20
Uzanto Consulting IA Summit 05 Making Magic Happen: Predicting User Behavior in AJAX applications Jonathan Boutelle Uzanto Consulting

Ajax Prediction

Embed Size (px)

Citation preview

Uzanto Consulting IA Summit 05

Making Magic Happen: Predicting User Behavior in AJAX applications

Jonathan Boutelle

Uzanto Consulting

Uzanto Consulting IA Summit 05

Who am I?

• www.uzanto.com

• www.jonathanboutelle.com

Uzanto Consulting IA Summit 05

Uzanto Consulting IA Summit 05

Uzanto Consulting IA Summit 05

HTML Server

Http request

Initial Html

User action

Html (with data embedded)

Dbase

Vanilla Web App (circa 2000)

User action

Html (with data embedded)

Uzanto Consulting IA Summit 05

HTML

Http request

Initial Html

User action

Html (with data embedded)

Vanilla Web App (circa 2000)

User action

Html (with data embedded)

Server Dbase

Uzanto Consulting IA Summit 05

Rich Internet App (circa 2003)

Http request for app

App (lActionScript and Flash) downloads

User action that requires data Data

UIClient

Manager Server Dbase

Uzanto Consulting IA Summit 05

Rich Internet App (circa 2004)

Http request for app

App (Flash and ActionScript) downloads

User action that requires data

Optionally preload data

UIClient

Manager Server Dbase

Uzanto Consulting IA Summit 05

Ajax Application (circa 2005)

Http request for app

App (HTML + Javascript) downloads

User action that requires data

Always preload data

UIClient

Manager Server Dbase

Uzanto Consulting IA Summit 05

Uzanto Consulting IA Summit 05

Building a model of user behavior

• Build naïve model

• Validate and refine model

• Be metric driven Responsiveness: % user data requests that were met

with pre-fetched data Efficiency: the % of pre-fetched data that ended up being

used

Uzanto Consulting IA Summit 05

Uzanto Consulting IA Summit 05

Preload Cost

the extra cost assumed by preloading:

{cost of download} *

(1-{odds data will be requested})

Uzanto Consulting IA Summit 05

Preload Value

value of reduced latency *

odds the data will be requested

Uzanto Consulting IA Summit 05

Download if value > cost

• Value = value of reduced latency * odds the data will be needed

• Cost = cost of download * odds the data won’t be needed

Uzanto Consulting IA Summit 05Odds the data wll be used 100%0%

value

cost

Value vs. Cost

Uzanto Consulting IA Summit 05

Odds the data wll be used100%0%

value

cost

Increasing Value of Preload

Uzanto Consulting IA Summit 05

Odds the data wll be used100%0%

value

cost

Decreasing Value of Preload

Uzanto Consulting IA Summit 05

Uzanto Consulting IA Summit 05

value

costOdds content will be requested

Importance of low latency