The international insight pipeline

x Sponsored content on Research Live and in Impact magazine is editorially independent.
Find out more about advertising and sponsorship.

Consecutive years of worldwide disruption have sent shockwaves through global economies and socio-political systems. Labelling this a period of Vuca (volatility, uncertainty, complexity and ambiguity) is perhaps an understatement. Business publications, thought leaders and other members of the commentariat are generally in agreement that, in these tumultuous times, businesses must be agile if they are to survive and prosper – and agility is underpinned by technology.

That’s as true for research as it is for other sectors, but when it comes to conducting international and multi-market research, the challenges are magnified. Language is the most obvious potential stumbling block, but there is a host of others, from varying cultures and consumer habits, to differing stages of economic development, to consistency in data collection.

Research technology – or ‘restech’ – has a growing role to play here; for example, by automating the sample process or offering clients ‘do it yourself’ access to data collection and insight. A sophisticated market research tech stack potentially offers the opportunity to conduct research at a scale and speed that was unimaginable until quite recently. That’s an enticing prospect, particularly for those involved in cross-border research, yet it hardly needs saying that researchers and clients must ensure rigour and accuracy are not sacrificed in a dash for quicker results and greater efficiency.

Advances in translation

Translation technology (transtech – see boxout on terminology, below) increasingly figures in international research projects, and getting reliable and useful results calls for considerably more than simply bunging some text into Google Translate.

Ruth Partington, founder and chief executive at insight localisation agency Empower Translate, asserts that there are commercial benefits to having some knowledge around transtech – for instance, in understanding terminology such as translation memory (TM), a database of previously translated text.

If you researched digestive biscuits, for example, and now want to research shortbread, there are obvious shortcuts in accessing this previous translation work. Some of the questions (and indeed answers) might be the same, so there are time and money savings to be had. Yet, Partington cautions, the savings of automation will not be as great as some researchers might hope, and there is still a necessary role for human quality control.

Consistent effort is needed to drive bottom-line results from TMs. However, explains Partington, TMs get “dirty” when approved translations are added to the bank and become mixed in with similar translations. If they are not regularly maintained by a specialist linguist, they will produce faulty translations.

“You can decide on what percentages and sensitivities you set and that speeds up the translation process reasonably significantly, cutting down the human-in-the-loop translation time by, maybe, 50%,” says Partington. “But where the real gains are for the research element is when transtech is properly layered into a project and into a client’s workflow. You can cut a timeframe for turning around translation by more than half.”

Working with InSites Consulting (now Human8 ), Empower reduced by 50% the time to market for multi-market medical surveys for a pharmaceutical client. At the centre of this was a cloud-based portal designed to eradicate some translation pain points for global researchers. By examining a cross-section of quotes produced by Empower’s team for research agencies over the years, the company saw the opportunity to create an algorithm that would produce instant quotes for global research projects. This reduced the time taken to commission from between four and 10 hours (depending on the project details) to 60 seconds: the time it takes for a client to upload its files to the portal for analysis and click the ‘confirm quote’ button.

Rebecca Cole, managing director at Cobalt Sky, currently chairs the MRS accredited company partner council. This is working on a project, headed by Partington, that aims to release a standard set of questions that research buyers can ask of translation agencies. Increasing transparency around transtech will be a big part of this initiative.

As someone who procures translation services and implements them into online surveys – as opposed to being a translations specialist directly – Cole considers the main effects of recent advances in transtech to be an ever increasing need for trust in, and a good relationship with, translation agencies. In her view, it has become more vital for the industry to provide easy ways for buyers of translations to differentiate the good from the bad.

“Transtech advancements have meant it’s sometimes a bit of a Wild West out there,” says Cole. “As with any technology, used correctly and in the right hands it can provide huge benefits to all parties: reduced costs, reduced time to market, stronger and more reliable results. But, in the wrong hands, or people using DIY tools such as Google Translate, it can lead to incorrect and lazy translations, longer time in the field because of increased respondent dropout, reduced quality of responses because of reduced engagement, and lack of professionalism being shown to respondents. Research buyers need to first identify who the quality translation agencies are, and what questions to ask them to find out how that company is using transtech, and to what extent.”

Not everyone uses translation agencies, however. Discover.ai, a tech based insight start-up drawing qualitative and cultural insight from online text-based sources (for example, blogs, forums and social media), focuses on multi-country projects and harnesses transtech, its own artifical intelligence (AI)-based model, and a network of local partners to access insights from multiple markets in their local languages. For example, it has partnered with PepsiCo in diverse cultural spaces, including hard-to-reach markets (the Philippines, the UAE and Honduras). This allows PepsiCo to access a deeper understanding of consumers’ needs, behaviours and attitudes in different geographies.

The platform uses Microsoft Translate for its automated translation services, with translation done as the data is collected. The English-language content is then run through the firm’s algorithms and analytical frameworks, before being analysed by strategists.

“It’s not just about opening up lesser-researched markets; it’s about getting cultural and expert angles, plus accessing hard-to-reach audiences with low incidence rates; they may be hard to recruit, but they often talk a lot online,” says Jonathan Williams, founder at Discover.ai. “There are limits here, though – but that has less to do with technology and more to do with cultural understanding. Agile solutions will give a good multi-country overview, but the more you can get this delivered by people with local cultural expertise, the more value you will get from it. There comes a point when full local research is needed to land strategies in a locally relevant way.”

Ensuring cross-border transparency

One common worry clients may have regarding international projects is inconsistency around methodologies across markets. Even if a research partner has respondent pools across a range of regions, they may use different methodologies or fulfilment companies to source respondents in each region.

Jamie Parks-Taylor, director of insight and analytics at advertising agency Cream, has concerns about discrepancies in approaches to sampling between markets that undermine cross-border comparisons. To improve agility without sacrificing quality or accuracy, Cream works with partners with on-the-ground support in each market – for example, Kantar’s software-as-a-service (SaaS) offering, Marketplace. Parks-Taylor sees increased demand for restech solutions from clients who are a little more sophisticated in terms of research understanding, or clients that operate in industries that are more inherently “fast-paced” or tech-savvy. Innovations in restech, such as DIY SaaS platforms and research tools that use AI, are often requested by tech clients who want results quickly and understand the pros and cons of AI in the research domain.

“The adoption of AI in research solutions is exciting,” says Parks-Taylor. “If predictive AI can replace the role of a human respondent in a survey, it has the potential to offer insights that are just as reliable and accurate, while saving time and money if the results can be delivered faster, with a smaller cost incursion that can be passed on to the client. But that’s a dream scenario.”

Carlsberg Group vice-president of insights and analytics Nick Rich agrees it’s an exciting time for insights professionals and insights function leaders. “It feels like we are leading among all marketing functions in the progress we are making to better operationalise and realise tech solutions into our function and ways of working,” he says.

With restech solutions appearing across almost every element of the research process and insights value chain – from data collection to analytics, through insights communication and longer-term knowledge management – they are resolving legacy pain points while addressing inefficiencies in the research process.

Trends observed by Rich are: restech that drives quicker and comprehensively uniform multi-market data collection; restech that enables enterprise data management and analysis; and restech that drives more effective knowledge storage and accessibility, such as Stravito, which Carlsberg uses. In the past, Carlsberg has been “somewhat underwhelmed” by certain AI solutions. However, Rich expects further improvements will probably have a profound impact. “Right now, that looks like solving some of those same legacy challenges we have always had – synthesis, summarisation, simplification, communication, and shortage of ‘time’.”

Technology as a driver of international affordability

Tim Wragg, chief executive at Hall & Partners, thinks the biggest trend from the rise of restech is a levelling of the playing field. With the right tools and talent, a direct-to-consumer start-up, for example, can make an ad, research and assess it with the same level of accuracy and insight as a giant fast-moving consumer goods player. In essence, AI and machine learning (ML) are democratising research at a market and cross-border level because the price point is affordable for everyone. “The big brands have to find competitive advantage from their scale in other ways now, whether that’s first-party data or something else,” says Wragg. “But no-one can afford not to take advantage of the restech revolution.”

Hall & Partners recently worked with the global insight team at a delivery company that sought to improve the immediacy and impact of insights from across the business. It wanted to accelerate the speed of reporting from three different trackers and find an alternative to time-consuming PowerPoint reporting and data requests.

Hall & Partners combined data from the trackers to create dashboards accessible to users across the organisation. The team used the information to publish internal articles inspired by important themes and findings. Once use of the platform became widespread, search, spend and app download data was added, and the audience widened to encompass every tier of global management. Sharing stories from different markets helped identify common problems and highlighted cultural differences for leaders to factor into their decision-making.

Phebi, whose software is used to find insights in voice and video data, has seen an increase in multi-country qual and customer experience work where clients are looking for a consistent, quantifiable way to do analysis across markets. On a project basis, says chief executive Mike Page, multinational agencies are collecting data via audio and video, and providing Phebi with audio or video files. The software then does machine-based speech-to-text transcription and translation, and makes them instantly available – along with ‘nonconscious emotion scores’ – to local and centralised team members via a portal, for analysis and for developing presentations that include highlight reels of people speaking.

“The most forward-thinking agencies are looking for an easy way to process, analyse, store and revisit all of their human speech assets for studies and trackers, and across studies,” says Page. “They typically want to be able to interact with the data in its original language and in a second language, usually English. Using technology and a centralised approach can result in more consistent and seamless analysis across countries, especially for text-based analysis of the words people use and for voice-based analysis of people’s nonconscious emotions.”

Phebi sees human speech as the richest source of insight on both a conscious and nonconscious level. Page points out that the technology needed to run smoother voice-enabled multinational projects, and gain new insights from them, is available and there is already an uptick in its use. Voice-enabled research software helps collect high-quality responses from often underserved groups with less access to technology, lower or no proficiency in writing, or difficulty communicating using keyboards. It is integrated with tablets and mobile phones used in the field, so respondents can use their own words.

“The transcriptions and translations are fully handled by software, quickly bringing quality results back to local and centrally located researchers in clear, usable language,” explains Page. “In addition, when nonconscious emotions are detected from speakers’ voices, researchers can tap into a new set of insights without hooking up specialised devices that early forms of nonconscious response measurement required.”

Insight Engineers managing director Jeff Deighton says the rise of harmonisation specialists, such as Wessex Insights, indicates a client need for harmonised data. On a separate point, researchers must make contingencies for survey platforms that only operate if connected to the mobile network and so can’t be used in rural or remote areas. Insight Engineers has encountered this problem on projects in Africa. The solution? Human interviewers and SurveyToGo on their laptop or tablet, syncing at night or when they are home and have Wi-Fi, instead of live using the Qualtrics data-collection platform.

Insight Engineers does a lot of online quant and qual work, and one of the trends Deighton has observed is a rise in “multi-mode projects and more do-learn-do, reconvening participants from phase one of a research project to show what has changed from that input and using them to hone further”.

Poor AI translations continue to be a “blight”, says Deighton. Instead, the company translates initially into other languages with professional translation companies and then back translates with an AI into English, to sense check.

That said, there are also positive signs of transtech improving in the past couple of years. Online qualitative research company High Tech Developments (HTD) recently ran sessions for the World Health Organization, which wanted to listen to and engage with a multilingual panel of doctors who were frontline Covid-19 responders. Japanese, Korean, Chinese, Taiwanese, Australian and Singaporean doctors had their own simultaneous translation so they could listen to everything being said – and so could the observers.

HTD offers AI, personally identifiable information-redacted transcripts direct from the Amazon AWS network. Chief operating officer and co-founder Rob Wallis says: “We also use AI to detect filters used by respondents to mask their true identity.” To date, Asia is the only part of the world in which HTD has discovered identity masking. “Quant and qual are merging and a lot of traditional researchers are not happy,” says Wallis. “The biggest reason for this is that the developers who design the platforms have no true understanding of what both worlds involve.”

It’s up to researchers to get on top of these issues. That is happening already. As Empower’s Partington explains, the insight industry is now “much more interested in intelligent integration” with partners who have a strong value proposition. Often, that will entail marrying restech and transtech to get the best of both worlds, settling on a smart balance of automation and human expertise.

This report was first published in the April 2023 issue of Impact.