Introduction to New Age Digital Media Businesses

The increase in netizens and social media platforms along with the proliferation of mobile internet has led to an unprecedented surge in User Generated Content (UGC) creation and consumption. Per a survey conducted by Global Web Index on social media consumption patterns, 36% of the respondents use social media as a source of entertainment, while 56% use social media platforms for video consumption through social media platforms 

(YouTube, and Instagram’s IGTV being the forerunners). However, with no limit to the diversity of user-generated content, where most of it is unfiltered, consumption for entertainment poses a serious threat to audiences. Audiences are often at the risk of being exposed to anything from mildly abusive and offensive content to hate speeches and extremely violent, mentally disturbing content. In such times, it is imperative to check if the user-generated content promotes healthy consumption, or spreads malicious content. A preemptive approach to content moderation rather than a reactiveapproach will help prevent inappropriate content before it is uploaded – nipping it in the bud. Algorithmic advances in latest tech can help platform owners develop mechanisms to screen audience profiles, mindsets and act accordingly.

Platform owners and content moderation service providers have developed innovative solutions by leveraging Artificial Intelligence (AI) and Machine Learning (ML) along with technologies like Blockchain and crowdsourcing. These solutions help combat fake content (image, video, text) by using attributes like geo-location, source of occurrence, etc., thus reducing the time taken to moderate content, and improving efficiency and quality of moderation. Implementing AI and ML also helps significantly bring down the overall cost of content moderation. However, owing to the miscellany of user-generated content in terms of language, geo-political and cultural context, organizations need to be agile enough to build hyper-local moderation models compared to generic onsite-offshore models. It is essential to have associates culturally aligned to the kind of content they will be moderating. In this light, service providers need to revisit their current operating model for content moderation to ensure that associates handling specific geos belong to that particular geo. Just knowing the language is not enough.

Modern-day analytics can help platforms build audience-profiling systems that help the AI/ML tools understand their users, their propensities, and likes and dislikes. It will help them know what kind of content the user generally engages with, and would be able to predict his or her next upload or activity. Platforms could also explore the idea of taking users through a short series of questions that gauge their mental state and hence, are able to predict the content being shared. While these steps might reduce the need for content moderation in the first place and make platforms safer, platform owners need to strike the right balance between keeping checks and balances, and providing users a platform to express.

Our client is an American multinational technology company that specializes in web-related services and products. Its top-notch team of 89,000+ in-house employees did not have answers to all the questions all the time. This meant that they would engage experts to compensate for the lack knowl-edge in house. To ensure both cost-effectiveness and profitability, the companypartnered with Wipro for increased expertise and demand of services.  Wipro’s digital solutions enable Internet service providers to ensure that certain key processes are delivering efficiencies and allows them to concentrate on building their core capabilities.

We provide expertise, diagnostics, and models that help deliver on their customer value proposition as well as identify and rationalize operational investments to improve margins and process efficiency. A significant portion of our operations work for our client entails content moderation, flawless implementation on ads going live and unimpeachable transactions of end users. This would help them deliver sustainable, significant value for their own organizations, their business partners and their customers. While ensuring customer satisfaction and business growth, we face a number of challenges to cater to and come up with solutions and models that would both improve the existing processes and benefit the business. To be classified as a high-performance vendor, DOP relationship had to satisfy two criteria – meeting minimum requirements, as well as capturing value beyond cost savings. This requires effective strategies but also the right technologies and processes, effective organization and change management skills, industry and domain knowledge, strong leadership, savvy governance, and operational capabilities.

Ways in which New-Age Digital Media impacts PEx

Increased Focus on the Customer

With the advent of social media, the performance of companies andproducts has become more transparent. Customers are able to compare prices and review performances with the click of a button. That on top of heightened competition requires Process Excellence initiatives to be more customer centric. We have been proactive in identifying the need of the hour to ensure value-adds lead to end-user satisfaction. 

Demand for Faster Processing Capabilities

As new-age digital platforms provide more personalized and on-demand web services, along with faster deliveries, PEx initiatives must assist companies to deliver high quality and efficient output in as minimal time as possible. Margadarshak, a tool to identify slur words through the transcript of the video helps reduce AHT of the process, efficiency gain and improved quality due to reduction in manual intervention.

Rapid Project Delivery Timelines

Companies need to update business models and offerings to adapt to faster changing customer expectations in the new-age competitive world. PEx initiatives can no longer go on for six months or so. Policy Assistant, an interactive virtual trainer identifying grey area scenarios and helping agents during doubts was to ensure highest quality delivered for machine learning with least errors. It was implemented in the form of a decision tree for five difficult policies within a month’s time.

Sometimes You Need More than just improvement

Process Excellence teams need to be able to distinguish between processes that merely need improvement and those that need total reinvention. Also, the training and wellness needs of employees working for these companies have to be kept in mind while designing these improve-ment projects. Wellness chat bots are designed to ensure wellness of employees as they are subjected to restricted content umpteen number of times and would help improve the system psychologically. They also ensure a holistic change in the mindsets of individuals.

New Opportunities for Automation

The new digital wave has made computer systems more sophisticated. Artificial Intelligence tools and advances in robotics are enabling those intelligent systems to be applied in new ways. Now, more tasks and processes can be digitized and automated even in areas that were previously seen as untouchable (because they require human judgement, for instance). Prime is a tool that uses a set of intelligent algorithms to review the metadata of the documents sent by the customers. It helps in auto validation of e-KYC documents and is able to differentiate between fake and genuine documents helping in identification of fraudulent customers and saving millions for the client.

Greater Process Complexity and Risk

Process Excellence teams need to be able to distinguish between processes that merely need improvement and those that need total reinvention. Also, the training and 

wellness needs of employees working for these companies have to be kept in mind while designing these improvement projects. Wellness chat bots are designed to ensure wellness of employees as they are subjected to restricted content umpteen number of times and would help improve the system psychologically. They also ensure a holistic change in the mindsets of individuals.

New Skills for PEx Professionals

All of the above means that just getting the Six Sigma black belt is not enough. New skills and awareness of other disciplines is becoming more and more important. Change management and other "soft skills" are critical for managing the people side of process change, especially during a time where these changes are coming thick and fast.

Transaction Monitoring

The relationship between vendors and clients required meeting basic financial objectives, performing at desired service levels and delivering consistent and predictable service. Having better metrics in place to measure process performance is a critical piece to the overall puzzle. The partnership has enhanced our experience in measuring end-to-end processes better as well as in including the provider in these processes. Quality is paramount in the content moderation space as the topics go live to audienc-es of millions with just one click. It is necessary to have a structured approach around measurement of quality so that errors are minimized and right transactions go live.

Currently the approach for measuring quality is two pronged. QAs perform blind reviews on the transactions and pass a decision based on their subject knowledge. The decision is matched/mismatched with the reviewers’ decision, which is validated further. Any discrepancy between them is considered an error. SOP for quality management is defined and the approach is followed diligently across processes.

Challenges

• Few checks on QAs concerning his/her subject matter knowledge and feedback delivery skills. In some cases, 5% audits of QAs are done by the MQAs
• Feedback documentation in terms of acknowledgement by reviewers is not mandatorily captured
• Grey area scenarios are only understood through client partners with no tracking of the same
• Training needs identification in terms of bottom quartile sampling and error trends are not done thoroughly
• RCA on all errors and defects are not carried out
• Sample size calculation is done per client needs
• Repeatability and reproducibility between QAs and MQAs are not being measured

Solutions

• Certification of QAs on feedback and coaching skills to be a criterion to measure their soft skills. A 90% score to be set as the target to reduce the variance between QAs and MQAs. Frequent calibrations between MQAs and QAs would be necessary so that there are less appeals and mismatches between QAs and reviewers.
• 100% feedback within 72 hours to be a compulsory requirement so that reviewer errors can be targeted not to recur after 3 days. Repeated errors by a single individual over a week to be tracked and performance improvement plan for the agent to be monitored and levied.
• Grey-area scenarios repository to be maintained after being validated by MQAs on the sanity of the same. Detect if any cluster signals are available so that a questionnaire can be made on the same that would help reviewers in guiding them to the correct answer. A virtual interactive training module   can be maintained for helping reviewers mark  the correct decision for a review.
• Stratified sampling on the basis of quintiles(for internal purposes) so that training needs can be identified and can be taken up for specific reviewers who have been weak in particular topics or policies over a period of time.
• RCA framework defined in terms of framework used for approaching a problem to be designed should be captured for any errors or mismatches between QA and reviewers along with MQAs and QAs. Every error or defect to be considered as a major error and RCA should be designed on the same so that reviewers get a clear picture on what and why a particular transaction went wrong. Agent, Process, People and technology error (ACPT) to be captured in the RCA. Below figure depicts it:

• Sample size is calculated per client protocols. We cover all the agents through internal sampling. Random sampling is used to cover entire agent span including greater samples being checked for bottom quartile folks. Predictive modeling can be used to define probability of failure/error in cases with a definite combination of factors. Firstly, the factors affecting the output Y are determined through chi square test considering the Xs to be discrete. Taking into consideration only factors significantly affecting the output, chi square distribution tree (CHAID) is drawn to  assess the probability of error possible for different factorial combinations. This way we would be able to churn out samples that have high probability of error, which would lead to increasing accuracy of decisions taken in the long run as soon as feedback is given according to the factors considered as affecting the output. A DMAIC approach can be  taken for increasing quality based on the  predictive modeling done for different combination of factors, which would eventually  lead to increased customer satisfaction.
• Variance between MQAs and QAs to be tracked for at least 5% of the audits done for QAsas well. Calibration should take place between MQAs and QAs regarding grey-area scenarios and cases where appeals are high, discrepancy between QAs and SMEs can be taken up as well. QAs performance stack can be monitored according to this parameter to adhere to this guideline.

Automations

In earlier generations of BPO, the focus was primarily on greater efficiency: standardizing and streamlining operations, running operations better and faster, and driving down cost wherever possible. However, companies can only pull the same levers of labor savings and process improvements for so long before the benefits sound to be too monotonous. Now, the legitimate question clients will be asking their vendors is: “Is there something else you can give me?” That “something else” turns out to be the ability to use in-depth domain and subject matter expertise and the ability to analyze data and functions about how and where RPA and automations can be driven to more predictably drive business outcomes.

As savvy clients look for differentiated technology capabilities from their providers, we came up with automation solutions for the client, that would not only help with reducing AHT of the processes, but also lead to efficiency gain and productivity savings without any harm to quality. Below are the some of the suggestions that we came up with that would help empower the business and ensure high performance:

Margdarshak

The aim is to improve customer experience by reviewing flagged content and taking appropri-ate actions like Parental Guidance, suitable for minors, safe for age group 5 to 7, mature adult content, etc. While reviewing the content for violations, the reviewer needs to vet the title and content of the video for profane words based on which CR1 / M10N rating is determined. Failing to identify the sensitive words in the metadata would lead to P0 escalations. This extension would help to list out all the profane words in the video, highlight sensitive words, and reduce escalations, which would lead to reduction in error rate and improvement in AHT and efficien-cy gain. Error rate reduction from 23.9% to 12.2% took place and 7 FTE gain was delivered to the client.

Policy Assistant

The process involves reviewing ads and enabling machine learning. It requires that the machine learn accurately about particular ads. Doubt clarification for grey-area cases with QAs and trainers usually takes a lot of time, which also involves high error rate due to wrong verdicts taken in identifying grey-area cases. 

An interactive navigator and virtual trainer was designed with the primary focus on policy-spe-cific grey-areas and top defect contributors to achieve consistency in decision-making and bridging the gap between policy page and reviewers. 

The modules are created in such a way that dummy cases can help a reviewer take a quick decision instead of waiting for SMEs to provide clarification on the subject. This led to improve-ment in FTR from 97% to 99%.

Wellness Chat Bot

The aim is to review flagged videos and take decisions on continuing to stream video online or whether some other action is required based on video content and policies defined by the client. The content of videos uploaded and flagged can sometimes be very disturbing and can impact the mental well-being of the Admin reviewing it. The Wipro team, via its engagement with the Wellness Manager and a Psychologist, has taken various wellness initiatives, but these resources might not be connected with on a 24/7 basis. Hence, it is imperative to identify ways to engage Admins in other wellness activities that could address their stress levels. 

The idea was to introduce an AI-based Digital Wellness Advisor (Chat BOT) that interacts with Admins on a proactive frequency or when the Admin seeks assistance. Admin’s stress level will be identified based on inputs such as text, voice, and pre-defined workflow solution. Conversations, puzzles, one-minute activities (as per suggestions from Wellness Advisors) and other engaging activi-ties are incorporated in the BOT to help reduce the Admin’s stress levels. It will also help to identify potential risks if the Admin shows early signs of being mentally disturbed.

Buscar

In this workflow, primary email/secondary email and creation date are identified and duplication of content is checked on three different tabs. All pages are checked for policy violations and then the review is submitted. This process involved multiple repetitive steps. Reviewers had to copy multiple information on different tabs for single review, which was an NVA activity. A Chrome extension was developed that would look for various information on the page and highlight it. It will also share the summary of the number of times the information is repeated. This reduced toggling time as well and led to 35% reduction in review time and efficiency gain of 9 FTEs.

Marvel

In this workflow, Sender & From Header lookup are performed manually for each review. URLs are launched by manually copying and pasting it into a new browser window. This process involved multiple repetitive steps. Reviewers copy multiple information for a single review and use Google to search for it. They also had to open the links manu-ally in the review by copy-pasting the links into a new window. To resolve the issue, a Chrome brows-er extension was developed that would copy various information in the review UI and search for results and also auto-open the desired links and use Google to search the sender’s address each time. This acts as an auto launcher for the URLs leading to efficiency gain of 11 FTEs. 

These automations have made our lives easier and paved the way for increased innovation and proactive resolution of issues. It has served as a value add for the customer in an elegant way and has brought in continuous improvements in the processes with enhanced results. Every inch of enhancement that we do leads to a wave of awareness and interest, which results in enthusiastic minds working to find realistic solutions to every problem.

Conclusion

Process excellence for media businesses has paved a new way of looking into innovations in the industry. Apart from historical audits to continuous improvement initiatives, opportunities in terms of automation and man-agement of metrics through different form of analytics have helped see a new picture of the business. There is great promise identified in our high-performance media accounts with greater focus on how to look at different metrics and how to define measurements in the content moderation space. With increased usage of social media, internet and video platforms, its necessary to be an innovator with each passing moment than being a follower or a laggard. Identify, gauge and implement should be our major focus while implementing brand new solutions for the 21st century client. Everyone needs just-in-time solutions with least defects. We have to ensure we tread that path along with all our clients. 

References

https://www.backofficepro.com/white-paper/business-process-outsourcing/

Upalika Palo

Assistant Manager, Quality and Process Excellence Wipro

Upalika has over two years of experience in Process Excellence, Six Sigma and Lean. Her experience spans across leading global organizations and industries such as media and telecom with a focus on process improvement methodologies and use of technology solutions for effective process management.