Analyzing the spatial and temporal dynamics of Snapchat

Levente Juhász and Hartwig Hochmair



June 12, 2018 @ VGI-ALIVE pre-conference workshop, AGILE 2018, Lund, Sweden

Instant messages called "Snaps"

  • Peer-to-peer
  • Disappear upon viewing







© Santeri Viinamäki CC BY-SA

Instant messages called "Snaps"

  • Peer-to-peer
  • Disappear upon viewing






© Jamie CC BY-NC-SA

Snap stories

  • Sequence of snaps

  • Available to view for friends
  • Available for 24 hrs

Snap Map

Location sharing + publicly available snaps

Snapchat by the numbers

  • 190 million active daily users
  • 300 million active monthly users
  • 47% of US teens consider it their main social media
  • More than 20k photos/videos per second

Is Snapchat of interest?

First hand information

Is Snapchat of interest?

Near real-time event detection

Goal: Understand the spatial and temporal characteristics of Snapchat

To see what kind of information can be expected

Data collection

Data collection

    Before February 2018
  • Mobile only app
  • No open API

  • Snap Map available at map.snapchat.com (Feb '18)
  • Can be reverse engineered
  • Public snaps on map can be collected

Data collection

Data collection

  • Heatmap from points corresponding to snaps
  • Regenerated every ~5 mins
  • Individual points visible for 24 hrs

  • Scrape locations continuously
  • Rebuild timeline and keep first occurence of points

  • Miami, New York City and Los Angeles
  • February 23 - March 3

Data collection

February 23 - March 3
Metro area# of tiles (area [sq km])# of snaps
Miami457 (2,860)25,155
Los Angeles1,029 (6,395)77,426
New York514 (3,105)54,243

Results

Spatial characteristics

Prominence of downtown areas with other local clusters

Spatial characteristics

Tile size: ~ 2.4 km x 2.4 km

25% of snaps are posted in 1% of tiles

Spatial characteristics

Gi* statistics

Temporal characteristics

Snap numbers increase by 61-75% over the weekend

Temporal characteristics

Temporal characteristics

Weekday
MiamiLos AngelesNew York
Most active17-1817-1821-22
Least active5-65-65-6
Weekend
MiamiLos AngelesNew York
Most active18-1920-2121-22
Least active8-96-79-10

Summary

Summary

  • Snapchat is used to gather first hand information
  • Strong clusters in downtown and touristic areas
  • Weekend activity > weekday activity
  • Peak: evening and night; Least active: early morning

  • Future work:
  • Longer period, more cities
  • Compare activity to Twitter

Questions?

@juhaszlevi

levente.juhasz@ufl.edu

Need volunteers for other research: research.jlevente.com