A major player in the publishing industry wants to capitalize on the emerging Brazilian middle class. Headquarters decides to translate 1000 technical books into Brazilian Portuguese on a limited budget and short turnaround times. After much discussion, analysis, and quotes, the client disappears with no commitment. Was it a real opportunity or an adventure not worth the time of our sales team?
How we failed to win a 100,000,000 word contract (GALA Istanbul 2014)
1. HOW WE FAILED TO WIN A
100,000,000 WORD CONTRACT...
Fabiano Cid & Diego Bartolomé
2. AGENDA
1. Fabiano talks
2. You interrupt him
3. Diego talks
4. You interrupt him
5. Fabiano talks again
6. You talk
7. We interrupt you
8. Diego interrupts Fabiano
9. We all shout at each other
(NOT NECESSARILY IN THIS ORDER)
3. AGENDA
1. Fabiano talks
2. You interrupt him
3. Diego talks
4. You interrupt him
5. Fabiano talks again
6. You talk
7. We interrupt you
8. Diego interrupts Fabiano
9. We all shout at each other
(NOT NECESSARILY IN THIS ORDER)
4. AGENDA
1. Fabiano talks
2. You interrupt him
3. Diego talks
4. You interrupt him
5. Fabiano talks again
6. You talk
7. We interrupt you
8. Diego interrupts Fabiano
9. We all shout at each other
(NOT NECESSARILY IN THIS ORDER)
5. AGENDA
1. Fabiano talks
2. You interrupt him
3. Diego talks
4. You interrupt him
5. Fabiano talks again
6. You talk
7. We interrupt you
8. Diego interrupts Fabiano
9. We all shout at each other
(NOT NECESSARILY IN THIS ORDER)
6. AGENDA
1. Fabiano talks
2. You interrupt him
3. Diego talks
4. You interrupt him
5. Fabiano talks again
6. You talk
7. We interrupt you
8. Diego interrupts Fabiano
9. We all shout at each other
(NOT NECESSARILY IN THIS ORDER)
7. AGENDA
1. Fabiano talks
2. You interrupt him
3. Diego talks
4. You interrupt him
5. Fabiano talks again
6. You talk
7. We interrupt you
8. Diego interrupts Fabiano
9. We all shout at each other
(NOT NECESSARILY IN THIS ORDER)
8. WHAT I SHOULD HAVE DONE
Now that I am looking back, I should have
qualified the lead:
• Publishing is not our target industry
• We have no experience whatsoever in this market
• Areas of expertise varied immensely
11. HERE’S AN IDEA!
• Using MT (Machine Translation) for books is
prone to result in bad quality
• Readers would complain and company’s
reputation would suffer
• How about release all titles for free download
with a BIG disclaimer; the most popular would
be properly translated
17. FIRST PROPOSAL
• Ccaps suggested a combination of machine
translation (MT) and human post-editing (PE)
• Since the material would be available to the general
public as high-quality publication, extensive human
interference by professional linguists was needed.
• Nevertheless, we provided a three-tiered proposal,
reflecting the three possible levels of human post-
editing.
19. MT ENGINE
We would have to build one engine from scratch for
each domain. This preparation process requires:
• bulk analysis of the source text
• detection of untranslatable terms
• compilation of a glossary for each domain to improve MT
output
20. MT ENGINE
We would also have to understand the way the XML
tags have been used to make sure the output respects
the original placement.
• Domain-Specific Engine Setup: X per domain
• MT Maintenance: included
• MT Storage: included
• Machine Translation Output: included
22. NO POST-EDITING
“With no post-editing, customer receives from Ccaps
the whole set of books translated into Brazilian
Portuguese with a raw output generated by the MT
engine. The format specified would respect the number
of XML tags, but these would be positioned in the same
location as the source text, requiring repositioning. “
26. FULL POST-EDITING
• Ideal scenario since it combines the best of
automated and human translation
• It will take longer and be costly than the other two
scenarios, but would greatly minimize the risks
involved.
• Post-editors are not necessarily experts in each
domain of the translated books, additional technical
review by an SME
• Resource to customer’s experts, who are already
familiarized with the subject matters
28. SOME ENGINEERING, DTP, ETC.
Engineering:
• Preparation and post-processing of translatable files,
• Extraction and placement of editable graphics in correct order and
location
• XML tag processing
• Charged by the hour; not possible to assess time needed
Desktop Publishing:
• Preparation of graphic files for translation
• Making necessary adjustments to graphic elements once the
translated text has been reverted back
• Charged by the hour; not possible to assess time needed
30. WE ARE READY!
• Not possible to predict the time to complete the whole
project
• Need to have a fairly large sample of data and more testing
time to evaluate quality and ascertain the effort involved in
the MT setup
• Based on initial analysis of samples provided, we expected
the following throughputs:
• Light Post-Editing: 7.500 words/editor/day
• Heavy Post-Editing: 4.000 words/editor/day
• We needed to know the scenario of choice to predict the time
and effort involved
32. CHANGE OF SCOPE
• Down to 100 books (rather than 1,000)
• Consider 10 domains
• Still 10,000,000 words
• Ccaps proposal still valid, as it does not
contemplate full amount
• Nope! Customer needs total cost
35. MASSIVE CHANGE OF SCOPE
“One of the options concerns a workflow in which the content
will be Machine Translated by another vendor first who will also
offer an online accessible infrastructure for you providing
Translation Memory, an environment for technical and final
review of the translation and tracking functionality.
Secondly I like to have a separate session with you. The agenda
points I then like to cover are:
• Brief introduction to your company
• Overview of your own approach to TM, MT, Terminology
Management, Staff selection and Quality Management
• Your opinion on advantages and challenges of a workflow with other
vendor
37. THIRD PROPOSAL
Specific amount:
• Domain-Specific Engine Setup: X
• Light Post-Editing: 2X
• Heavy Post-Editing: 5X
Turnaround time:
• Six months to complete
41. FIRST CONTACT, FIRST MEETING
- Contacted by prospect on April 10, 2013
- They had a project of "mass translation" and wanted
to arrange a meeting for that same day
- Meeting held at prospect's office.
- Project scope was explained and prospect seemed to
be familiar with the need for machine translation.
- Rough numbers provided and Fabiano promised to get
back to him the following day with some ideas.
42. FIRST CONTACT, FIRST MEETING
- Contacted by prospect on April 10, 2013
- They had a project of "mass translation" and wanted
to arrange a meeting for that same day
- Meeting held at prospect's office.
- Project scope was explained and prospect seemed to
be familiar with the need for machine translation.
- Rough numbers provided and Fabiano promised to get
back to him the following day with some ideas.
43. FIRST CONTACT, FIRST MEETING
- Contacted by prospect on April 10, 2013
- They had a project of "mass translation" and wanted
to arrange a meeting for that same day
- Meeting held at prospect's office.
- Project scope was explained and prospect seemed to
be familiar with the need for machine translation.
- Rough numbers provided and Fabiano promised to get
back to him the following day with some ideas.
45. WHY?
"LSPs see themselves as the keepers of a proven method to
produce translations and often push back on what they see as
unreasonable customer requests. They strive for stability and
predictability instead of providing solutions that will rock the
client’s world. Their customers want to outsource their global
communications to a trusted partner that has their interests at
heart. They don’t want to hear about difficulties; they want to
hear about solutions that will allow them to beat their own
competition."
Hélène Pielmeier, Common Sense Advisory