Metrics to measure the impact of artificial intelligence-based trade optimization system.

The commercial optimization software based on artificial intelligence must have feedback on the commercial actions carried out, new sales made, levers and promotions used. All this information will be used to feed back to the predictive models, adjust the learning of the models and learn how to optimize levers and promotions to achieve a greater impact on the business.

The latter is fundamental to any commercial or marketing action software. Business IntelligenceHaving clear KPIs that measure the effect you are having on the commercial channels.

What KPIs should an artificial intelligence-based software provide?

Let's focus on which KPIs an AI-based software that encompasses all the capabilities described in this article should provide; we can make the following summary:

• Opportunitiesespecially those that are of range extension:
  • Number of range expansion opportunities identified.
  • Number of converted opportunities, which have generated sales.
  • Opportunity conversion ratio: No. opportunities / No. converted ops.
  • Net sales achieved by the opportunities in the period of validity of the opportunities. This period is the period between two opportunity calculations by SAIL based on new data (mainly sales, customers and products). The effective period is usually monthly or weekly.
  • Recurring net sales of opportunities. Previously won opportunities that continue to generate sales for the customer, having been incorporated into the assortment of that point of sale. This recurring sale is usually measured during the twelve months following the month in which the opportunity was converted.
• Impact on the channel
  • Relative impact on the channel of sales achieved from opportunities with respect to channel sales.
  • Percentages of the KPIs of the previous point with respect to customers and the channel.
• Impact on customers where they have become opportunities

• Customers who have accepted opportunities in each opportunity period and comparisons with historical data.
• Impact of materialized opportunities on customer sales, comparative net sales at points of sale before and after the opportunity, taking into account seasonality. 
• Business growth in customers where opportunities are achieved. 

The KPIs above are intended to measure not only that the system estimates good opportunities for range extension if not that generate recurrenceThe use of these products will help the customer to accept them in his "regular" order and help to build customer loyalty.

►You may also be interested in: Sales KPIs. What they are and which are the best

In order to measure development opportunities and as a whole the impact of the optimal order is to apply the most common sales KPIs: customer growth, overall net sales, sales per customer, etc. 

It should be remembered that the artificial intelligence is not something magical, it is based on mathematics and therefore it is necessary to use statistics to validate impact. It is not enough to use instinct or take a specific case to say that it does not work, everyone has experience of Amazon recommendations that make no sense at all or interactions with Siri that are totally absurd. And yet many of us continue to use Amazon and virtual assistants and are reasonably satisfied.

In our business we must make informed decisions on the basis of the data..

When implementing an artificial intelligence system, one of the best tools to check that it really works is to use the statistical technique of dividing our population (in our case, customers), into two groups: experimental and control. The predictions and actions given by the commercial optimization system will be applied to the experimental group and the control group will continue with the commercial tools already used by the company. The experimental and control groups should be equivalent in terms of customer characteristics, sales, products and orders.

The KPIs of the experimental and control groups will be measured, and if the artificial intelligence system is well implemented and the groups are equivalent, the benefits of the new system will be evidenced by an improvement in the KPIs.

This experimental group vs. control group tool is best used in digital channels, as it allows you to quickly create different groups with different customers and verify that by measuring the KPIs in more or less relevant periods of time (weeks), the results of the experimental groups are always better than the control group.