Case Study BBCi Search

1
Case StudyBBCi Search

• Why search is not a technology problem

2
me.

• What do I do?
• I work for the BBC
• Q -Whats your perception of what the BBC do?
• I work in the New Media dept.
• Information Architect?
• Interference Architect!
• Previously
• Creative director at Sapient London
• Creative director / launch designer of BBC News
  Online, London Times and Sunday Times
• Even further back
• Wasnt an IA I was just an A
• What am I going to talk about?
• Why search is not a technology problem.

3
Some context
4
Search is strategic to us.

• business imperative

5
Henley centre study online life
6
Project context

• The BBCi initiative

7
Pre-BBCi geocities
8
User perception
9
BBCi perceptual model aimed for
10
technology /content context

• Existing, disparate search engines
• Muscat
• All around the site in different implementations,
  different user interfaces
• Often bolted-on
• Autonomy
• News and Sport
• heavy duty research driven use.
• Not possible to search all BBC web content.

11
Business advantages of shared search

• Shared development means we could invest in a
  long-term shared solution
• Research and resulting improvements shared by all
• Possible Licensing advantages/economies

12
User advantages of shared search

• Able to search across all of the BBC for the
  first time
• One common experience at the core of any search
  within the BBC

13
How we started
14
Project process
15
Market research driven positioning

• We currently over-serve a 20-45 male techno-savvy
  audience
• Want to reach newbies, mainstream consumers,
  especially young mums
• Our market research told us they often used the
  search engine on their default homepage (as set
  by ISP)
• Or often Ask Jeeves
• moms -)

16
persona driven design

• Introducing Gayle Steve

17
Use-models
18
use-models
19
use-models
20
use-models
21
Which turned into
22
use-models
23
use-models
24
early encouragement!!!
25
Building taxonomy from user-centred seeds

• Limited time and resources to build a taxonomy
• Focussed on analysing the most entered search
  terms from our logs as a starting point to give
  most value.

26
Taxonomy tool
27
Taxonomy tool top level screen
Top level of the hierarchical tree structure
taxonomy. Shows all our subject areas plus a test
node for playing around and a xyzxyz node which
reflects our most popular query nothing! i.e.
"". We couldn't have an entry for nothing so we
had to equate a search for no string with
something to give us a node. We then had to have
that node as top level node because otherwise it
would inherit context from its parent.
28
Taxonomy tool node level screen
Shows screen for Question of Sport which has a
parent - BBC programmes - and children - the
presenter's names. Each node also has urls and
synonyms attached. Each url has a label
(title), a description, score (crude method to
indicate which url is most relevant to the node).
Each url has the possibility of having an expiry
date which can be reviewed and the url can be
archived after expiring.
29
Taxonomy tool synonyms
Shows a node with numerous urls attached and
scored Plus numerous synonyms The children are
all aspects of WW2 often these are based on
programme content on BBC TV or radio (e.g.
documentary on Monte Cassino being shown,
therefore people likely to enter as a search term)
30
Taxonomy building process
31
Taxonomy/editorial team
32
UI design and testing

• Round and round and round

33
Iterative design development

• 5 rounds of user-testing and iterative design
  over a 2 month period
• We probably could have cut it down if we had done
  more research upfront
• We started getting diminishing returns
• Avoiding Non-designer Kneejerk
• Great to invite marketers and managers to tests,
  gets them bought into UCD brilliantly
• Manage their breakthroughs sleep on it!

34
What we tested paper prototypes
35
What we tested
36
What we tested
37
What we tested
38
What we were looking for

• The search is supposed to allow users to
• Zoom in (refine search)
• Step sideways (to another BBC content area
  scope)
• Zoom out (widen to BBCi or Web scope)
• Do users understand these options? Can they
  navigate through them? Do they want them?

39
Key findings
40
Best links area
Key findings

• The title was clear. The www icon was better. But
  still not perfect.

A site that's not to do with the BBC. Not run by
them. It would be nice to see the whole name
www.NIassembly.gov.uk
User no. 4 , Pippa
41
Best links area
Key findings

• The position of the icons confused one user
  (novice).

Don't know what they are doing there. You'd
expect to just click on the words on the left and
get what you want. Is it just a link to return to
the BBC easily?
User no. 4 , Pippa
42
Zooming in
Key findings

• Some users understood that searching is a
  dialogue

Finding the right search word is very
frustrating. I think one thing, but the computer
thinks differently.
User no. 4 , Pippa
43
Zooming in
Key findings

• going from general to specific.

I'd expect to be asked to be more specific. It
would prompt you type in more words -- recipes?
flowers? boat show? Whatever
User no. 5, Laura
44
Zooming in
Key findings

• So encourage dialogue be polite and positive,
  however general the initial query was

It would need to make you feel that you got it
right, even though you weren't specific
enough. It should make you feel you're on the
right track.
User no. 2 , Gillian
45
Zooming in or stepping sideways?
Key findings

• Users tended to view the specific sections
  links as a way to zoom in.
• They thought about the links in that section in
  terms of their search domain, not in terms of BBC
  organisational structure.
• They dont know anything much about BBC
  organisational structure.

See all the regions -- all the different
divisions they have. Listed A-Z... Regions of
Northern Ireland
User no. 1, Fred
46
Zooming in or stepping sideways?
Key findings

• The specific section links area didnt seem
  specific enough

I'm normally quite rushed on the web. I haven't
got time to concentrate. So the specifics section
should say all the main headings for the subject
(Vikings). Boats, gods, battles, food The
specifics section would not help you out. It
doesn't tell you that it can look for specific
specifics.
User no. 3 , Emma
47
Stepping sideways
Key findings

• The planet of the apes search worked best because
  the sideways steps (film, cult, news) clearly
  refined the search.

You'd get used to that idea of the buttons
User no. 3 , Emma
48
Zooming out
Key findings

• Users cant articulate what a search engine does
  but they know what to expect when they see the
  term.

Oh! I didn't think it as a search engine. Yes I
know I said it would search the web. But I didn't
click!
User no. 2 , Gillian
49
Live!
50
Searching across all the BBC
51
Searching just BBC News
52
Searching the UK Web
53
More Lou!
54
lessons learnt... post-live assessment

• Marketing campaign starts April 20th
• Started a V1.5 testing and design blitz in early
  February
• 2 weeks, 3 user-tests with 2 days to iterate
  design in-between each!

55
V1.5
56
v1.5

• Whole team contributed to the design.
• Posted out new designs every evening
• Covered walls
• Fashion!
• Operating system trends getting more important.

57
What we learnt from the post-live testing

• People did NOT see the best links!!!
• All that effort, all the value useless if UI
  hides it
• Numbering is a curse
• It was contributing to skipping the good stuff.
• Relevance is irrelevant
• People loved the tabs once they found them.
• But they only found them on the 3rd or 4th visit.

58
Design lessons

• The tyranny of the search engine idiom
• Migrate habits
• Get team to collaborate on use-models early
  generates understanding early
• Gives whole team something to go back and
  question fundamentals usefully
• The whole team and the business understands where
  the value is
• The taxonomy
• The human factor

59
Conclusion
60
My Soylent Green Moment
Oh my God, Search IS PEOPLE!!!
61
QA
62
Thanks!

• Hope it was useful/interesting
• http//www.blackbeltjones.com/presentations/asist2
  002/