H&M Bets Big on Machine-Learning to Survive
In an effort to come back from its multi-year slump, H&M is turning to machine-learning. Is machine-learning the answer to H&M’s problems and is it too late?
Fast-Fashion Industry Dynamics
The rise of fast-fashion brands such as Zara, H&M, Top Shop, and Forever 21 has contributed to the decline of the traditional bi-annual fashion seasons and the emergence of near weekly “micro-seasons”. The success of the fast-fashion business model hinges on anticipating micro fashion trends and bringing them to market quickly and at low cost. Since fast-fashion retailers are focused on predicting rather than creating fashion trends, it is critical that their predictions are correct; otherwise, they risk getting stuck with inventory that they can’t move once the next “micro-trend” begins. As a result, fast-fashion retailers are turning to machine-learning to help detect trends and avoid an unpopular and costly product cycles.
H&M Stock Hits 10-year Low as it Struggles to Keep Up with Competitors
H&M has struggled to keep up with other fast-fashion retailers in predicting retail trends and localizing their merchandise to appeal to consumer tastes. In September of this year, H&M’s stock price hit a more than 10-year low (see Chart 1) after reporting that pre-tax profits shrank nearly 20% from the previous year.
H&M’s declining performance can be attributed to two key factors. First, H&M has consistently failed to predict and respond to fashion trends ahead of competitors. In March 2017, Goldman Sachs reported that H&M’s supply chain lead times are double those of Zara. As a result, H&M’s inability to execute quickly has left the company with nearly $4B of unsold inventory. Second, H&M failed to understand consumer preferences in key markets. According to Forbes, “you could walk into any H&M store whether it was located in Sweden, the United Kingdom or the United States and it would carry very similar merchandise”.
H&M Looks to Machine-Learning for Turnaround Efforts
In an effort to improve performance, H&M is turning to machine-learning. The Wall Street Journal reports that H&M plans to analyze store receipts, returns, and loyalty card data to better align supply and demand and reduce reliance on markdowns. H&M piloted this approach in their Östermalm, Stockholm store. The store had previously been stocked with basics for men, women, and children—but after using machine-learning to analyze purchase history, they learned that most of the store’s customers were women. As a result, the store was able to reduce the number of items it stocked by 40%, adding more fashion-forward items for women and completely removing its menswear line.
Emboldened by the early success of the Stockholm pilot, H&M is now investing heavily in machine-learning to inform assortment and demand planning. Rather than relying on merchants to predict trends, H&M has built a team of 200 data scientists, analysts, and engineers to analyze data ranging from external blog posts to internal purchasing data. In addition to using machine-learning algorithms to build better assortments, H&M is investing in automated warehouses, with the ultimate goal of achieving next-day delivery for 90% of the European market. Long term, H&M is hoping to implement RFID technology in its stores to further improve efficiencies in its supply chain. The RFID technology would allow customers to scan labels and receive personalized recommendations based on their purchase history or interests.
Thinking Beyond Machine-Learning
H&M is making a big bet on machine-learning to turn the company around from a failing chain retailer to a digitally integrated brand. Unfortunately, this effort may be several years too late. The positive results from the Stockholm store pilot are encouraging, but given H&M’s massive 4,288 store portfolio, I recommend further validating its investment by piloting the technology in a critical mass of stores that is indicative of H&M’s global store portfolio prior to rolling this initiative out to all stores. Also, rather than solely focusing on the rapid implementation of technology, I recommend that H&M invest in radically re-building its culture and bringing in fresh talent that aligns with its new company vision. In an interview with Women’s Wear Daily, H&M’s CEO, Karl-Johan Persson, rejected the need to change company culture explaining that “the recent reasons why we did some mistakes connected to the H&M brand and physical stores is because we haven’t been customer focused enough, we haven’t lived [our] values well enough, so it’s more revisiting that.” I think that customer-focus is exactly the cultural mindset that H&M lacks. Unfortunately for H&M, by the time its senior management team realizes this, no pivot will be able to turn the company around. If machine-learning is in fact the answer to H&M’s problems, given its 3-year slump and all-time low stock price, does the company have the luxury of time to see through the benefits that the technology can offer?
Word Count: 791
 “The Future of Fashion: From Design to Merchandising, How Tech Is Reshaping the Industry.” CB Insights Research, 28 Feb. 2018, www.cbinsights.com/research/fashion-tech-future-trends/.
 Marr, Bernard. “How Fashion Retailer H&M Is Betting on Artificial Intelligence and Big Data to Regain Profitability.” Forbes Magazine, 10 Aug. 2018, https://bit.ly/2QAczkQ.
 “H&M’s Q3 Pretax Profit Falls More than Expected.” Thomson Reuters, 27 Sept. 2018, reut.rs/2B412oo.
 Ringstrom, Anna. “H&M Invests in Supply Chain as Fashion Rivalry Intensifies.” Thomson Reuters, 30 Mar. 2017, https://in.reuters.com/article/h-m-results-idINKBN1711G5.
 Chaudhuri, Saabira. “H&M Pivots to Big Data to Spot Next Fast-Fashion Trends.” Wall Street Journal, 07 May 2018, https://on.wsj.com/2rr9qs2.
 Marr, Bernard. “How Fashion Retailer H&M Is Betting on Artificial Intelligence and Big Data to Regain Profitability.”
 Chaudhuri, Saabira. “H&M Pivots to Big Data to Spot Next Fast-Fashion Trends.”
 Marr, Bernard. “How Fashion Retailer H&M Is Betting on Artificial Intelligence and Big Data to Regain Profitability.”
 Diderich, Joelle. “Karl-Johan Persson on Strategy and Culture.” Women’s Wear Daily. 15 February 2018. https://bit.ly/2RMYMr2
Student comments on H&M Bets Big on Machine-Learning to Survive
I question your concern about whether it’s “too late” for H&M, given I’d argue it’s a brand people buy for a specific utility – aesthetic at low price – and not for the H&M brand itself. Consequently, if investing in ML can enable the company to bring more relevant apparel options to the consumer still at low cost, then as long as they can get that product in front of the consumer’s eyes (whether via retail or digital), they seem well-positioned to grow again.
I wonder if they could be using machine learning to actually predict shopping trends more efficiently. As we discovered in our marketing case on GAP, a lot can be gleaned from google search metrics and social media, specifically on the items and style’s promoted via those channels. I think there’s also something to be said about their lack-luster eCommerce site coupled with the lower quality of their clothing. As a consumer, I’m less likely to purchase low quality clothes online than I am to purchase something that’s durable. The way they optimize their store is important, however and they should and possible still can take additional steps to address those issues.
This was an awesome read; thank you for sharing. Your piece touches on a critical topic that we haven’t really discussed at length in class, but I think is actually more reflective of the economy at large. In a utility driven-fields, do legacy companies have the runway to turnaround their fundamental approach to product ideation and development before consumers deem them obsolete? And, how bone deep do corporate changes need to be for a company to thrive in a digital world?
As you rightly point out that turnaround process starts not with introducing new processes or staff, but rather with an extensive paradigm shift on how ideas are born and vetted. I don’t agree with the CEOs assertion that its simply a matter of getting back to Company values. For me the situation requires a level of rigorous analytical thinking that may not be M.O. of the organization today.
Moreover, I disagree with ksimmons that there isn’t a time element here. It is really hard to come back from being seen by consumers as obsolete/irrelevant. Particularly when you consider that H&Ms former position has not created a vacuum but rather has led to the rise of challenger players that consumers actually like.
Tina – I really enjoyed reading this! I’m personally fascinated by the current revolution of brick-and-mortar retail and whether or not they can transition to survive in the long run. I agree with ksimmons that if machine learning can help them better serve their customer, that this should be something that the company focuses on. While culture is critical, I also think it is critical for a retailer to entirely focusing on best serving their customer. Separately – pretty interested in the RFID scanning option. In a world where omnichannel is so important, I wonder how to integrate the online and in-store experience for a customer and this seems like an interesting approach.
Thanks for the great read. My first thought is on the kind of data that H&M is already capturing, especially given the heavy reliance that they will have when moving into machine learning. For a fashion company to use ML to predict and respond to trends quickly, I would imagine H&M could potentially tap into open source data instead of proprietary data, assuming they are going for macrotrends in the industry. In addition, while you mentioned that the culture needs to be addressed with fresh talent, I wonder if you are referring to a shift in talent from design skillset to more data-science skill-set, as I could imagine that the recommendation to move towards ML might render many roles obsolete to the company. This would be a major shift and one that definitely needs to be carefully managed.
Tina — thank you for sharing this! A really interesting read. I’m curious to learn more about how H&M has used machine learning today in their stores. You mention that they used some machine learning techniques to understand that most of their customers in their Östermalm store were women; this, however, seems like an insight that could have been gleaned from a simple visit to the store.
I also wonder how they might be able to use machine learning in other parts of the their retail struggles. Could they potentially use AI to optimize their store footprint or help predict the customer trends you mention they have struggled to keep up with?
Looking forward to seeing what the future holds for them!
It’s smart for H&M to leverage machine learning here and it’s actually not too far off from their current strategy. H&M’s success is based on emulating high fashion designs and producing them at a lower cost given that they pay much less for designers and marketing. Applying machine learning to receipts, returns, and loyalty card activity is a great way to understand the popularity of specific designs, colors, and sizes, which could inform future ordering. It could also give insight into specific products that are often defective or uncomfortable (e.g., specific SKUs with high returns). I’m curious to see how this pans out because obviously many firms within the fashion/clothing industry could leverage this technology, including larger retailers such as Target and Walmart.
Interesting read! My general sense from reading this article is that H&M is leveraging ML to drive operational efficiencies. While operational improvements can positively impact bottom-line, they will only bring “incremental” changes to the company. In my opinion, what H&M needs at this point is a “radical” change, similar to the case where IDEO helped the Latin American theater-company reshape its customer experience. So, in that way, I agree with your point that H&M should focus on its culture, talent, brand vision etc. versus over-investing in implementation of ML and other such technologies.
Thanks for sharing this piece, Tina! I am fascinated by machine learning in the fast fashion space and am very excited to see how it continues to inform the industry’s future. While machine learning may help H&M better understand its target customer (e.g., women vs. men) I believe that their use of machine learning will not be meaningful on its own in turning around the company’s performance. As you mentioned, one of largest problems H&M currently faces is its inability to respond to trends quickly enough. In order to succeed in competing against others in the space like Zara, H&M must first improve its current manufacturing / supply chain. Only then will it be able to take the learnings from its machine learning technology to inform the products they are creating for consumers.
Tina! This is a very well written article. I found this article very insightful as I am not a fashion expert and did not know about this type of fast-copy-fashion, where companies try to pick up on “micro-trends” in the market and then respond fast enough while the trend/signal is alive. You mentioned in the article that: “H&M has built a team of 200 data scientists, analysts, and engineers to analyze data ranging from external blog posts to internal purchasing data.” I would recommend that H&M goes even further with their Machine Learning activities – beyond analyzing internal data and external blog posts, they should analyze their competition websites and try to see what comes up on their websites, how long does it stay, and what comments people have on those items. Perhaps, some companies are using advanced dynamic pricing models, so by using Machine Learning we could reverse engineer their models to identify the trends they are saying. Given H&M’s terrible situation at the moment I think they need to be very aggressive with their ML strategy and use any data available at their disposal.