HTML, CSS and JavaScript, with separate pages for the homepage, avatar customisation, roadmap and user hub. We designed the homepage to introduce Cairn and guide users through the experience. Users can create and customise their own character, choosing different appearances and accessories to make their journey feel personal.

The user hub brings the different features together, allowing users to view their roadmap, track completed milestones through a dynamic progress bar and return to the avatar editor to update their character. We also developed a wardrobe page to support future character customisation features.

By separating the application into individual HTML, CSS and JavaScript files, we were able to develop different components simultaneously and connect them through page navigation, creating a cohesive and interactive user experience.

For the the purpose of the demo, prototype is implemented as a collection of static Html pages linked together.

For the University Roadmap generator, the user enters their goal and optional details such as their intended course, year group, and entry year, which is passed to a JS script that makes a call to the Gemini API (using the gemini-3.5-flash-lite model) which searches the web, generates a roadmap with timestamps in a structured JSON output, which is then passed back to the frontend and presented in a timeline format with a dynamic progress bar and interactive milestone "blocks" within a matter of seconds.

1. Choosing a model
Initially, we used gemini-3.8-flash, which returned responses after 2 minutes, which greatly slowed our programming and debugging process. We worked through it by researching and trialing other models and APIs (such as other Gemini Models, OpenAI models and Exa) to find the one with the lowest response times, and eventually settled on the gemini-3.5-flash-lite model. Even though it is a lightweight model, we decided that it was sufficient for what we were generating (short titles, dates and descriptions).

2. Role distribution

Because of the timing of tasks, some of us ended up waiting on others, and sometimes two people were working on the same thing. Once we spotted this, we reorganised how we split the work, finding new tasks and dividing them differently so everyone could keep moving.

3. Timing and expectations
We expected things to come together more quickly than they did, which put pressure on our schedule. We handled this by scoping the design to fit the time we had, focusing on what we could realistically deliver well rather than trying to build everything.

1. Running a successful survey
We gathered over 50 responses by reaching out across multiple platforms, including LinkedIn. This let us reach the demographic we wanted, a broad range of people from students through to those already in their careers, and gave us real evidence to base our decisions on. 73% said they often feel overwhelmed by the number of foals in front of them, which is why every goal in Cairn becomes its own road. 92% said breaking a big goal into levels or steps makes it feel more achievable, and 87% find knowing their next step helpful, so each road is made of levels and steps with the next one highlighted. 83% said having several goals organised in one place helps them stay on track, which led to toggling between roads, and 81% said seeing their progress laid out visually keeps them motivated, which led to the progress bar on each level. 88% said smaller deadlines help them stay on track, so each goal has a deadline and the steps within it. 56% said they have found it hard to find someone who has already taken the path they are following, which led to sharing roads and asking someone a step ahead.

2. Building a working website
We built a working website that allows the user to create an avatar, enter their goal and additional details, generate a custom roadmap (using the Gemini API) and view their progress via the dynamic progress bar or the roadmap timeline.

3. Teamwork
We worked well together from the start, reaching agreement quickly and communicating openly. We learned that this kind of easy collaboration kept us moving, even when plans and timings changed.

Features we would add

• Rest days and flexible roads. Users can take a rest day, and adjust steps and deadlines without penalty, so a busy week doesn’t break their road. People are more motivated when they have choice (Patall, Cooper and Robinson), and Duolingo uses streak freezes so a missed day doesn’t erase progress.

• A calendar that builds itself. An expandable calendar fills automatically from the deadlines already on each road, so users see their long-term goals without entering anything twice. Start-of-term and new-month dates motivate new beginnings (Dai, Milkman and Riis), and planning the “when” turns intentions into action (Gollwitzer and Sheeran).

• Social roads. Users can share their finished roads, follow each other’s, and see how someone a step ahead got there. Studying a worked example helps beginners more than starting from scratch (Sweller), and social interaction strengthens behavioural outcomes in gamified learning (Sailer and Homner).

• As our prototype is a collection of static html pages, it was not possible to do the typical thing of having our Gemini API key stored in a .env file so that it would be hidden from bots crawling Github for vulnerable API keys. Our current solution has the user enter a Gemini API key into an input field, but if we were to deploy the actual app, we would use Node.js and hide our Gemini API key in a .env file.

• When deploying, instead of having to use the paid Gemini API to accommodate large amounts of users, we could create and train our own model for generating timed milestones based on search data.

How we would grow it

Start with a pilot at one university, beginning with first-year students. Universities have a reason to try it: in a Jisc survey, 46% of careers services named student engagement as their top challenge, and many students don’t actively engage with employers until their final year. Prospects found 44% of first years attended a careers event, versus 51% of finalists.

Next, universities, careers services and employers publish official roads, such as “Securing a summer internship” or “Getting into [company]”. This matches how existing platforms work with institutions: RateMyPlacement is trusted by 112 universities and works with 210 student employers, and Forage shows employers will build content for students.

After that, we widen beyond internships to other first steps, such as school leavers and career changers, through the Career and Personal categories.

How we would improve it

• Live help from someone a step ahead. We would let users reach someone further along their road in real time, matched by goal and experience. Be My Eyes shows how well this model can work: each request goes to several volunteers, usually 10, and the first to accept is connected, so help arrives quickly and no single person is a bottleneck. We would build our own version, with verified accounts and safeguards before opening it up.

• Roads that fit the person. We would ask a few quick questions about their experience and how much time they have, and shape each road to match. Duolingo asks learners their level and daily time commitment before the first lesson, and Noom builds its whole plan around the user’s goal.

• Timely nudges. We would add gentle reminders at the moment a step is due. Behaviour happens when motivation, ability and a prompt come together (Fogg), and plans that say when and where are far more likely to be followed (Gollwitzer and Sheeran).

• Your story so far. We would add a view that looks back over everything a user has completed. Ibarra’s research on career change describes making sense of what is happening and arranging new events into a story of who we are becoming. In our survey, 79% said looking back at what they have completed motivates them to keep going.

• Testing with real users. We would test with real screen reader users and refine from what we hear. VoiceOver is the most popular mobile screen reader (70.6%) and NVDA the most commonly used on desktop (65.6%) in WebAIM’s survey, so we would test on both.

Monetisation

Cairn is free to use, so anyone can take their first step.

• Free. Users can have up to five roads active at once. Every completed step earns a stone, and a cairn is a pile of stones that marks a trail, so each one is proof someone has walked this way. Users spend their stones on vehicles and on accessories for their character and vehicle. Stones are earned by completing steps and are never needed to finish a road, so progress is never locked behind a payment. Customising a character increased how strongly players identified with it (PMC study), and progress shown visually energised people more than numbers (Cheema and Bagchi). Duolingo uses the same loop: learners earn gems by completing lessons and spend them in a shop.

• Cairn Pro. A paid plan for people working on more at once: more than five roads at a time, and double stones on every step, so unlocks come faster. Duolingo takes a similar approach, with a free tier, gems earned in the app, and a paid Super plan that adds perks.

• Pathway extension. The larger opportunity is for universities, careers services and employers to pay to publish official roads, such as “Securing a summer internship”, while students follow them for free. Handshake runs on this kind of three-sided model, connecting students, universities and employers. We would test it in the pilot before building it fully.

References:
Patall, Cooper and Robinson (2008): https://pubmed.ncbi.nlm.nih.gov/18298272/
Duolingo streak freezes: https://uxdesign.cc/duolingo-analyzing-all-engagement-and-retention-techniques-3e73b79120cf
Duolingo gems, shop and Super plan: https://blog.duolingo.com/duolingo-101-how-to-learn-a-language-on-duolingo
Dai, Milkman and Riis (2014): https://faculty.wharton.upenn.edu/wp-content/uploads/2014/06/Dai_Fresh_Start_2014_Mgmt_Sci.pdf
Gollwitzer and Sheeran (2006): https://kops.uni-konstanz.de/entities/publication/2e749bfb-8533-437c-8203-7e788c910c5f
Sweller, worked example effect: https://repub.eur.nl/pub/61680/art-3A10.1007-2Fs10648-010-9134-7.pdf
Sailer and Homner (2020): https://opus.bibliothek.uni-augsburg.de/opus4/frontdoor/index/index/docId/109056
Prospects Luminate, student engagement (Jisc survey): https://luminate.prospects.ac.uk/how-to-improve-student-engagement-with-careers-activities
Prospects Luminate, first and second year students: https://luminate.prospects.ac.uk/how-employers-can-engage-with-first-and-second-year-students
RateMyPlacement: https://RateMyPlacement.co.uk/about-us
Forage: https://career.kzoo.edu/?p=3593
Handshake: https://sp-edge.com/companies/202894
Be My Eyes: https://community.bemyeyes.com/language/english
Duolingo onboarding: https://www.reallygoodux.io/blog/duolingo-user-onboarding
Noom onboarding: https://www.revenuecat.com/blog/growth/web-to-app-onboarding-funnel
Fogg Behavior Model, Stanford: https://behaviordesign.stanford.edu/resources/fogg-behavior-model
Ibarra, Working Identity: https://flora.insead.edu/fichiersti_wp/inseadwp2004/2004-21.pdf
WebAIM Screen Reader Survey #10: https://webaim.org/blog/screen-reader-user-survey-10-results/
Avatar customisation (PMC): https://pmc.ncbi.nlm.nih.gov/articles/PMC8765231
Cheema and Bagchi: https://news.vt.edu/content/news_vt_edu/en/articles/2011/08/082311-research-goalvisualization.html