Predictive Analytics and Big Data
The retailer of the future is a data driven and fact-based decision maker that will harness vast amounts of new data in order to better align products with market demand, improve consumer engagement, develop better customer relationships and make better business decisions.
Most retailers have reasonable information reporting in the forms of real-time dashboards and historical reports. However, this information leaves retailers stuck in the past. The future of retail clearly looks forward, and the two tools that most empower retailers to proactively create their futures include predictive analytics and big data.
Predictive Analytics
There are many consumer interactions that if properly depicted with predictive models empower retailers to improve performance. When working with new retail clients I typically recommend beginning with predictive analytics to forecast offer responses, sales conversions and up-sell lift. Here's an example of how one offer-response model may work.
By tracking consumers' offer-response behaviors to varying types of offers across various channels, the CRM system learns what types of offers consumers respond to and then categories consumers into offer-based segmentation – such as:
- Offer-Induced—these are the persuadable consumers and the strategy here is to deliver highly relevant and personalized offers (generally for higher margin goods) in order to increase customer share and incremental sales.
- Offer-Unnecessary—these consumers demonstrate repeat purchase patterns with or without offers so the retail strategy here is to avoid making offers for related products which consumers would purchase anyway and instead only send offers for product categories from which they have not purchased.
- Offer-Denied—these buyers only buy based on explicit needs, so retailers should avoid sending them offers as this results in selling products at reduced margin that would have otherwise been sold at full price.
- Offer-Adverse— these buyers don't like offers, and may respond negatively to being targeted. I often call these buyers the sleeping dogs. You should just leave them alone.
This buyer behavior recognition is one part of a predictive analytics model which increases offer conversions and margins by targeting receptive consumers and not offering discounts where they are not necessary. Adding additional consumer intelligence delivers even more proactive strategy and forecasting accuracy. For example, including RFM (Recency Frequency Monetary) analysis strengthens the model and gives retailers even more predictability.
CRM systems can automatically tabulate consumer purchases from the POS into RFM categories. This is often done as part of a loyalty application but a loyalty system is not required. The RFM analysis displays consumer purchase patterns that when combined with additional data such as customer segment, purchase history or information such as the prior Offer-Response behaviors further shows how retailers can identify the best promotional opportunities. Below is a simple RFM table which illustrates how to align campaigns based on consumer intelligence.
Recency |
Frequency |
Monetary |
Action |
<90 days |
1-2 |
High |
Nurture campaign promotions of higher margin goods. |
91-180 days |
3-12 |
Moderate |
Increase promotions of complimentary or bundled goods. |
>180 days |
13+ |
Low |
Life cycle nurture campaign promotions, including close out and higher discount goods. |
The above actions are overtly simple. Each combination of Recency, Frequency and Monetary values will adjust campaign designs by promotion frequency, items and triggering events.
Expanding the table with additional consumer intelligence shows how that additional information improves the campaign strategy.
Recency |
Frequency |
Monetary |
Offer Response |
Action |
<90 days |
1-2 |
High |
Offer Denied |
The RFM analysis alone would have included this consumer in an automated campaign. However, also recognizing the consumers Offer-Response behavior changes that to exclude the consumer from the campaign. |
91-180 days |
3-12 |
Moderate |
Offer Induced |
In this table example, this consumer offers the most predictable upsell and margin lift. This consumer should be placed in a nurture campaign of higher margin up-sell, cross-sell and complimentary products. |
>180 days |
13+ |
Low |
Offer Unnecessary |
This consumer should be placed in a campaign that only offers products from categories which the consumer has not previously purchased. |
Increasing the consumer intelligence with additional consumer behaviors – such as purchase history by product category, customer satisfaction scores (using NPS or CSAT), the loyalty reward rate, the loyalty break rate or Customer Lifetime Value (CLV) to name only a few — will continue to improve customer segmentation, behavioral forecasting and campaign performance.
Retailers have long struggled with the Right Product / Right Price / Right Channel / Right Time objective. IMHO, with the volume of products, fluid market conditions and fleeting consumer behaviors, attempting to achieve this all important objective in real-time without automated technology is an uphill slog that will never realize the timeliness, accuracy and results of predictive models.
Next - Retail Big Data >>
          
|