moravia - taus tokyo forum 2015
TRANSCRIPT
Machine Translation for After-Sales Support: How to Get It Right
Vladimir Reiff, Director of Sales APAC
Agenda • Pre-sales vs. After-sales • Business considerations - addressing the distraction • Marketing considerations – CS trends • Metrics, metrics… • Customer Engagement Lifecycle (CEL) • MT and CEL • Usual Suspect & Other Content • Case studies
TAUS Executive Forum, April 9-10, Tokyo
Pre-sales vs. After-sales
TAUS Executive Forum, April 9-10, Tokyo
After-sales: Distraction or Opportunity?
• After-sale support is often seen as a distraction – Every resource (human or financial) used on after-sale is a resource
which can’t be used on pre-sale, product development and other core activities
– “Writing knowledge base articles” is not a corporate goal for any manufacturer or developer
• However, this is changing as after-sales may increase client retention through increased engagement
• If you delight customers during after-sales, they are more likely to come back
TAUS Executive Forum, April 9-10, Tokyo
CS trends according to Forrester • Customer service is moving from cost center to differentiator • Voice of the customer programs are operationalizing insights • Analytics are improving the end-to-end experience • Outsourcing is slowly gaining market share
TAUS Executive Forum, April 9-10, Tokyo
Increased complexity & mixed budget ownership
Difficulties to find owner for all diverse CS activities
Metrics, metrics… • If CS is to work as differentiator, it has to provide tangible
results • Metrics are key but not very wide-spread in after-sales space:
– Pre-sales has tons of them (ROI, # of leads, brand awareness…) – Actual production does too (l10n headcount, costs, fully-loaded CpW…) – But how to measure after-sales support?
• Customer satisfaction • First contact resolution • Support call deflection
TAUS Executive Forum, April 9-10, Tokyo
“every time self-service deflects a contact, it also helps ten or more additional customers who never would have called”
TAUS Executive Forum, April 9-10, Tokyo
Call Deflection and MT • Call deflection can only happen when problem is solved without
direct assistance • Direct assistance can only be avoided if there’s relevant, helpful
support content • This can work internationally only as long as the content is
available in customers’ languages • All CS content cannot be translated by humans – cost, speed… • …but it does not even have to be as the main point is helpfulness
to solve problems, not linguistic perfectionism!
TAUS Executive Forum, April 9-10, Tokyo
Customer Engagement Lifecycle
TAUS Executive Forum, April 9-10, Tokyo
Post-‐sales training Connec0ng local customers in the same market
Connec0ng customers to
product/service consultants
Crea0ng or reinforcing user
groups
B2C as well as C2C
Online forums
Live internet chat
Knowledge base ar0cles
Telephone support
Email support
Social Media Management
Extended warran0es,
subscrip0ons, etc.
Product mailings and
announcements
Managing upgrades and renewals
Process returns, refunds, etc.
Immediate Help Ongoing Engagement Community & Lifestyle
Usual Suspect - Knowledge base articles
• By far the most content • Natural “first line of defense”
– Customer with an issue will first go to company’s website to seek help
– Website’s ability to help solve the problem will determine the experience and impact the brand • Humans have negative bias – people will share
bad experience on social media 50% more likely than good experience
• Impossible without MT
TAUS Executive Forum, April 9-10, Tokyo
What about the Other Content?
TAUS Executive Forum, April 9-10, Tokyo
• Online forums, blogs, wikis • User-generated content • Live internet chat, to support real-time customer interactions • Telephone support • Email support
How to handle the content effectively?
• Pre-editing – Improving the source to increase productivity of following steps (better
to fix one English source than 25 target languages)
• Post-editing (none, selective, full) – Post-published post-editing (“P3”, Chris Wendt, MS) – Publish first, post-edit later based on demand and feedback – there
where selecting the text ahead of time is not practical or possible and the risk of omitting important/relevant information is high
• MT to support self-service • MT to lower internal overhead for support
TAUS Executive Forum, April 9-10, Tokyo
Typical Target Companies… • …have high customer turnover • …have low switching costs • …are concerned with international/global expansion • …have large volumes of customer support materials as a natural “by
product” • …”live by the brand” and perception of poor customer service can
damage the brand • …are typically found among:
– Software products and services developers/providers – Hardware and consumer electronics manufacturers – Travel, cruise lines, vacation and hotel chains
TAUS Executive Forum, April 9-10, Tokyo
Case Study 1 – Consumer Electronics
TAUS Executive Forum, April 9-10, Tokyo
• Expansion to new countries results in rising support costs • Support center is becoming inefficient – too many people are
involved in resolving queries which are difficult to handle in multilingual environment
Challenge • Analysis of the support activities (phone, email and online chat)
identified many cases which could be addressed by self-support • MT solution implemented together with support writing team –
repeated or serious issues documented and immediately published to the website
Solution • Assisted help focused on complex issues • Increased customer satisfaction • Lower support costs
Result
Case Study 2 – Software Developer
TAUS Executive Forum, April 9-10, Tokyo
• Knowledge base articles are well managed in English and Tier 1 languages with MT PE • Decisions to localize into target languages are taken per tier • Issues in different countries however differ – some of the content is not used, some seems to be
missing • Countries in Tier 2 and Tier 3 languages suffer poor client satisfaction due to limited support
Challenge • Decision to decrease quality of the support content to allow more content in more languages • MT as a starting point (with exception of a few languages with insufficient MT quality, at least at
the beginning) • PE applied for often-researched topics or topics with low ‘resolution rate’
Solution • Amount of content available across all tiers increased dramatically • Positive response especially from countries with low-to-no previous local support • Thanks to higher PE rate to not alienate customers used to get high quality support, Tier 1
languages didn’t show any significant satisfaction issues Result
Case Study 3 – Do Not Panic!
TAUS Executive Forum, April 9-10, Tokyo
• Support content for consumers (end clients, developers, integrators) • Translation is provided for full set of supported languages Challenge
• MT is used wherever possible to minimize or even remove HT to reduce the cost and time associated with human translation Solution
• Still a lot of human translation involved because: • High Quality MT:
• One of the top languages with solid MT output (due to enough content to train the engines) • Content type which is suitable for MT Result
How to get it right? • MT (PE) quality and more content/languages beats HT quality and
less content/languages • What’s important depends on the clients – let them choose & fine
tune based on demand/feedback • MT (PE) can decrease support costs while increasing customer
satisfaction • After-sales can be a differentiator, treat it as an opportunity to
delight customers & increase repeat purchase
TAUS Executive Forum, April 9-10, Tokyo
Thank You! [email protected]