You are on page 1of 7

How Machine Learning is

Transforming International
Market Research
Only a decade ago, self-driving cars and voice assistants that tell you the weather would have
sounded like something from a sci-fi movie. Today though, artificial intelligence (AI), and more
specifically machine learning, are rapidly transforming our lives and the way we do business.

Machine learning has revolutionised industries from manufacturing and healthcare to retail
and financial services. Innovations have given organisations the power to dramatically improve
efficiency and ultimately increase profitability. According to a survey by McKinsey, 45% of
executives who haven’t yet invested in AI are afraid of getting left behind by the competition.

The international market research industry is undergoing a similar transformation, with AI at the
heart of it. Insights that previously took hours or days to generate can now be produced in a
matter of seconds.

In this article, we take you deeper into how machine learning is being used in international
market research and explore its impact on translation. We also look to the horizon and ask the
question, “Will machines replace humans in international market research?”
Let’s dive in and find out.

Where are we now?

Historically, international market research has So where are we at the moment? These are
been a slow and expensive process with the developments currently leading the way in
language barriers adding to the complexity. AI international market research:
is changing this by automating repetitive tasks
to allow people to work more efficiently,
augmenting the translation process, and Machine translation of data
generating enhanced insights from vast With the rising importance of content
amounts of data. localisation and increasing volumes of big
data, a need for cost-effective, fast translation
There’s a lot of buzz around what may be is driving demand for machine translation. The
possible with machine learning – automated technology is now at a point where it is
facial coding, chatbots, automated sentiment achieving meaningfully high levels of accuracy
analysis, machine translation, and real-time in and is being employed in the translation of
language coding, to name just a few. While open-ended responses or verbatims, among
some of these approaches are commercially other uses. Although not yet able to deliver
available, the reality is that some are still a 100% accuracy, machine translation plays a
long way from being ready, and others fall crucial role in intern ational market research.
somewhere in between.

2
You now have access to significantly greater volumes of
raw data. Machine translation allows you to process it
and select the pieces of data that you really want to go deep
on and improve quality. It gives clients access to more
information, more quickly.

As Romina Franceschina, Business Solutions they arise. Once sufficient data is inputted,
Director at Language Connect, explains, the model is able to suggest codes to assist
“You now have access to significantly greater the linguist, automatically apply codes, and
volumes of raw data. Machine translation highlight areas that may be of interest.
allows you to process it and select the pieces
of data that you really want to go deep on and Coding the data directly in the native
improve quality. It gives clients access to more language is hugely beneficial as it removes
information, more quickly.” In a highly the need to translate data into English before
competitive industry which is dependent on it can be investigated. Not only does this
speed to market, this creates a competitive reduce turnaround time, but it also increases
advantage for those who use it. accuracy as any nuances in the responses are
not missed.

In-language coding in real-time


Steady advances in Natural Language Social media analytics
Processing (NLP) mean that models capable In an effort to understand today’s ever-evolving
of understanding both written and spoken digital consumers, market research agencies
words are becoming more reliable. This are seeking to combine the old with the new.
makes it possible to rapidly process Research methodologies that make use of
open-ended responses on a large scale and machine learning are mining social media and
automate what would otherwise be a chat apps to better understand consumers’
time-consuming, manual process. thoughts, feelings, emotions, and ideas.

Machine learning is used to create dynamic Christelle Muller, Associate Director, Business
models which can translate and code survey Solutions at Language Connect explains,
responses in real-time. The model is trained “Agencies are now able to do brand analysis
by the input of expert linguists who perform via the information they find on social media,
in-language coding on survey responses as rather than asking a question directly to the

3
respondent. Often what they'll do is askthe
respondent and also look at social media,
and create a combined response.”
It still doesn't understand
Using a ‘no-questions’ approach in this way, the difference between
market research agencies are able to when we're talking about
gain deeper insights without leading or
influencing responses.
streaming on our TV versus
a river stream. It just can't
decipher the two.
The benefits are substantial

Although it’s still early days for AI in the


international market research industry,
the benefits for both agencies and clients But what are the challenges?
are numerous and substantial.
AI has considerably shortened certain
“The main benefit is return on investment”, workflows in market research, however,
says Romina. “Far greater volumes of data automation is very much a work in progress.
can now be processed and time to market
has significantly reduced. And then there The most significant hurdle is that manual
are the cost savings, of course. You can no intervention is still required. NLP can quickly
access more data with either the same or process high volumes of data, but cannot
a smaller budget.” capture the full complexity of a language.
“Take translating questionnaires for example.
In addition to the ability to do more with less, You need to be able to capture the voice of the
machine learning algorithms are able to researcher who’s asking the questions
uncover previously unseen patterns in data and artificial intelligence isn’t ready for that
that far exceed the ability of humans using yet,” explains Mickey Brooks, Senior Vice
traditional methods of analysis. President of Sales at Language Connect.

4
“It still doesn't understand the difference Although the insights industry is under
between when we're talking about streaming pressure to deliver cost-effective results more
on our TV versus a river stream. It just can't quickly, it is imperative that these gains do not
decipher the two.” come at the expense of quality if we are to
keep moving forward.
Accurate translation requires in-depth
knowledge of a language, its syntax, grammar, “Poor experiences are going to keep people
and any subtle differences that can drastically from wanting to adopt this new technology,”
change the meaning. explains Mickey. “Market researchers are
traditionally quite slow to adapt to new
Christelle agrees. “At the moment, machine methodologies because they are so
translation will more or less translate word for accountable. They’re hesitant to take risks on
word. It doesn't catch any of the cultural new technology and new methodologies
aspects of what's been said and doesn't pick because if it doesn't go well, their clients are
up on subtle nuances. So for example, if you going to lose confidence and ultimately might
have a French speaker who's asked to go somewhere else.”
translate verbatims from French to English
and there is a certain phrase that's been said in “It's also potentially linked to the pressure and
a context specific to his market, they will know the timings agencies work under. It needs to
what is actually meant by it. Whereas at the be right first time,” adds Christelle.
moment with AI, we're just translating word
for word.”

5
So what does the future
look like?

Based on its research, Gartner estimates that overcome before AI provides an established
AI-based analytics software will free up as solution; even then the perfect solution will
much of one-third of a data analyst’s time, likely always be a combination, a balance,
allowing them to tackle bigger projects or between machine and human actions.
other tasks.
People are very much key to the process.
A host of exciting developments are on the Without humans to ask the right questions,
horizon that promise to make market research feed in the right data, choose the most
faster, more efficient, and more insightful. Now suitable algorithms, and train the models,
is the time for organisations to start preparing machine learning simply won’t provide
and adapting to realise AIs full potential. Those useful results.
who don’t, risk being at a significant
disadvantage to those who identified The goal of employing AI in international
opportunities and experimented to get ahead. market research is not to replace human input
but to complement it and maximise efficiency.
Given the enormous potential of AI, Moving into the future, the key will be
the obvious question is “will machines understanding how AI and humans can
eventually replace humans in international augment each other’s strengths. “Machine
market research?” learning projects are tackling the ‘lights out’
type of workflow steps which presents a huge
The short answer is no, at least not in the near opportunity for humans to focus on high-value
future. As Romina explains, “We have all these tasks and let the machines do the low-value
wonderful developments, but as an industry, tasks,” says Romina.
we are still very much in the early stages of
exploring and experimenting.” Despite the While we can only guess at the long-term
exciting opportunities that the technology effect that AI will have on the international
presents, there are still many hurdles to market research industry, one thing is clear –
humans will be key to success.

You might also like