How can an app help me if it doesn't know me yet?

PursueUpdated 10 October 20265 sources

Short answer

By asking, briefly, instead of guessing. A 2023 systematic review (Yuan and Hernandez) treats the 'cold start', a new user with no history, as its own problem, and asking the person directly is one approach that works from the first day. Later, fresh answers should outweigh old patterns, because preferences drift (Koren, 2009).

01The cold start

Personalisation quality drops sharply for someone with no history. A systematic review treats this 'user cold start' as a separate problem and maps the ways systems try to solve it, including simply asking the user^[1].

Once there is data, personal baselines help. In a pilot with 28 students, prompts triggered by each student's own first week raised activity in the following hour more than prompts based on a population average, though neither raised activity over the whole week^[2]. Algorithms can learn each person's good moments over time^[3].

02People change

Preferences drift. A model that tracked that drift, instead of treating old data like new, predicted Netflix ratings better than anything before it^[4]. And what a person doesn't do is a noisy signal: something they never saw can't be counted as a 'no'^[5].

03Where the evidence stops

Most of this comes from recommender systems for films and products, plus small health pilots. It supports asking and updating; it gives no rule for how many questions are enough.

Something to try today

When you start using any personal assistant, tell it the one thing it most needs to know about your day in a sentence. It will guess less.

Pursue

How Tre-Up approaches this

In its first days Pursue knows almost nothing about you, so it is designed to ask for a 90-second check-in instead of pretending: energy, focus, time, what's in the way, and the one thing that matters. What you tell it today outranks anything it inferred last month.

Tre-Up is an iOS app in development. Pursue is the AI inside it. This note describes research and design intent, not results from the app. Nothing here is medical advice.

Sources

Summaries are written in our own words from each source's abstract, public record or official text. Follow the link to read the original. How we write these notes

1.Yuan H, Hernandez AA (2023). User Cold Start Problem in Recommendation Systems: A Systematic Review. IEEE Access 11:136958–136977.Affiliation not yet confirmedOur summary doi:10.1109/ACCESS.2023.3338705
2.Ikegaya M, Foo JC, Murata T, Oshima K, Kim J (2025). Using Personalized Intervention Criteria in a Mobile Just-in-Time Adaptive Intervention for Increasing Physical Activity in University Students: Pilot Study. JMIR Human Factors 12:e66750.Shizuoka University · University of Alberta · Heidelberg UniversityOur summary doi:10.2196/66750
3.Liao P, Greenewald K, Klasnja P, Murphy S (2020). Personalized HeartSteps: A Reinforcement Learning Algorithm for Optimizing Physical Activity. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 4(1):18.University of Michigan · IBM Research · Harvard UniversityOur summary doi:10.1145/3381007
4.Koren Y (2009). Collaborative Filtering with Temporal Dynamics. Proceedings of KDD 2009, 447–456.Yahoo! ResearchOur summary doi:10.1145/1557019.1557072
5.Ma H, Xie R, Meng L, Feng F, Du X, Sun X, Kang Z, Meng X (2026). Negative Sampling in Recommendation: A Survey and Future Directions. ACM Transactions on Information Systems 44:1–44.Shandong University · Tencent · University of Science and Technology of ChinaOur summary doi:10.1145/3793855