TFR: What is GFAIVE?
Maria Makarova: At its core, GFAIVE is a company focused on simplifying the life of fashion brands and retailers. We built an AI SaaS Platform that combines elements of assortment planning, demand forecasting, and new model sales performance predictions. With the help of our solution, companies can grow their revenue steadily, reduce costs, and minimize waste associated with inventory distortion.
TFR: What is the entrepreneur story that inspired the launch of the company?
MM: GFAIVE is a family-owned business. It was founded by me and my father, Andrei Makarov. Initially, we got the idea of starting this company after speaking to multiple retailers and discovering that regardless of size, companies typically rely on Excel spreadsheets for demand forecasting and collection planning.
This came as quite a shock for us because, with the emergence of new technologies, we felt it was evident that there are better approaches our there than simple Excel calculations. After all, human brains can’t process enormous amounts of data and notice minor year-to-year changes. A machine, on the other hand, can.
So, we assembled a team that combined data science experts and fashion industry professionals to come up with a solution that the retail sector could truly benefit from. After our hypothesis was confirmed by experts during various accelerator programs, we knew we had to stick with it.
TFR: What are the main challenges in apparel in relation to assortment and inventory management?
MM: The biggest challenge fashion retailers face is predicting consumer demand in advance and prior to production start, up to 12 months ahead.
The truth is existing solutions provide numbers solely based on the analytics of historical sales. Only a few provide insights about future trends and moreover, none of them fuse with external trends relevant for a specific retailer. Thus, retailers are left with doing the guesswork and predicting these numbers themselves.
TFR: What changed in the retail industry compared to 5-10 years ago and in relation to data and tech?
MM: Well, first of all, there are many more technologies available now than 5-10 years ago. Yet, retailers are only starting to use them. The pandemic has also had an enormous effect on the industry and assortment planning in particular. Before it, everything was about growing the number of items, increasing sales, producing more and more. However, the results weren’t that great because it was an endless cycle.
Now, it’s kind of like COVID came in and made a necessary adjustment. Sales in some areas dropped significantly and assortment got the chance to become much more balanced. Thus, leading to more sustainable profitability.
For instance, we are already seeing that there are less discounts and sales. There’s been a forced optimization. Now we don’t blindly produce. Instead, brands want to follow a more strategic and thoughtful approach by optimizing their resources. In the long run, it’s going to result in a better market, even though there might be some growing pains right now.
TFR: A recurring challenge I faced when working for different apparel brands is how chaotic is their data structure, architecture and mapping. Data is shared and gathered in different systems and manually managed. Different roles (designers, buyers, planners, allocators…), using different data and processes. Reaching an integrated planning capability is more complex than expected. Then, a lack of historical data or product description could diminish the effectiveness of predictive analytics.
How do you deal with this lack of consistency or maturity in regard data management? (e.g. product attributes, customer segmentation, store clustering, etc)
MM: Since our team is well versed in the assortment structure and has in-depth knowledge of fashion retail, we can identify the hierarchy formats that are needed for different clients. Hence, the data that goes into the hierarchy will always be unique to each retail brand that comes to us.
So, an assortment will always have a category, gender, or upper and lower label categories. Different brands and retailers value specific attributes which only they might work with. For instance, some might work with “trendiness”, others might value a certain material and prioritize it. So, in this regard, we value the uniqueness of the client, and we offer statistics and algorithms based on these characteristics too.
Generally, we need a minimal number of fields and then it’s all about customization. With our platform, up to 15 attributes can be added. An interesting thing is that we specify attributes from photo catalogues and have computer vision and neural networks process them as well.
TFR: Brands are realizing that assortment is as important as store location. Designers and planners should develop specific assortments at customer/location level. Some countries might have preferences for a specific color (e.g. Black), product line (e.g. Casual) or even categories (e.g. Outwear). But not only countries, also cities or neighborhoods could show specific consumer behavior patterns.
How do your assortment and demand forecasting solutions help brands improve business efficiency?
MM: Efficiency is improved thanks to the higher accuracy of forecasts that our solutions provide. By taking into consideration Instagram accounts that are relevant to a specific retailer and monitoring Google Trends changes, we’re able to identify products that will perform best. Typically, with up to 98% accuracy.
Since even 1% increase in forecast accuracy can boost revenue by 3%, our clients tend to be very happy with our platform.
TFR: Many companies, from large to startups, are offering analytics solutions powered by machine learning. What are the benefits brand should expect from such solutions? What are the most important variables when choosing a tech partner such as yours? (e.g. data-scientist team, better algorithm, faster results, etc)?
MM: The best part of machine learning is that its algorithms can go through enormous amounts of data that no human brain can process. Plus, they can find patterns that we simply wouldn’t be able to do ourselves or it would take too much time. Hence, speed and accuracy are the main benefits of ML-powered solutions.
However, when choosing a tech partner, it’s important to look at more than just their claim to be machine learning driven. Do they have expertise in your industry? Are they acknowledged by well-known organizations? Do they actively participate in industry events? Does their team have actual data scientists? All these questions should be asked when you’re looking for a tech partner.
TFR: Could you briefly explain a couple of business cases where GFAIVE participated?
MM: We had a couple of interesting business cases already.
First, we worked with a national kids wear retailer who relied heavily on legacy merchandise planning and once the pandemic hit, was stuck with useless past years benchmarks. We worked closely together and implemented our platform with a simulator feature so that the client could run multiple scenarios and evaluate possible outcomes. In the end, the client was able to gain insights on how to structure future collections and could optimize its assortment by up to 20%.
Another client was a footwear and accessories retailer with more than 900 stores in over 300 cities across Russia, Belarus, Kazakhstan, and Ukraine. The team was looking take advantage of data and transform its pre-production planning process. So, our team enhanced their tech stack with our demand forecasting tool. Currently, the company expects a reduction in stock levels of 30%.
TFR: Demand for tech roles was high but increased even more with Covid-19. How do you manage to find these skilled people? What is their motivation?
MM: First, we try to find talent across multiple regions. Currently, we have a highly international team with members from Russia, Armenia, Kazakhstan, and many other European countries.
Secondly, we value flexibility and try to create a trusting work environment. We place an emphasis on performance as opposed to the number of hours spent in the office.
Lastly, we’ve transitioned to a completely hybrid work scheme. Thus, we meet regularly with team members face-to-face, but don’t go into the office every day. We listened to what our employees wanted when restrictions began to lift and decided to stick with a hybrid work arrangement. As a result, everyone was happy, and productivity didn’t drop at all.
So, I think these are the main factors that help us attract skilled talent within the IT sector.