SKU Innovation

An innovation product, or New Product Development (NPD), can range from a new flavour or pack-size of an existing product, through to a completely new type of product for the organisation or category. There is no magic formula as to what makes a successful product, and many innovations will fail. But innovating is incredibly important for organisations, particularly in the Fast Moving Consumer Goods space, to retain and grow category share.

The decision of whether to launch a new innovation product will consider a range of factors from predicted demand, based on market research or benchmarking existing products, to expected cannibalisation of existing products. So how can we use data to identify the best innovations, and how can we maximise the benefits and minimise the risks when these products are launched?

In our experience, predictions of demand for innovation products are not strongly rooted in data. The better examples identify a similar existing product and benchmark demand expectations to that, while on the poorer end others make sweeping assumptions on best and worst case market share and make business cases and launch plans on a rough mid-point. Even in the case on benchmarking, there is often more that can be done to further refine this estimate.

Firstly, you can apply machine learning to identify which factors drive demand, from pack size/format to brand to price point. With these insights, your machine learning model can give a truly data-driven demand prediction, and also identify the 10 most similar products on a demand basis which can be used to validate the prediction. This is best done using the average store demand for each product to avoid bias from different products being sold through different numbers of stores.

Secondly, the results from consumer testing can be added to consider the attractiveness of the new product to the other recent innovations. By analysing the historical correlation between different consumer testing results and consumer uptake, you can further refine the prediction of expected demand.

When an organisation is entering a new category with a product innovation, it can be challenging to find appropriate benchmarks in the existing product portfolio. However, there is a wealth of information which can be mined on competitor’s products using Nielsen or IRI ePOS data. Even when launching a product innovation in an existing category, these competitor benchmarks can help to better estimate expected demand. Adding competitor’s products gives the machine learning model more data points to learn what is driving demand, and more options to find the most similar products to give a tighter confidence in the demand prediction. 

To leverage these demand insights, the store-level demand prediction needs to be scaled up based on the number of customer stores the product innovation will be sold in. Promotional activity around launch and cannibalisation of other SKUs also need to be accounted for – more details on these are in our Trade Promotions article. With this holistic demand forecast, the next decision is how much of the innovation product should we order/produce? Too little and you will stock-out, too much and you will have excess, unsold inventory. So where is the sweet-spot? 

One option would be to find the order quantity which maximises the margin for error in either direction, exactly between stocking-out and excess inventory. However, to go one better, you can apply the opportunity costs of stocking-out vs excess inventory to find the order/production quantity which minimises your total financial risk. This optimisation is described with the following chart.

Earlier we discussed how machine learning can be used to get a more robust and data-driven prediction of the demand a new product is likely to achieve. There are numerous advantages to this prediction being much more data-driven, particularly earlier in the innovation cycle. For example, an initial demand prediction can be made before any investment in R&D or consumer research, using the planned product characteristics. This demand prediction can then be refined at each decision gate, based upon consumer testing results and commercial decisions around pricing. This brings more rigor to prioritising and progressing innovation opportunities.

Further, the same machine learning techniques can be used to optimise elements of the product. For example, we have created a solution with a top 20 CPG firm which identifies potential changes to new products to increase their predicted and/or profitability. These changes range from using different product sizes/formats to branding products as healthier choices. The same solution supports decisions around pricing by predicting the demand for different price points, converting that into volume, revenue and margin KPIs and identifying the price point that will maximise the bottom line.

A second use case for the same CPG helped the innovation team identify promising opportunities for new products at the start of the process. Instead of predicting the demand for a specific set of product characteristics, we applied the machine learning model to screen millions of combinations of product type, flavour, format, brand etc to shortlist potential new products. A key decision here was to include competitor products which enabled the tool to identify products that were significantly outperforming the category. The tool identified a range of product renovations and new potential products with predicted demand and financial KPIs 30-50% above the product innovations currently in the pipeline. 

By de-risking the launch of new innovation products and helping to prioritise innovation opportunities, organisations can become much more efficient and effective at bringing new SKUs to market. Along with an optimised delisting process to efficiently remove underperforming SKUs, companies can create a much more productive portfolio with higher category share and margin. 

Purpose AI is an AI and Analytics Consultancy delivering data-driven solutions for complex Supply Chain and Commercial decision making. We work across the Consumer Products and Retail Industries and have worked with many Top 20 CPGs and Retailers in Europe and North America. Our USP is that we have developed industrialised solutions to common industry problems, reducing the time and costs to deliver value. But unlike SaaS providers, there is no licensing costs, we customise our solutions to your requirements and embed a strong business process to ensure they deliver value.