Traditional brick-and-mortar (B&M) stores are under immense pressure due to the threat from Amazon and are often trying to compete by putting many products on attractive promotions. While Amazon uses sophisticated predictive analysis to personalize their marketing strategies and promotions, most brick-and-mortar stores do not have the resources to experiment and learn. In fact, 30% of promotions run by B&M stores could be toxic – making less money than the baseline situation. This happens when the margin depletion due to discounting is not adequately made up by a sales lift.
If done wisely, promotions can indeed add to the bottom line, but the problem is particularly acute in price inelastic items. The quantity or quality of promotions run for such items does not hold any value as retailers do not get much sales lift (in exchange for the discount). This can turn the entire discount process into a loss and can happen due to multiple reasons:
- For certain essentials, it is difficult to bring customers away from the competitors due to their brand loyalty
- For certain other items, a small discount does not help the cause of the customers looking for offers that represent a sizeable share of their budget.
- In some other cases, discretionary/indulgent customers might be more concerned about product attributes in these cases rather than pricing.
The solution lies in being able to use data to make decisions. Retailers need to accurately forecast the impact of proposed promotions and only run those promotions which will add to margin. They need to be able to identify items which are not going to get an adequate lift from the discounting and save the margin loss. Less is more here, in the sense that about half of the value comes from avoiding toxic promos, with the other half coming from making good promos better. However, devising and executing profitable promotions come with their own challenges:
Challenges in Forecasting promotions
When it comes to promotions, it is essentially a sales forecasting problem under conditions of the promotion. Many retailers try to use spreadsheet-based simple tools to do this forecasting. This can be a very tedious, time-consuming process and prone to inaccuracy. While most of the margin impact happens due to the sales lift, there are other effects that need to be accounted for to determine the overall impact of a promotion accurately.
Estimating Secondary effects
- Cannibalization: While the promotion might apparently boost margin for the promoted product, it might deplete sales for a similar product and worse still, that product might have a higher margin. Another way the cannibalization of the same item happens in a future period is known as pull-forward /pantry-loading effect.
- Halo: There might be loss leader items which help boost sales of other items. This might be due to the other items being complements or the promoted item drawing footfalls to the store and boosting sales across categories. It is always better for a retailer to tolerate some degree of margin loss on one small category to generate a higher profit on a larger category.
With multiple marketing channels, often attributing the ‘true incremental’ marketing spend for a promotion requires careful calculations and estimations. Retailers also incur operational costs for running promotions, which need to be considered. In addition to the challenges above, most of them struggle to estimate the impact of a promotion in a newly introduced item or in an item that is being promoted for the first time or is being promoted at a discount level not seen earlier.
Adding to the woes of B&M retailers are the multiple stakeholders involved in the promotion decision-making process. These stakeholders have different, and in some cases, opposing incentives. For example, a vendor would want to have his/her brand on promo and cannibalize the other brands, whereas a marketing head would want to improve the marketing ROI. This leads to a lot of negotiations between vendors and merchants, and between merchants and marketing. Most of such negotiations can be quickly concluded if data-based answers are readily available.
A simple yet comprehensive solution is one where retailers are provided with a robust platform to manage promotions – a software application (not spreadsheet tool) with following features:
Predictive Analytics: Highly accurate estimation of promotion impact along with secondary effects needs a wise mix of machine learning algorithms and coded business judgment rules. Ensemble forecasting performs well – items with sufficient history are predicted well by linear or Poisson regression models with necessary variable engineering and regularization; items with sparse history are often predicted well (after pooling) by ‘black-box’ neural net, XGBoost models.
To handle new and sparse history items, similarity analysis is done by using feature models and Natural Language Processing on item descriptions. The system needs to simulate different promo states (i.e. different discount percentages or marketing modes) – models are converted to APIs using cloud infra and server-less cloud functions.
An Intuitive and Comprehensive Workflow that allows vendors and merchants to submit proposed promos, simulate the promos with different price points and marketing support. An innovative tool that shows which promo will generate profit at what price points using certain marketing elements (in-store display, ad display, etc.) and allows allocation of resources and reports the impact metrics of the promos during and after execution.
In simpler times, promotions were a simple concept. Reduce price to sell more or lower your price than the competition to capture all sales in a market. In the current landscape, with aggressive competitor strategies and the availability of rich transaction and pricing data, it is imperative that today’s retailer utilize the information available to stay profitable.
Join Our Telegram Group. Be part of an engaging online community. Join Here.
Subscribe to our NewsletterGet the latest updates and relevant offers by sharing your email.
The writer is a Director of Product at Impact Analytics. He has years of experience in creating and deploying large scale products in areas such as banking, media and enterprise mobility. He specialises in managing cross-functional teams and combining the best of technology, data, and design to execute a successful product vision.