FEATURE11 July 2019
Distilling data
x Sponsored content on Research Live and in Impact magazine is editorially independent.
Find out more about advertising and sponsorship.
FEATURE11 July 2019
x Sponsored content on Research Live and in Impact magazine is editorially independent.
Find out more about advertising and sponsorship.
For a new whisky brand, launching in Australia required more depth than qualitative research could offer, so it turned to data-driven modelling to help it estimate how well the tipple would sell. By Katie McQuater
Scotch is no longer consigned as an old man’s drink. In recent years, distilleries have joined gin and rum makers in launching new brands as they look to target a younger demographic of drinkers.
Pernod Ricard-owned Chivas Brothers launched its Allt-A-Bhainne single malt in 2018. The product uses semiotic codes that aren’t typical of the single malt category, which historically trades on tradition and heritage – in contrast, the Allt-A-Bhainne distillery opened in the 1970s.
As such, Chivas Brothers wanted robust data to help it build confidence internally and secure marketing budget around the launch of the new whisky in the Australian market.
Rather than commissioning qualitative research, the company worked with insights agency Skim to produce a volumetric prediction model aimed at forecasting the brand’s potential success and highlighting marketing priorities.
“The client was looking to test the concept, profile the potential buyers for marketing purposes, identify key occasions that consumers would associate to this new single malt, and get a robust estimation of the potential volume of sales in the first three years from launch,” says Tommaso Gennari, senior analytics consultant at Skim Group.
Skim conducted an online quantitative survey of 1,200 whisky buyers in Australia to test concepts, profile potential buyers and estimate consumer selections through choice-based conjoint.
Pernod Ricard Australia also gave the company three years’ worth of sales data to use for the expected competitive set.
The team could draw on both survey data and survey-based choice modelling, and the sales data, to develop a volumetric estimate of seasonality and the impact of distribution on whisky sales.
The conjoint was designed to measure the impact of awareness of the new single malt concept on consumers’ choices.
The best price at which to set the whisky was also investigated as the conjoint simulator – which was calibrated to market share – included the possibility of switching between two prices and gave the base share of preference of the new whisky that was then used in the volumetric predictive model.
The researchers could predict how many cases would be sold each week for the three years after the launch by combining the share of preference from the conjoint simulator with the seasonality and distribution effect from the sales-data analysis. They also combined output from the simulator with other aspects of the research to illustrate what impact different price choices, distribution and marketing activity could have on sales and positioning.
The most challenging aspect of this approach, according to Gennari, was summarising and explaining the relative role of the assumptions the researchers were making when using the predictive model.
Alongside the volumetric model, the survey research found that the Scottish name was a barrier for some consumers, while many liked the bottle’s visual appearance and the taste of the peat.
Though Skim did not give a ‘yes or no’ answer to the price question, it outlined the implications of each pricing choice in terms of volume and positioning.
“Maybe unexpectedly for us, the client chose to go to market with the higher price, despite a lower price generating more volume,” says Gennari.
This was because the company wanted to keep a “differentiation in premiumness of offerings within the portfolio,” says Gennari, but was supported by the “evidence that it could still hit its volume target at the higher price – if it was able to produce a certain level of market awareness and reach a certain distribution”.
0 Comments