Behind the scenes of OnOneMap.com

In 2005 I was involved in creating ononemap.com, a property search site that became very popular (and controversial), and was acquired in 2008 by dotHomes. This is the story of the user experience revolution in property search.

Transcript

2. •  Admire my Powerpoint theme •  Problem •  The OnOneMap solution •  Technology •  Shuffling data •  Cool tech
3. Search Rightmove.co.uk for ‘Richmond’:
4. My concept of Richmond Richmond town centre, between bridge and A316 Rightmove’s concept of Richmond (with help) Richmond upon Thames, borough of London 2.8 sq km 57.4 sq km
5. •  One property per map •  Branded by 3rd party •  Opens in new window
6. •  User experience designed to appeal to agents, not house hunters •  Pre-conceived ‘regions’ and boundaries •  Subjective information given priority –  Who needs to know who the agent is? •  Information trickle •  Location is key...
7. •  Where’s the school? •  What’s the council tax? •  Is it on a flood plain •  Noisy main road? •  Is there open space nearby? •  Is it commutable to work? •  Is it on the ‘right’ side of the railway line? LOCATION
9. Sites  such  as  rightmove.co.uk  and  primeloca6on.com   are  crumbling  ruins  compared  to  www.ononemap.com   Unlike  many  insipid  rivals,  this  one  is  fantas6c   Could  this  be  the  easiest  way   to  find  a  new  home?   Mrs  Website  of  the  Day  is  the  main  property   website  expert  in  our  house,  and  she  was   quite  excited  by  this  site   an  enterprising  property  search  site  
10. •  Google  Maps   –  no  brainer  at  the  1me   –  S1ll  a  good  choice   •  PHP  for  all  searching,  crawling  and  indexing   –  Sta1s1cal  distribu1ons  to  provide  powerful  filtering  at  high   zoom  levels   •  MySQL  with  spa1al  extensions   –  MBR  search  5x  quicker  than  lat/long  range  
11. •  Master-­‐slave  replica1on   •  Main  index  update  during  off-­‐hours   –  User  profile  data  replicated  real  1me   –  Ad  clicks  and  logs  collected  locally  and  aggregated  daily.   •  Cookies  for  mul1-­‐server  sessions   –  Previously  shared  using  real  1me  database  replica1on:   unreliable  and  unscalable  
12. •  SubmiPed  from  700  registered  sources   –  Four  submission  technologies  available   •  Crawled  from  major  sites   •  Links  verified  on  con1nuous  cycle   –  Data  mine  content  to  verify  key  fields   –  Specific  paPerns  developed  for  popular  sources   –  Social  queuing  priority  
13. •  Conflict  resolu1on   •  Clustering   •  Weekly  vs  monthly  rents  (mul1ply  by  4?!)   •  Shortlist/blacklist   •  De-­‐duping  property  records   –  Some1mes  adver1sed  at  different  prices!  
14. •  Phew   •  No  more  bullet  points.  
15. School  search  
17. ‘Nearest  match’  disambigua1on  
19. Mobile  phone  masts  and  flood  data  
25. Proximity  filter  
28. Rename  proper1es  and  add  your  own  notes