“I liked the shaver so much, I bought the company”…


Nike, the world’s biggest apparel brand, swallows retail predictive analytics firm Celect.


(Credit: Getty Images)

On 6 August 2019 Nike announced the acquisition of Celect, a leading retail predictive analytics and demand sensing firm founded in 2013 and backed by $30m in VC funding. Using Celect’s technology, Nike will be able to predict which styles of footwear customers want, when they want it and where they want to buy them.


Why is this important?


Nike was just one of Celect’s many clients and within 24 hours of the announcement Celect’s website had gone dark - replaced by a link to the announcement. Nike will “immediately integrate Celect’s team into Nike’s Global Operations Team” – meaning the core engineering and data scientists anyway. The company is effectively no more.

While the price paid was not revealed, it is safe to assume that it would have been a multiple of the $30m poured into the company ($15m as recently as December 2018). This was for a company with around 60 employees and 20+ customers. 20+ (now unhappy?) customers who may have gone through years of disruptive integration but who no longer have access to Celect’s demand prediction capabilities.


Nike’s COO commented: “Our goal is to serve consumers more personally at scale…we have to anticipate demand. We don’t have six months to do it. We have 30 minutes.”


Nike’s move is significant for the comparatively nascent world of the application of machine learning in retail demand prediction. While they already had full access to Celect’s expertise as a customer, they felt compelled to both own the capability and remove Celect from their competitors and the market in general.


Nike has effectively bought an experienced predictive analytics team (at an estimated cost of circa $3m+ per employee being retained – as any value in Celect’s customer base has effectively been extinguished in the deal). Nike has identified that the application of AI driven predictive analytics (predicting demand for new products on an item by item basis) is a core capability that will be of fundamental strategic importance to the future of their business – and probably every retail brand - as the ‘supply chain’ continues to evolve into the ‘demand chain’.


So what does this mean for retail – and Product Lifecycle Management (PLM) platforms in particular?


While predictive algorithms are initially trained on historic sales data, they are most powerfully applied to new products – ie how much should I order of this proposed design at this MRP? That decision is made during the design, merchandising and buying process, a process that sits at the heart of all PLM platforms. It follows that the integration of new product demand prediction capabilities is a candidate for every PLM provider’s roadmap.


While AI powered predictive analytics is a powerful but currently complementary capability to any buying team (PLM enabled or not) the integration of real time consumer insight in PLM platforms could provide the catalyst to trigger the integration of predictive analytics in PLM.


The accuracy of demand prediction algorithms is inevitably throttled by the quality of the data that powers them. Currently this data is almost exclusively gleaned from item attributes and historic and real time sales data. Real time voice of the customer (VoC) input provides a wealth of normalized, weighted and benchmarked metrics for any new proposed design and represents a uniquely powerful predictive dataset. If sales data is the gasoline of predictive analytics, VoC is nitro.


As soon as PLM platforms are generating new item designs as well as VoC data it is a small and comparatively simple step to integrate historic and real time sales data feeds into PLM, at this point all the data ingredients are present for machine learning to get to work, predicting demand for any new proposed design at any given MRP – from a sketch, 3D CAD or sample image. At this point we are likely to see a steep jump in business performance for PLM customers evidenced by more appealing ranges, more accurate buying and smarter allocation. As a result, PLM platforms will evolve from being simply data infrastructure frameworks to true business intelligence platforms and the lines between ERP and PLM will become increasingly blurred.


Nike ‘just did it’ by buying Celect, and with the relentless pressures on fashion retail demanding increased speed, efficiency, sustainability and the smart management and application of data, others will not be far behind.



Copyright SoundOut 2019