Wordparrot.com 商業計劃書

國立臺灣大學管理學院企業管理碩士專班 碩士論文

Global MBA College of Management

National Taiwan University Master Thesis

Wordparrot.com 商業計劃書 Wordparrot.com Business Plan

周安里 Alec E. Jones

指導教授：陳家麟 博士 Advisor: Chialin Chen, Ph.D.

中華民國 106 年 6 月 June, 2017

Acknowledgement

I would like to thank my thesis advisor Professor Chialin Chen. He has been

instrumental in helping me organize my thoughts and develop a coherent approach towards positioning my company for its future growth. I would also like to thank the professors from the oral defense committee, Professor Ling-Chie Kung and Professor Kevin Chen, who were measuredly harsh in their criticism and helped me realize the most crucial need for branding and value proposition in advance of a site launch.

I’d like to thank all those who allowed themselves to be interviewed for this paper, including Max Jung, Olwen Zhang, Lee In-Young, and Maggie Lee. And finally I’d like to think my friends and family for their emotional support and thoughts about developing the site. There are sure to be many challenges ahead.

Abstract

The issue of how to translate something is a problem that inevitably comes up in one’s academic, professional, or personal life. The most immediate instinct one might have is, “do I know someone who can help me take care of this?” And in many cases, that need can be quickly and painlessly resolved by a friend or colleague. But in other cases, this solution becomes more complicated. What if this person can’t be bothered to get the job done? What if speed or quality is an issue?

The idea for Wordparrot was borne out of a number of experiences in Taiwan with people who have, in their past experiences with translation companies, lamented the high cost, inconsistent quality, or both at once. At the same time, it is the result of the founder’s lifelong fascination with languages, foreign cultures, and the funny results that often occur from attempting to bring the real meaning of a phrase from one cultural understanding to another.

The essential goal of Wordparrot is to use the open source tools and APIs available on the 21^st-century Internet to build a translation marketplace for the end consumer.

Customers can create accounts, upload files, and be connected to a global network of translators who will perform translation services for a fee. It is neither a revolutionary idea, nor one that requires a Ph.D. in computational linguistics. The intended drivers of profit for the company will be a high volume of low-margin transactions through a globally accessible platform. The company’s differentiating qualities will be the convenience features of the web application and the company’s ability to globally source and showcase translator expertise.

Until computer scientists figure out how to make computers render translations that are indistinguishable from that of humans at large scale, human beings will have a role to

play in making it work. Languages are constantly evolving, and machines still have great difficulty in capturing the shared, unspoken understanding between speaker and listener that is necessary to express ideas accurately on paper. That’s not to say that computers don’t have any role to play – they have revolutionized the field. From the standpoint of future business, it is better to think not in terms of “machine translation” but rather that “machine-augmented human translation” as the new standard for some time. The chapter ‘The Automation Question’

goes deeper into this important issue.

Utilizing machine-translation APIs like Microsoft Translator to increase productivity is an eventual goal for the company. But for now, the company believes value can be delivered and profit can be made by achieving much simpler goals – finding customers who need translation, either singularly or as a regular business/institutional requirement, and bringing them together on a platform with translators available remotely for work on a regular or short- term basis.

The platform’s eventual goal is to offer quality and convenience by encapsulating the various types of technologies that can enhance the convenience for both sides – social media- based authentication, file uploading and storage, and payment processing not just with credit cards but also Paypal, Apple Pay, and Android Pay. More ambitious goals are to develop web-based chat, videoconferencing, and project management tools to help translators and customers coordinate difficult projects that require multiple stages of revision.

Table of Contents

MASTER THESIS CERTIFICATION BY ORAL DEFENSE COMMITTEE ... I ACKNOWLEDGEMENT ... II ABSTRACT ... III TABLE OF CONTENTS ...V LIST OF ILLUSTRATIONS ... VII LIST OF TABLES ... IX

CH1. THE INDUSTRY ... 1

1.1COMPETITIVE ANALYSIS ... 1

1.2MARKET SURVEY ... 7

1.3IDENTIFYING PAIN POINTS ... 12

1.4OFFERING SOLUTIONS ... 14

1.5 THE AUTOMATION QUESTION ... 17

CH2. THE BUSINESS MODEL CANVAS ... 20

2.1 CUSTOMER SEGMENTATION... 20

2.2 RELATIONSHIPS ... 21

2.3 CHANNELS ... 22

2.4 KEY ACTIVITES ... 22

2.5 KEY RESOURCES ... 23

2.6 KEY PARTNERS ... 24

2.7 REVENUE STREAMS ... 25

2.8 COST STRUCTURE ... 26

CH3. PLATFORM FEATURES OVERVIEW ... 27

3.1 REGISTRATION PROCESS AND ENTERING THE SITE ... 27

3.2 CREATING AN ORDER AND SUBMITTING FOR QUOTE ... 29

3.3 PAYMENT & RETRIEVING FINISHED PRODUCT ... 32

CH4. FINANCIAL PROJECTIONS ... 34

4.1DETERMINING PRICING & MARGINS ... 34

4.2 SCENARIO ANALYSIS ... 36

4.3 ESTIMATING FIXED COSTS ... 37

4.4PROJECTING BREAK-EVEN ... 39

CH5. ASSUMPTIONS & PRIORITIES ... 40

5.1REVIEWING ASSUMPTIONS ... 40

5.2INITIAL MARKETING AND BRANDING ... 41

5.3 LAUNCH-DAY PRIORITIES ... 42

5.4 MEDIUM-TERM PRIORITIES ... 42

5.5 LONG-TERM PRIORITIES ... 43

CH6. REFERENCES ... 44

CH7. APPENDIX ... 46

I

1A – P

’

2A – B

M

C

T

4A – S

O

T

P

P

Target Markets Prices Per Word

Basic Expert

Exchange Rate

Converted Basic

Converted Expert

Taiwan

MeetTheTranslators.com NTNU Translator

2 4 0.033 0.066 0.132

Services 3 6 0.033 0.099 0.198

Writepath 0.11 0.11 1 0.11 0.11

United States

onehourtranslation.com 0.164 0.192 1 0.164 0.192

UK

intercombase.com 0.07 0.08 1.114 0.07798 0.089

Korea

Editage 70 150 0.000889 0.0622 0.133

Japan

Gengo 0.06 0.012 1 0.06 0.12

Average (in USD) 0.091 0.139

–

4B R

& C

–

4

–

– 36

Annual Rate

Months after Launch Pessimistic Realistic Optimistic

6 1000 2000 4000

12 2000 4000 8000

18 3000 6000 12000

24 4000 8000 16000

30 5000 10000 20000

36 6000 12000 24000

0 50 100 150 200 250 300 350 400 450

1000 2000 3000 4000 5000 6000

Number of Orders

Revenue & Contribution Margin

Revenue Gross Margin

–

4D

– 36

4E –

– R

0 200 400 600 800 1000 1200

0 6 12 18 24 30 36

Months after Launch

Revenue Growth Scenarios (Annualized)

Pessimistic Realistic Optimistic

0 50 100 150 200 250

0 6 12 18 24 30 36

Months from Launch

Period Gross Margin Scenarios

Pessimistic Realistic Optimistic

–

4

.

–

Inside.com.tw listings (June 2017)

Developer

Entries

Monthly 10% Annualized

# ID Description Salary Premium (x13)

24415 Front-End Engineer 38000 41800 543400

24417 Front-End Engineer 40000 44000 572000

23865 Web Developer 40000 44000 572000

24405 Full Stack Developer 42000 46200 600600

22771 Front-End Developer 32000 35200 457600

Software Engineer -

20627 Backend 45000 49500 643500

UI/UX Entries

22602 Web UI/UX Designer 35000 38500 500500

23870 Web/UI Designer 40000 44000 572000

24406 UI/UX 50000 55000 715000

24402 UI Designer 55000 60500 786500

Sales Entries

24235 Business Development 30000 33000 429000

24072 Business Development 40000 44000 572000

24057 Sales Development 30000 33000 429000

–

4

,

REALISTIC SCENARIO

4

–

,

0 10 20 30 40 50 60 70 80 90

6 12 18 24 30 36

Months from Launch

Period Salary Expenses

Developers Sales Administrative

0 20 40 60 80 100 120

6 12 18 24 30 36

Months from Launch

Period Fixed Expenses (Realistic Scenario)

Salaries IT Marketing Legal Misc.

4

–

,

-

4

–

,

/

3

-80 -60 -40 -20 0 20 40 60 80 100

6 12 18 24 30 36

Months After Launch

EBITDA Per Period

Pessimistic Realistic Optimistic

-300 -200 -100 0 100 200 300

6 12 18 24 30 36

Months after Launch

Accumulated Profit/Loss

Pessimistic Realistic Optimistic

C

1 -

I

1.1 – C

A

The translation industry is a vast and fragmented market. According to a press release by Gengo, the global translation industry is estimated at $33 billion USD per year. The EU Directorate-General Report on the Status of Translators indicates that in 2012, approximately 330,000 people were employed as translators or interpreters worldwide. Seventy-eight percent of translators in Europe worked as freelancers. In the U.S. more than 3,000 firms employ over 55,000 professionals in various aspects of language services. More than 50,000 people work as translators and interpreters, according to the U.S. Bureau of Labor Statistics.

Translation is a vast industry, encompassing interpreters, written documents, Youtube captions, movie scripts, Amazon product descriptions, and website localization.

Unlike an industry like the smartphone industry, the translation market is extremely fragmented and full of privately-held companies who do not disclose financial information that would be helpful in determining market analysis figures such as market share or consumer segmentation analysis of some of the competitors. Instead, it is more useful to break down the industry itself with Porter’s 5 Forces, and then take a look at some of the business models associated with the industry, which range from Fortune 500 companies exclusively dealing in software down to freelance business operators.

Bargaining Power of Customers: Medium to High. The price of simple translation is

constantly driven downwards. However, companies who can establish a reputation for reliability and difficult-to-find expertise can command higher prices from consumers.

Bargaining Power of Suppliers: Medium to Low. As with bargaining power of customers,

suppliers (translators themselves) with more expertise and reputation have more negotiating leverage in any transaction.

Threat of New Entrants: High. No enforceable industry barriers or government regulatory

schemes restricting market entrance exist, and companies can operate across borders.

Threat of Substitute Products: Medium. The biggest substitution threat here would be

machine translation, which currently is only capable of producing quality at the lower threshold and cannot meet the demands of any customer who requires native or near-native fluency as a quality requirement.

Competitive Rivalry: Medium to High. Again, this aspect of competition directly correlates

with how much a company or translator can differentiate themselves with hard-to-find expertise.

Common Business Models in the Industry

The types of competitors in the translation market can be roughly categorized by several different types of business models. These categories are listed below, starting with the largest in size and technological resources and ending with the smallest and least technologically oriented. A sample of companies fitting these categories is also included.

Machine translation software

This business model focuses around the concept of using computers to create translations. It is best used in an assistive role to reduce the human workload, as purely computer-driven translation is still problematic.

Google Translate, Microsoft Translator

These global technology companies use machine learning and neural networks to hone complex translation systems, which they provide access to via APIs. They are not ideal for longer and more difficult translations, but serve well in cases where simple sentences and casual translation requirements are the order of the day.

Translation Tools

Companies employing this business model may or may provide translation services to endconsumers, but they focus on selling translation tools that make it easier for other companies to translate for their own needs, or for agencies/freelancers to use. Some translation tools include website/content management system (CMS) localization and project management

tools.

Lionbridge

This company operates a number of translation software solutions for businesses. Its recently-released GeoFluent SaaS offering allows companies operating call centers around the world to translate audio and text on-demand between customers and call center employees.

Trados

Trados is a German company that sells a rules-based computer translation software package.

Software licenses can be purchased by freelance translators and corporate teams to simplify their workflow and increase productivity.

Crowdsourced Translation Platforms

These platforms match customer translation needs with a large network of translators around the globe. Their target segment is smaller translations requiring less complexity and specific expertise. The platform creator is responsible for setting prices and wage rates, and enforcing a necessary minimum level of quality for its translator network. This business model is closest to the intended model of the company.

Gengo

The Japan-based platform company operates on the simple principle of accepting documents and returning translations as quickly as possible. It offers two different tiers of expertise, basic and business level, at a constant price-per-word pricing model. It specifically warns customers that it will not accept documents from the legal or medical field.

Flitto

Flitto is a Korean company that operates several parallel translation businesses. The first is a crowdsourcing business. It leverages a large translator network for customersubmitted content of very short lengths, such as a short blogpost, banner, or even a

tweet. For document translation, it provides a one-on-one translator matching service like Gengo.

Finally, it also has a corporate sales team for more complex translation needs. However, unlike Gengo it does not standardize pricing for one-on-one service and instead simply charges a fee on top of whatever price is cited by the translator.

Freelancer Yellow Pages

This business model doesn’t derive its income from the translation process itself, rather than simply charging a fee to allow individual translators or small agencies to access its substantial user base. They are best at providing a closer look at their large networks of service providers and give you the choice of whichever fits your needs.

Upwork

In addition to translators, this company focuses on variety of freelancer types, such as web and mobile developers. Customers will create their job requirements, and the website will scan the posting and attempt to suggest good-fitting candidates. Payment to the freelancer is handled directly on site, and the site takes a percentage of the total fee paid out to the freelancer.

Facebook/Yahoo/Craigslist

Large social media sites that allow classified postings are a good source of freelancers and

are often free for both the freelancer and customer. However, the sites have no responsibility

for enforcing a minimum level of quality and don’t get involved with payment processing, and therefore have a ‘buyer beware’ quality about them.

Local Agencies

Companies matching the description of ‘translation agency’ are too many to list here.

Translation agencies excel at providing a specialized level of expertise and the greatest level of customer service and order customization. Often times they focus specifically upon one type of area, such as college applications, legal documentation, or website localization.

Agencies may have multiple branches or be as small as just a few people in size. Their technological capabilities are generally more limited than the examples previously, and generally include a brochure website with features such as accepting file uploads in order to process quotes for customers. They rely on free or low-cost software such as email and Dropbox and Google Drive, or are consumers of the above software packages mentioned earlier.

Freelancers

Freelance translators are as numerous as small translation agencies and can vary significantly in quality, price, and expertise. Like small agencies, their technological capabilities tend to be limited to simple brochure websites, email correspondence, and whatever free software is available on the internet. They may consider translation a career, or may merely occasionally offer their services in order to supplement their income.

1.2 – M

S

The market for translation and document preparation is international, heterogeneous, and largely unregulated. Given the low barrier to entry of providing translation services, customers are faced with a plethora of options when it comes to meeting their translation needs. These range from the casual (asking a friend or colleague and paying in cash) to the very professional (formalized contracts and bidding processes with pre-approved agencies), depending on the needs of the customer or organization. Important aspects in determining the appropriateness of a vendor include price, convenience and speed, reputation for quality, and ability to grant credit payment terms.

The following are summaries of a number of interviews conducted with people involved in the process of sourcing professional translation, either internally or through translation agencies.

Name: Max

Nationality: Korean, based in Seoul

Role: Project Manager, Korean Trade-Investment Promotion Agency Description:

As a project manager, Max deals directly with translation agencies on behalf of the KOTRA government agency. Most recently he is responsible for contacting agencies to translate important brochures for the agency’s high-level talks with Malaysian trade authorities. The Korean government is attempting to promote Korean-manufactured electronics and cosmetics in Malaysia. His organization restricts him to sourcing amongst

approved agencies only, which are reviewed through an annual approval process. Currently 5 agencies are approved.

Experience:

Max requires a 5-page document to be translated. This document is legal in nature and

requires expertise and precision. Max contacts the agency via email. He negotiates with a single contact at the agency, who himself is a translator. The agency’s turnaround time on the document is 48 hours, and Max receives the final product as a MS Word document attached to an email. Quality is high and the agency is known by reputation for reliability.

Typically, documents contracted to agencies like this can range from a single page to a 200- page annual report. The organization produces many reports and published papers originating in Korean and translated for international audiences.

Name: Olwen

Nationality: Taiwanese, based in Taipei/Hualien Role: Translator, part-time

Description:

As a translator, Olwen handles document preparation for a small agency that deals almost exclusively in the field of immigration. Wealthy Taiwanese looking to obtain immigration status for their children in the United States or Canada pay high fees for standardized documents required by Canadian immigration authorities. The clients also require consulting by professional accountants, since they are required to disclose important financial information and especially information related to penalties they may have paid for aggressive tax avoidance practices.

Experience:

While Olwen remarked that the agency offered a competitive wage, she also had the

impression of dishonesty and tolerance for poor quality in her agency. Olwen was not allowed to have any contact with the clients themselves, despite providing services for her assigned clients entirely on her own. Aside from a single foreign lawyer, all employees of the agency are Taiwanese. The agency pays a good wage, but turnover is high and the boss is known to be extremely controlling over the entire process.

Typically, Olwen would be responsible for 50 pages’ worth of documents and a turnaround time of 2 weeks. All work occurred within the office on office computers and no work was allowed to be removed from the office. Clients negotiated on price, and typically paid by check. Work involved frequent revisions, which required messages to be relayed back and forth through a salesperson.

Name: In-Young

Nationality: Korean, based in Seoul

Role: Public Relations, Baxter Pharmaceuticals Description:

In-Young is responsible for contacting a large Singapore-based translation agency for

periodic translation of press releases and other corporate materials. These materials frequently but not always remain internal. All Baxter branches in the Asia-Pacific region are required to contract with this agency. However, the company has also contracted with freelancers and smaller local agencies in the past.

The pharma company specifically avoids contracting exclusively with any one company, preferring instead to maintain a competitive bidding process. Each translation job requires solicitation of bids from multiple agencies. The company prefers the Singaporean company specifically due to past issues with translators lacking necessary expertise with the pharmaceutical industry and associated terminology.

Experience:

In-Young communicates with the agency almost exclusively via email, and via phone on

rare occasions when time is an issue. The person with whom she spoke is the single point of contact within the agency, and is responsible for both client relations and translation.

Although expertise in the pharmaceutical industry is required, the agency’s reputation and status as ‘registered’ is the primary signal of quality.

In-Young first contacts the agency by attaching the relevant document within an email, and expects a quote within a few days. Once the process has started, In-Young can expect multiple back-and-forth emails as she and the translator discuss numerous necessary revisions. The translator uploads the document to a cloud host such as Google Drive and gives In-Young the ID and password needed to access it. After the job is complete, In-Young will deliver an invoice provided by the agency to her company’s finance department, and the agency will be paid in a lump sum.

Name: Linda

Nationality: Malaysian Chinese, based in Singapore Role: Translator/Executive Assistant, Singapore

Description:

Linda is a full-time employee of a Japanese multinational’s branch office in Singapore. The

corporation provides a variety of specialized construction services for large infrastructure projects in the Southeast Asia region. Linda’s job includes but is not limited to translating internal documents created by the Japanese headquarters and intended for the company’s Singapore-based engineering teams. Linda translates 2-3 documents per week, ranging from 1-5 pages in length. The documents always originate in Japanese and are translated into English or Chinese, which are both languages she speaks fluently. She also performs some executive-assistant tasks and translates a number of documents for the branch manager’s consumption.

Experience:

Linda is not a native Japanese speaker, but attended a Japanese university as an undergraduate full-time. A confounding issue that she had with the job upon hiring is that proper translation required knowledge in mechanical engineering, and Linda had no engineering expertise. In her words, she took ‘a full year on the job’ in order for her knowledge of engineering terminology to develop to meet the needs of the job. Now, she said, she is fully knowledgeable about the required terms and translating most documents is routine.

However, over time she has become bored with the lack of variety in this type of translation work, and she stated her intent to leave the company within two years.

1.3 –

I

A common concern for customers is the need for quality. Determining quality in something like translation is highly subjective and difficult to quantify. Each customer’s “quality threshold” is different. For one client, the need for correct word choice and native-sounding fluency is crucial, whereas for another it is only important that the translation be comprehensible. Going through multiple stages of revision in a document is inevitable and often expected.

One of the confounding issues is that customers of translation are often unable to discern quality from mediocrity if the target language is one that they don’t understand. In other cases, the people interviewed often have competent skills in English, and are able to discern a quality translation from a mediocre one. However, they don’t wish to have too much of their time and attention removed from their jobs in order to ensure that the product meets their quality standard.

The Expertise Problem

In at least three of the interviews conducted, the client company was in need of specialized translator expertise, including knowledge of pharmaceuticals, medicine, and mechanical engineering. Sourcing translation of this type is an inherent tradeoff between availability and quality. This complaint elegantly describes the major problem:

“Professional translators don’t have the knowledge we need, and the people who have the knowledge don’t have the time or desire to be translators.”

In many companies, this need only occurs occasionally, and the company deals with the problem by asking its full-time employees in non-translator capacities to translate for them. It’s a common “make or buy” problem that multinational corporations will face.

However, asking engineers to translate or correct very often will amount to a mis-allocation of corporate resources, especially if the employee is being taken off more important responsibilities in order to correct a translator’s inadequate work. In Linda’s case, the company solved the issue by investing significant resources into training a person with no engineering expertise. That solves the issue, but it also leaves the company vulnerable should she leave.

The Convenience Problem

Customers procuring translation on behalf of organizations have ready access to today’s fast, full-featured social media platforms in their individual lives. They expect a similar level of convenience and design when required to choose the tools on which their job depends. When attempting to use a web platform for their jobs, they will expect the process to be reliable and not confusing to use. If visiting a translation agency’s website (or many other sites, for that matter), the average user will only give the front page a cursory glance before making a judgement about whether the site can meet their needs. In many cases a judgement about quality will be made based on the website’s design quality and not the people working behind it.

1.4 –

Create quality signals – Translator web profiles, inspired from social media platforms such

as Twitter, can act as publicly available vetting tools to bring additional comfort to clients concerned over the agency’s ability to meet their requirements. Clients can go onto profiles and download samples of the translator’s past work, be linked to any journal articles published by the translator, and get a comprehensive view of the translator’s curriculum vitae.

Electronic vetting – Web-based testing of candidate translators has value in separating great

translators from merely decent ones. In addition to the normal HR processes like job interviews and resumes, the company can develop online tests that pepper job applicants with tricky questions and weed out applicants with deceptively strong resumes but insufficiently weak writing skills.

The Expertise Solution

Attract skilled experts in a part-time capacity – The platform is being built with

the full knowledge that potential translators may not intend to do translation as a career. At academic institutions across the globe, graduate students who may have already accumulated several years’ worth of experience in their fields will commonly re-enter universities to acquire Master’s or Ph.D degrees and increase their long-term employability and earnings potential. Especially in the United States, students may do so and either take on significant student debt or otherwise be forced to accept research positions under tenured professors that are below their earnings potential in the industry.

Such individuals are ideal translators for the platform. The ideal job for them is one that provides additional income for their lifestyle needs and provides flexible work schedules and requirements without requiring a commute or adherence to a strict hourly schedule. For longer documents, a job can be split up into multiple pieces and one translator assigned to an editing and compiler role in order to ensure that the work of a team working remotely attains a necessary standard.

Attract a larger pool of expertise through technology – A contracting agency’s

ability to source rare expertise is directly related to geography and red tape. Simply put, an agency’s ability to find talented people is restricted to those people with whom it can realistically conduct business. For example, a Taiwan-based agency attempting to hire translators in the United States has several obstacles to overcome, including communication across time zones, language barriers, and local regulations (especially with regards to payment).

A web-based translation agency may be able to overcome ‘expertise scarcity’ by enlarging its pool of potential contractors through a capable web-based interface. The world of social media is now instantaneous, with slick chat messenger systems and (for larger orders) multinational teams of translators communicating job requirements and working remotely.

So why is something like professional translation still using long, cumbersome email chains, traditional payment methods like checks, and Dropbox? It’s a workable solution but not an optimal one for clients or for translators. The idea behind this company is based on the idea that building convenient and integrated remote solutions will be better at luring and retaining both sides of the platform.

The Convenience Solution

Social Media Login – New users hate making new accounts on websites, and frequently forget the credentials of their account almost immediately after supplying them for the first time. This ensures that the platform’s potential customer never returns to the site.

Enabling common social media platform logins through the Oath2 authentication standard is the best solution to this problem currently.

Encapsulate Important Functionality – As In-Young pointed out, translation

agencies use the tools available to them, such as 3^rd party cloud host and document storage.

However, remembering the login details for this cloud host is not convenient. Having a web platform that hides the details of web storage from the user will provide a better user experience.

Communication and Version Control – A common identified issue is the need for

multiple revisions during the process of translating documents. While this is unavoidable, the pain can be lessened by incorporating an interface that integrates the main components of the translation process, which are client-translator bilateral communication and document uploads/ downloads. The envisioned interface in its final form will incorporate the convenience of a work-based messaging platform like Slack with the version control

capabilities of a site like Github.

1.5 –

In a market with very low government regulation and low barriers to entry, the most important question for potential investors is whether competing firms can achieve profitability. This question also looms in the face of ongoing technological innovation in the field of machine translation, which continues to pop up in some of the world’s most popular applications. For example, Skype’s recently released Translator application combines its video conferencing software with voice recognition and machine translation in order to render fully computer-translated text for two users communicating via different languages.

Estimating the profitability of entering the translation market therefore involves answering two important questions: 1) Will machines completely replace humans in the field of translation? 2) If no, can human translation be profitable after factoring in the consumer demand lost to increasingly sophisticated machine-driven solutions?

For now, the answer seems to be no, and yes. The strongest argument against

computers completely replacing human beings in translation work is that translations involve a heavy amount of context and inference. When attempting to bring a meaning that is as close to the original sentence as possible, a translator must weigh cultural differences, tone of voice, innuendo & double meanings, formality and gender, and numerous other factors. When attempting to translate between languages of different language families, especially between East Asian and Germanic/Romance languages, context and meaning are especially

susceptible to being garbled.

Computer-driven translation generally falls into two different types:

Rules-based Translation (Yahoo Babelfish)

Computer algorithms are designed with a complex set of grammar and lexicon rules in order to render accurate translation. These rules are implemented by linguists and refined over time. Translation is therefore rendered by matching text that is analyzed by the computer algorithm with pre-established patterns. The main drawback with this approach is the large human effort required to develop reliable rulesets and improve upon them over time. Another limitation is that rules-based translation is less effective for rare language pairs.

Statistics-based Translation (Google Translate, Microsoft Translator)

In a statistics-based approach, the computer algorithm is not given any pre-defined rules by linguists. Instead, algorithms employ machine learning techniques and neural networks in order to develop their own sets of rules. Using huge datasets and massive computing power, the neural network analyzes many different types of texts. Translations by these systems are the result of calculating a complex set of probabilities that improves over time. Statistical translation has largely replaced rules-based systems in the past 10 years with advances in computing power and machine learning techniques.

Machine Translation – Advantage or Threat?

It may be more beneficial to view translation systems developed by the world’s largest technology companies such as Google and Microsoft as key partners rather than adversaries.

Statistical machine learning is already being employed in situations like generating human- like reports of daily stock price movements and sports scores. In these cases, grammar and syntax are repeatable, and the acceptable threshold for quality is low.

In more specialized cases where fluency and accuracy are critical, or where subtext in language is more important, machines cannot yet replace human beings.

An area where machine translation can actually benefit human translation as a business is the process of “first pass” translation. This is the process by which human translators use a machine translation tool to generate a rough draft of a translation, and then fix errors and re-mold sentences to achieve a more desirable and fluent result.

The difficulties of machine translation are elegantly summarized in the Youtube video,

“Why Computers Suck at Translation”, by Tom Scott, a British computer programmer and public speaker on such topics as computing, cybersecurity, linguistics, and scientific innovation. This video is highly recommended for anyone interested in getting a 5-minute overview of the difficulties faced by the machine translation field.

C

2 – T

B

M

C

2.1 – C

S

The long-term objective of the platform is to appeal to a large variety of consumer needs. In chapter 5, “Assumptions and Priorities”, the plan lays out how the site’s overall strategic objective is to initially focus on a niche market with a small team of translators, and to gradually expand strategic scope into a generalist e-ecommerce tool as features mature and the volume of customers increases.

The types of industries/fields in need of translation are too numerous to list here.

Instead, the customer segments can be broken down into three distinct categories, to comprise six different main segments in total:

Frequency (Isolated, Repeat) - How often is the customer planning to require

translation services? A potential customer may be only in need of a single purchase. Other customers may require services in large bursts occurring at certain times of the year, or they may have a steady need all year. Due to the difficulties associated with evaluating the needs of first-time customers, the company should prioritize repeat customers in its pipeline.

Size (Individual, Organizational) - Is the customer contacting the company doing

so on his/her own merits, or on behalf of an organization? The difference can be large in terms of understanding requirements and determining payment terms. Organizations tend to have more specialized needs and more stringent requirements in terms of quality and speed of work. They can also pose bureaucratic problems if responsibility for determining work requirements is delegated amongst two or more people. However, they also have budgets and

may be willing to pay a premium for reliability and speed. The company should prioritize organizational clients over individual ones.

Expertise (Generalized, Specialized) - This attribute measures the difficulty of translation requirements. Here, specialized can be defined as requiring at least a master’s level (current student or graduated) in a given field. The company should not assign inexperienced people to translations requiring prior expertise and familiarity with esoteric terminology such as in the engineering, pharmaceutical, or legal fields.

2.2 –

Developing a relationship with the customer and understanding customer needs is one of the strengths of orienting the business model around technology. Much of the human effort required in the initial contact can be automated – the site walks a first-time customer through the steps of providing order info, uploading files, and understanding the company’s workflow.

Using 3^rd-party payment systems allows the company to offer a range of payment options, such as credit cards or Paypal. And in the future, the company plans to build upon the platform to enable more customer-facing features such as real-time chat with salespeople and even teleconferencing.

Just as important to these relationships is our ability to collect data through the site.

Each order allows the company to continue building a dataset over time, in order to fine-tune our pricing strategy and determine the most favorable types of customers & industries to target for a growth strategy.

2.3 –

The site is intended to unify the various tools required in the translation workflow, including customer account creation, file storage and downloads, and payment processing. In the future, it may be possible to diversify the types of services offered beyond the site, but for now, the website is the exclusive location where company services are offered. At the outset, a web application is the sole focus of software development.

Native mobile applications for Android and iOS will be available in the future pending financing availability, as those software platforms require different skillsets and therefore different developer teams. Mobile applications may be less relevant to this business model as word processing and spreadsheets continue to be in the domain desktops and laptops.

2.4 –

Platform Development – This requires a development team in order to continually add new features and fine tune existing ones. An important aspect of the company’s future growth depends on diversification of revenue streams.

Business Development – A sales team is required in order to promote the web platform and

draw in new clients. A large churn of single-order clients is expected, and therefore the company must be proactive in seeking business with organizations who can reliably make orders with regularity. The matter of what type of promotional activity the company should use in order to attract first-time customers is still under debate, but may include “first order is free”, or deals to that effect.

Talent Development – At the outset it will be difficult to attract and retain high-level expertise in required fields to perform translations. The company’s challenge is to source enough regular business from higher-paying customers to retain translators with specialized expertise.

Customer Relationship Management – The translation industry is an inexact process, and

customer needs will vary highly with each order. A major challenge will be to manage customer relationships appropriately in order to minimize failures in processing orders, whether those failures are in quality, speed, or misunderstanding requirements. Tools such as chat interfaces and customer surveys are essential in order to understand requirements, complete orders quickly, and produce happy customers who return again and again.

2.5 –

Intellectual/Technological – The web platform itself. The platform is built on Angular, a

software development platform whose current iteration was released in 2016 and developed by internal teams at Google. Although its APIs are relatively new and are bound to experience some changes as the framework matures, Angular offers a full set of features for single-page applications under a permissive open-source software license. It is a popular development choice for the enterprise space and has been built with application scalability and the latest web standards in mind.

Human – This includes the development team required to build and maintain the web platform, constituting the company’s largest fixed-cost expenditure. It also includes a sales team, which is important for achieving company growth. The translation industry has low

barriers to entry and is highly competitive, and therefore the company must be proactive in acquiring new customers. The translators within the platform are obviously a critical human resource as well.

Financial – The company will have to spend its early-stage capital wisely and keep fixed

costs low. It will expend some effort acquiring customers at the outset, and due to the business model, a large percentage of them will be one-time customers. Therefore, in the early stage the company must wisely choose where to spend money and count on growth from word-of- mouth advertising.

2.6 –

The company’s partners include the various companies providing paid API services, and will increase as time goes by. These currently include:

Stripe + Braintree (payment processors) – These companies provide feature-rich payment

processing APIs, and using them together gives the company the capability to process transactions using credit cards, Paypal, Apple Pay, or Android Pay. The APIs work by forwarding customer data directly to their servers, which allows the company to maintain compliance with important financial data security standards and easily handle currency exchanges. In addition, the APIs offer such features as processing refunds, fraud protection, and analysis tools.

Mailgun (Email Service) – This service allows the company to avoid the issues involved

with creating its own automated email services. Customers require emails for important

aspects of website use including registration and order receipts. From the outset, Mailgun offers companies up to 10,000 free emails a month, with paid service for each email after

that.

2.7 – R

The company expects to derive revenue from rendering services directly. Depending on consumer need and company policies, customers will pay for services upfront, in installments, or at a lump sum upon completion, with more favorable payment terms granted to larger orders and orders with repeat customers.

Revenue streams are characterized as low profitability and high-volume. Therefore, the company’s main challenge is to achieve sufficient scale to overcome its fixed costs. Credit should be collected within 30 days of order completion. The smaller the typical order, the lower the risk of uncollectible debt poses to the company. As of right now, it is still a major strategic consideration whether the company will focus on a higher volume of easierto-serve, lower-expertise translations or focus more on businesses and more exclusive expertise needs in the manner of a boutique consultancy.

2.8 –

Salaries – By far the most significant expense for the company. Building full-time teams of

developers, salespeople, and administrators is the most significant early-stage cost.

IT Infrastructure - IT infrastructure costs include hosting, Domain Name Services (DNS),

third-party email API services, and other miscellaneous costs. The company domain name has been already registered and will not need to be renewed for two years. The application is built using open-source frameworks with permissive licensing schemes and have no costs associated.

Marketing Budget – The company must spend a significant amount of funds in order to

promote brand awareness and drive traffic to the website. The marketing budget includes things such as purchasing Google AdWords or ads on Youtube. The marketing strategy should remain focused on social media and search, since those are more likely to deliver click-throughs to the site itself.

Legal Expenses - The company should expect to need the services of an attorney from time

to time. In the beginning, these requirements are for such things as company registration, drawing up contracts for the first employees, and reviewing End User Licensing

Agreements. Expenses are considered to go up considerably should angel investors/VC firms show interest in the company.

Miscellaneous – Always present. Significantly missing from the financial projections in this

business are costs associated with office space rental. The workflow of the company from the start is being designed around the idea that remote work can and should be prioritized to achieve significant cost savings.

C

3 – P

F

O

The Wordparrot application in its initial stage is designed to provide four primary features for individual customers and B2B clients:

1) Convenience - Account registration via social media logins, file submission, and requesting quotes can all be done without any direct communication required with staff.

2) Security - Online payment through trusted 3^rd-party payment processors provides a variety of different payment options, fraud detection, and compliance with financial data laws.

3) API integration – Interacting with the agency, tracking order progress, and uploading and downloading files is encapsulated within the application rather than being fragmented across different unrelated services (email, cloud hosting, and traditional bank payments).

4) Quality Assurance – Service quality and accuracy can be signaled in a variety of ways through website features, such as translator profiles, progress bars, and the ability to communicate with translators directly through the site.

3.1 – R

Since the company is required to evaluate the needs of a client, it is necessary that a prospective client cannot submit files anonymously. Therefore, all clients are required to create an account. However, this requirement has its downsides. Users hate having to remember new logins and passwords for sites they may use infrequently or perhaps just once.

It’s a major inconvenience that will result in lost business for the site. Ideally, the registration process should present as little a barrier to a potential customer as possible.

Thankfully, the proliferation of popular social media websites such as Facebook allow our site to mitigate this issue. These sites offer public-facing APIs using the Oath2 Authentication Protocol. An important initial goal for the site is to offer potential customers the option of registering an account using their authentication data from social media sites. They will later have the option of adding/removing social media logins to their account, as well as to define site-specific login information if they wish. The company’s objective is to offer authentication mechanisms for Google, Facebook, Yahoo, Naver, & WeChat within the first 6 months of launch.

This authentication policy is subject to review at a later date, should the company find that account registration poses a significant barrier to new customers trying out the application.

Once logged in, the user is taken to the splash page. For now, the splash page contains basic information linking the user to different places in the site. In the future, we intend to derive marketing value from this front page. As the company’s ability to offer new and varied services grows, we can introduce our existing user base to bundling/upselling opportunities.

The updates feed is located to the side of the splash page container within the home page.

With this feed, the user can get up-to-date information about orders in progress that are pending on their account. The application has been designed to scale for organizations requiring multiple orders and language pairs simultaneously. If orders are small and only consist of a single document, then the update feature merely informs the customer and sends them an email when the order has been completed. However, with orders of larger complexity, the client can be informed when each document is finished, rather than waiting for all documents within an order to be finished and creating a lag time between completion of

3.2 – C

Clicking on the ‘New Order’ link brings you to a page that introduces the basic flow of the order process. Going from the first order form to the payment screen requires four steps:

a) Provide important background information about the order b) Upload all files one wishes to have translated

c) Submit a quote and wait for a response d) Approve and make payment

The order flow is intended to mimic the commonly known aspects of popular e-commerce websites that sell products online, such as Amazon. Whereas some translation agencies open potential transactions with an exchange of emails and discussion of the parameters of the job, the website’s order flow is intended to reduce the amount of time and man-hours required to negotiate over the job, determine a price, and arrive at a decision.

In the future, the company will have some sales people to handle these responsibilities, as some customers will certainly wish to maintain this one-on-one communication pattern.

However, the process by which the company interacts should be oriented around technological capabilities and the advantages in automation and convenience that they bring.

One of the difficulties of applying the e-commerce order flow to a service industry is that services are not mass-produced and standardized in the way that products sold on Amazon would be. The needs of each client will be different. Ideally, the company should be able to accommodate a wide range of service needs, from short translations requiring a day or less, to longer orders requiring multiple employees and payment broken out into installments. The preliminary order form attempts to categorize the needs of the typical client

as much as possible before order submission. For clarity, the choice of languages available for service are referred to as “locales”. Every order features one and only starting language, or “locale origin”, and either one or many target locales which constitute “locale branches”.

Each branch requires one or more translators to be assigned to it to perform work.

The form has a number of selection menus designed to categorize the type of service required. In addition to specifying the locales required in the order, a number of options are available. Examples include:

Type: Translate (Convert a document from one language to another), Review

(Modify and polish an existing document without translation), Create (develop a document from scratch)

Process: Same (polishing a previously translated document), Single (translation

between two languages), Multiple (From one language to many languages)

Field: Academic (high school or college essays and applications), Publishing

(graduate level or Ph.D work intended for academic journals), Signage (Restaurant menus, billboards, and other publicly viewable literature), Marketing (brochures, flyers, press releases etc.)

Speed: Regular, Express (Express option will move the job to the front of the queue for the company’s translators. More developments on this will come later).

Successful submission of the background information form will save the prospective order’s data to the database for the first time. The user can leave the website and return at a later time, and the order will appear in the shopping cart. However, the user is required to submit at least one document per target locale before being allowed to submit a quote. This document or documents must conform to a number of pre-specified formats, such as PDF, DOCX, TXT or XLSX.

In addition to the file requirement, the user is allowed to upload a number of supplementary files for each branch. These are files that are not to be directly modified by the company’s translators in any way, but are simply permitted so that the translator may have more context and information about each order’s service requirements. For instance, if the customer is submitting an order involving some file translation, the supplementary file aspect allows the customer to include a memo specifying the translation requirements that has been provided by his or her superiors, rather than needing to write down these requirements for the translator and needlessly repeating himself or herself.

3.3 – P

After the client has customized their order and met the minimum uploading requirements, they can submit this order for a quote. This will enable the quote to be viewable by administrators on the administrator-only portion of the site. Administrators can download the files, review them, and make a judgement about a reasonable price to charge the client. This payment can be arranged to take place upfront or structured as installments. When the price is assigned, the client can receive an email about this quote, and their front-page feed will update to show that this quote has been issued.

For the time being (and subject to change in the future), the client is not allowed to directly choose a translator through the website interface. There is also no negotiation feature outside of the website to discuss price, although these conversations can occur outside of the site.

This is also subject to change in the future. If the customer agrees to the quotes, they can enter an e-commerce-like checkout page and choose their method of payment. They are allowed to simultaneously pay for any and all orders for which they have received a quote.

Payments are handled through one of two payment processors, Stripe and Braintree. Both providers offer conventional credit card charging, as well as a number of web-based payment systems. Braintree (a Paypal subsidiary) offers payments through Paypal, Apple Pay, and Google Wallet. Stripe offers payment via Alibaba’s Alipay, which is a critical component in any strategy for attracting mainland Chinese customers.

These third party payment systems allow the company to achieve cross-border compliance with legal regulations regarding consumer financial security. They also employ the latest machine-learning strategies for detecting and stopping financial fraud. Finally, they

are essential for promising financial security to clients – payment details are forwarded directly to trusted payment processors and never stored on the company’s servers.

Once an order has entered the system, it will become an order in process. The administrator will determine the translators required for the order and assign them to the order.

From the point of view of the customer, the website’s layout allows them to:

a) List all orders in process or completed, searchable by a number of terms

b) Monitor the individual progress of each order and download files when completed by translators

c) Use a web form to contact the company with any questions/concerns

d) Browse the profiles of translators assigned to the order, which include resume information, some biographical information, a photo, and samples of previous work Translators can update their progress if the order takes more than a day or two, and the order will display a rough estimate of completion across all files in process. When each document has a file uploaded (to indicate completion), the user’s front page feed will update to attract the user’s attention that the file is available. Finally, the user will receive an email as well as any feed updates when all documents have been completed.

C

4 –

4.1 –

&

Unlike retailers selling physical goods, businesses that offer services to end consumers or other businesses often don’t know exactly what their costs will be, and therefore base their pricing on estimations. Translation is no different. The clearest method of pricing a translation service is to estimate the price of a document using a cost-per-word rate. This rate may be kept uniform across an order, or it may be discounted for larger orders.

Many translation agencies simply don’t offer information about their pricing structures upfront. For good reason – estimating potential labor cost (in man-hours) can vary highly even with documents of a similar length. Agencies making naïve cost projections based on document length may find themselves fixing up highly flawed documents at a loss. If they underestimate the appropriate price, the agency may also alienate their translator base by not paying at a value commensurate with the labor hours to produce a quality result.

The following is a small sample of translation agencies companies located in South Korea, Japan, the United States, UK, and Taiwan, and the prices they offer upfront on their websites. Although each company has its own method of categorizing orders, the statistics compiled here roughly correspond to each company’s ‘basic’ and ‘professional’ pricing tiers.

This data is not meant to offer a market-wide measurable average but simply to give the reader a glimpse into what kind of prices the market can currently bear (table 4A).