COMPANY CONFIDENTIAL. DO NOT DISCLOSE.
More than skin deep
Avon’s community tradition is more than just a good look – it’s good science. With text analytics tools in JMP® and JMP® Pro, R&D scientists mine consumer insight to inspire the next generation of beauty products.
_____________________________________________________________________________________
Challenge: In the beauty industry, some of the richest consumer insights come in the form of survey comments, product reviews and testing feedback. Unstructured text, however, has historically gone underutilized due to the time-consuming nature of parsing qualitative information by hand.
Solution: R&D scientists at Avon are now using text analytics tools in JMP® and JMP® Pro to tap previously opaque consumer data streams. The insights they generate through statistical approaches guide decision making in product development. “Using JMP allows us to take away our lens of bias and look strictly at the themes that are emerging from the consumer,” explains Senior Chemist Elizabeth Knapp.
Results: By putting JMP and JMP Pro in the hands of scientists, Avon has empowered them to maximize the team’s consumer research spend and become more nimble in applying consumer insight to formulation work. Text exploration and sentiment analysis have helped scientists to identify new applications for proprietary technology, identify improvement opportunities and optimize the entire R&D pipeline from blue sky to development to inline commercial product.
_____________________________________________________________________________________
With 135 years of cosmetics expertise, Avon knows its audience. The iconic beauty brand has long celebrated women’s voices and supported causes near and dear to the hearts of many: breast cancer research, anti-hunger campaigns and entrepreneurship. These values have served to craft Avon’s story into one that is today well-known around the world. Less widely known, however, is that Avon’s attentiveness to the consumer has also shaped the science behind its breakthrough beauty.
“Cosmetics are by nature emotional,” says Ashley Hutson, Principal Scientist at Avon Products. Hutson, who leads the color emerging space in Avon’s product development, says the most successful cosmetics are those that evoke an emotional response. Just as modern notions of beauty have come to represent many things to many people, consumer appetites are equally complex. Understanding what consumers want allows Avon to remain nimble in the beauty industry’s trend-driven marketplace.
Text analytics unlocks key insight from previously unutilized data streams
Consumer insight data comes to Avon in many different forms. Some feedback is actively solicited from consumers through surveys and focus groups, while some is collected from online product reviews or other customer support channels where buyers share their opinions organically. While traditional quantitative surveys are easier to parse using standard statistical methods, Hutson says the consumer data with the richest insight is often qualitative, unstructured text.
“JMP has been critical in terms of allowing us to maximize our time, reading through survey results and looking at responses on an individual level. By adding stop words or filtering the data in different ways, we can get to something meaningful and identify what the consumers are really saying,” explains Hutson’s colleague Elizabeth Knapp, a Senior Chemist with Avon’s upstream Applied Research and Formulary Design Lab. Vetting early-stage ideas by looking at consumer insight in JMP, she says, is what drives Avon’s product lines into the future for a better customer experience.“The trick is figuring out how to distill all this [text data] down and do it quickly without losing sight of the sentiment that consumers are trying to convey,” she explains. Until recently, much of Avon’s text data was underutilized – or not utilized at all – given both the time-intensive nature of analyzing unstructured data and the format’s susceptibility to research bias. The underutilization of any data, however, can come with a real bottom-line cost. That’s why Avon has adopted JMP®, an industry-leading software with robust text analytics capabilities that enables scientists to open up previously untapped data streams.
A unified data management strategy enables R&D scientists to maximize their consumer research spend
Avon has long commissioned third-party vendors to gather consumer research across its worldwide markets. The company’s move toward an increasingly unified analytics culture and their deployment of JMP, however, has motivated managers to bring much of this work in-house. Instead of budgeting for costly external contracts, Avon now invests a fraction of the cost each year in JMP licenses that put analysis in the hands of subject experts with deep product knowledge.
This democratization of analytics has empowered the R&D group to maximize their consumer research spend by designing survey instruments in a way that can be applied to multiple projects in the same product category. “Using JMP and being able to do [all the analysis] within our upstream group enables us to plan out studies in a way where we can really maximize those testing dollars,” Hutson explains.
Keeping consumer research in-house also affords teams more control and ownership of the data. “By incorporating JMP into our online survey tools, we're able to collect that data all in one place and then sift through it to either correlate different data that we're collecting in the lab, or use JMP text analytics,” Knapp adds.
Word clouds help explore potential new applications for Avon’s ProtinolTM technology
Working at the blue-sky stage of Avon’s R&D pipeline, Knapp explores how emerging technologies from the beauty or beauty-adjacent industries can be formulated to improve product aesthetics and enhance performance. It’s a creative space where she and her colleagues go out into the world in search of inspiration and put unconventional ideas to the test.
One of the team’s latest breakthroughs is Protinol™, a technology designed to restore collagen in damaged skin for a healthier, younger look. In the wake of Protinol’s early success on the market, Knapp is now evaluating new ways to leverage the technology across other product lines – a process that involves both iterative laboratory experimentation and consumer research. Her goal is to identify connections between the molecule’s biological endpoints and consumer perceptions of its value.
“I've been able to look at open response questions in [our online survey tools] to better understand the benefits we're driving with [Protinol] and the consumer perception is of this space,” Knapp explains. As groups emerging from the data begin to unite around particular likes or rationales – skin firming or all-natural ingredients, for example – she can then target a product proposition in her formulary work.
Knapp’s surveys put a product idea in front of the consumer and ask them to evaluate the pros and cons. “To have that visualized in a word cloud helps us see what we need to work on and how to prioritize,” Knapp explains. “Some things really pop out visually and help drive the conversation: how does what the consumer is telling us fit with the overarching company goals? It helps drive towards maximizing our testing dollars.”
Sentiment analysis in JMP Pro yields insight into polarized product reviews
As one of Avon’s JMP Pro users, Knapp also uses the tool’s sentiment analysis capability to generate more nuanced consumer preference insight. Sentiment analysis combines text analysis with predictive modeling to determine which words and phrases are most relevant to the problem at hand, using data to evaluate both the direction and strength of associations between terms. This approach, Knapp says, is imperative because it eliminates the need for researchers to make inferences as to how consumer appetites are grouped, as would be necessary if assessing qualitative data by hand.
Sentiment analysis is especially critical in situations where consumer opinion is polarized. Understanding what is creating the polarization around a product can help the team to identify opportunities for improvements in both product quality and packaging optimization.
“Having JMP Pro lets me leverage some of those more unique tools and move forward changing how we do things in the upstream to make use of the software as much as we can,” she explains, siting a recent case in which a line of color products that went to market very quickly received mixed reviews.
Following the product launch, she explains, the team began monitoring the public domain space for feedback, including reading product reviews from customers on avon.com. Though the line’s average ranking was 3.5 out of 5 stars, when the team dug into the data, they realized that this average was the result of a marked polarization: Some customers loved the product, while others had very strong reactions against it.
“It was essentially bi-modal,” explains Diane Dabkowski, Manager of Avon’s R&D Commercialization Technical Services Organization, who worked with Knapp to evaluate the issue. “It was only through sentiment analysis in JMP Pro that we quickly could see from the reds and the greens that there was something going on.” Perhaps the cosmetic’s shade was different than what the buyer expected. Perhaps the shade changed over the course of wear. The results of this analysis eventually led the team to rework the formulations of a number of shades identified as problematic and redesigned the product’s packaging to more accurately represent the line’s color palette.
Improvement opportunities like these fall squarely within Dabkowski’s purview as manager of regulatory compliance and quality assurance across Avon’s global markets. No matter how well-received a product may be, there are always ways to leverage the R&D group’s deep product and process knowledge. Optimization, however, requires a strictly data-driven perspective. “We spend so much time with these products that they become our babies. It’s so important that we have an objective lens and JMP helps us with that,” Dabkowski says.
“We may think that consumers will react one way, but using JMP and its text analytics [tools] allows us to take away our lens of bias and look strictly at the themes that are emerging from the consumer directly… and make sure that we’re not influencing the way that this data comes back to us,” Knapp explains.
New data strategies and JMP, validated by time savings and improvement
The product and process innovations achieved across Avon’s R&D pipeline since the rollout of text analytics have in many ways validated the company’s push to promote statistical literacy and data-driven decision culture. Though scientists across the R&D pipeline from blue sky to development to inline commercial product are now empowered to make consumer research a part of their standard process, the organization was not always this way. Dabkowski explains that as a Minitab user early in her career at Avon, she began assessing other tools when the limitations of the existing toolkit became clear. Others soon followed suit.
“As I was introducing JMP to the different work teams within R&D, they participated in the workshops and said, ‘Wow. JMP can just blow away Minitab.’ So they ended their Minitab licenses and migrated over to JMP,” she recalls. Now, she says, “We benefit from JMP’s applications in so many ways – it’s not just the text analysis. We use it for clinical applications, package testing, design of experiments.”
JMP users across the organization hold an annual workshop and consult with JMP engineers and educators at times when additional support is needed. But, Dabkowski adds, as the organization becomes leaner and more nimble with the increasing democratization of analytics through tools like JMP, in-house statistical support is no longer needed: “JMP has been a wonderful enabler to understand statistics.”
Gloria Portocarrero Huang, Senior Research Scientist at Avon, was working in the physical science department when she was first introduced to JMP as a tool for digging into the biology of skincare products. “Before I had JMP, there was definitely a lot of work that had to be done by hand. But once I got JMP, I really leapt at the chance to use the tools that could give us the agility to digest our data,” she says, noting that JMP in many cases confirmed some of the thoughts the team had had early on, but had lacked evidence to take action.
“JMP really helped with not only removing some of [the noise] and saving time but organizing it into a way that we can actually use that information,” Huang adds. Visual, scientific approaches like text analytics not only advance the science behind Avon’s product, she says, they also help upstream technical experts discuss consumer appetites with marketing partners and upper-level management, defining how upstream products fit into Avon's overall calendar and platform strategies. Intuitive analytics and visualization in JMP level the conversation so that everyone understands the data – and that is the very definition of a knowledgeable organization.
