By Adam Kelleher (Sr Data Scientist, BuzzFeed) BuzzFeed has developed the technology to attribute pageviews to a referring user. Using these data, we can construct diffusion graphs for our articles. These graphs introduce a whole collection of new performance metrics, and their complexity opens the door for a new assortment of complications to go with them. I'll mention some past work that has been done to create similar data from old pageview events. Then, I'll work through how we process these data into graph objects (avoiding some pitfalls), and mention some of the new ways of looking at web analytics implied by these objects. I'll talk about how we can take advantage of the structure of these objects to make certain algorithms more efficient. Finally, I'll cover some of the future applications we're particularly excited about!

2. HISTORY OF VIRALITY

4. THE DATA

5. THE DATA: OLD VERSION
Article being viewed
User viewing article
Time of pageview
Referring domain

6. THE DATA: NEW VERSION
Article being viewed
Time of pageview
Referring domain
User viewing article
Referring User

7. DIFFERENT PERSPECTIVE:
Pageviews are a process on a graph!

8. WHAT THE GRAPH
LOOKS LIKE:

9. WHAT THE PROCESS LOOKS
LIKE:

10. WHAT THE DATA LOOKS LIKE:

11. WHAT CAN DO YOU WITH
OLD PAGEVIEWS?
(Educated)
Guess!

12. CONNIE

13. OLD GRAPH RECONSTRUCTION:
MODEL-BASED INFERENCE
Probabilistic: You can infer connections that aren’t
there!
Error Prone: Graph statistics can be susceptible to
small changes in the graph
Gets larger when differences in
pageview times gets smaller

14. SIMPLIFIED VERSION:
Observe:
Guess:

15. SIMPLIFIED VERSION:
Guess:
Reality:

16. Check out a toy implementation here!
github.com/akellehe/pyconnie

17. NEW GRAPH RECONSTRUCTION:
TRIVIAL
These are
actually
Unique
Visitors …

18. LIFE IS A LITTLE
MESSY…
This is
more like
what the
Pageview
graph
looks like

19. PROBLEM: DATA MUNGING
• Lots of potential for heuristics!
• How do we get promotion attribution
from propagations?
• Trees are important: how can we be
sure we get them?

20. PROBLEM: STREAMLINING
ANALYSIS
• How do we work from a common set of definitions?
• How do we avoid repeating analysis?
• How can we streamline data visualization? EDA?
• How do we share optimized analyses? And avoid
inefficient (but correct) algorithms?

21. DEFINE DATA
STRUCTURES!
• All data munging happens “under the hood”
• Data pre-processing is unit-tested
• No room for heuristics: standardization!
• Hard math definitions can be consistency-checked!

22. PROPAGATION SET
For one article
For the site (or other set of articles, S)

23. PROPAGATION SET
Pageviews to article b
in time T
Pageviews to the site
in time T
The simplest data structure. Just a
representation of the raw pageview logs.
Represented as a generator of UserEdge objects

24. PROPAGATION
GRAPH,

25. PROPAGATION GRAPH

26. PROPAGATION GRAPH

27. INFLUENCE GRAPH
Propagation graph together with a map,
That measures the influence of the origin user in p
on the pageviewing user

28. CONSIDER:

29. PROPAGATION
FOREST

30. PROPAGATION FOREST
The propagation graph is great, but we’d also like a
concept like unique visitors!
If there is attribution ordering in the graph, we can
trace content back to its source!

31. PROPAGATION FOREST: FIRST
PARENT ATTRIBUTION
n pageviews One UV

32. PROPAGATION FOREST
gets the credit

33. RESULT: ALL GRAPHS
ARE FORESTS
Promotions have 0 indegree,
Users have 1 indegree
total edges in connected components:
Trees!

34. CAREFUL FOR EDGE
CASES: MISSING DATA?
All connected components should be rooted at a
promotion source.
What happens if we lose the first edge (e.g. use the
wrong T)?

35. PROPAGATION FOREST:
CYCLE BREAKING
Consider … Cycle is not broken by
first-parent attribution
Traversal algorithms go
on forever!

36. PROPAGATION FOREST:
CYCLE BREAKING
Consider …
As long as they’re not
equal, the can be
ordered, say
Then, there is a node in the
cycle with an out-edge
younger than its in-edge:
The original pageview for
that node must have been
lost. Cut the in-edge
(FPA!).

37. SUCCESS!
Cycle-breaking + FPA = Trees!
Each tree is the UV graph downstream from a
promotion source: promotion attribution!
Additional Benefits:
Most information diffusion analyses model trees growing on
graphs.
Many algorithms simplify when run on trees!

38. SUPERTREE
We may want to run an algorithm, or calculate a tree
statistic from a whole forest, instead of just one
tree. How can we do that?
Merge all the roots (promotion sources) together into
one “super-node”
The whole forest becomes a SuperTree!

39. SUPERTREE: EXAMPLE

40. SUPERTREE: EXAMPLE

41. APPLICATION:
LARGE SCALE
DATA VIS

42. WHY IS IT SLOW?
Layouts often consider repelling each
node from every other:
time complexity
Good for a few thousand nodes

43. OPENORD: SIMULATED
ANNEALING
Linear main layout
Quadratic settling Phase
Implemented in Gephi

44. OPENORD
Good for ~10k Users
Slow for ~100k Users
Messy! (if you skip
the quadratic step!)

45. TAKE ADVANTAGE OF
TREE STRUCTURE!
Traverse the tree to decide where to place nodes!

46. H3 LAYOUT Each parent is in the center
of a hemisphere.
Children are laid out on the
surface of the hemisphere
They become centers of
smaller hemispheres (if
they’re parents)
Etc.

51. A NEW IMPLEMENTATION
pip install pyh3

52. WITH D3

53. MORE
APPLICATIONS

54. ATTRIBUTION
Instead of

55. CASCADE PREDICTION

56. GRAPH AND TEMPORAL PROPERTIES
ARE IMPORTANT!

57. TEST THE INFLUENTIALS
HYPOTHESIS

58. IMPROVE CONTENT
TARGETING

59. FINDING THE CAUSES OF
VIRALITY
Consider Fitting a Model:
User Features, content features,
context features, User pair
features

60. UNDER CONSTRUCTION:
Online Regression!
Real-time feature weights tell which features
correlate with propagation probabilities!
Drives hypothesis-building!

61. THE TEAM