prteek/user-journey-dau

Simulated daily active users of an app

667,236 records of daily user activity in an app: 51,480 users active between 1 October 2020 and 31 October 2023. Each row is one day on which one user was active. The data is simulated, based on historical activity data of a SaaS app; it contains no real users.

It is the data behind the paper Probabilistic Model of User Journey on alphaXiv, which places users in seven states each day (new, current, reactivated, resurrected, at-risk weekly active, at-risk monthly active, dormant) following Duolingo's growth model, estimates a Markov chain of transitions between them, and forecasts daily, weekly and monthly active users. Code: github.com/prteek/probabilistic-model-of-user-journey.

Two subsets:

• default (dau_data.csv): the activity records.
• states (states.parquet, 32,111,109 rows, 225 MB): every user on every day from 1 October 2020 to 31 October 2023 with their state, as built from the activity records by prepare_states.r in the code repository. Building it takes a while and about 2.5 GB as CSV, so it is shared here ready-made.

default: activity records

column	content
user_id	anonymous user id (UUID)
date	a day on which the user was active (2020-10-01 to 2023-10-31)
registration_date	the day the user registered (2020-04-20 to 2023-10-31)

Days without a row for a user are days on which the user was inactive.

states: one state per user per day

column	content
user_id	anonymous user id (UUID)
date	the day (2020-10-01 to 2023-10-31)
is_active	whether the user was active that day (true on the 667,236 days in the activity records)
state	the user's state that day (below)
registration_date	the day the user registered

state	meaning	rows
new	active for the first time	50,175
current	active today and in the past week	585,129
reactivated	active today, last active in the past month but not the past week	16,847
resurrected	active today, last active more than 30 days ago	15,085
at_risk_wau	not active today, active in the past week	604,449
at_risk_mau	active in the past month but not the past week	1,683,057
dormant	inactive for at least 30 days	29,156,367

import pandas as pd
states = pd.read_parquet("hf://datasets/prteek/user-journey-dau/states.parquet")