Using AI in pricing FAQ

This article answers the main questions related to insights module

Niko Naakka avatar
Written by Niko Naakka
Updated over a week ago

Our AI optimization algorithms can help you gain tremendous amounts of profit when configured properly. In most cases the default configurations are already quite optimal. However, depending on your own use case you may want to sometimes adjust or change the configurations. This FAQ is here to help you with that.

Table of contents:

When should you use product groups vs. when to optimize individual products?

Optimize as product groups if any of the following are true:

  • You want your products to have identical prices across all optimization rounds (e.g. sizes of clothing, colors of clothing or other similar products)

  • You want your products to change price in harmony, even if they are differently priced (retaining the difference in prices across the items)

  • You want to optimize non-performers with very low sales volumes, that get too much negative feedback on 0-sales days otherwise (grouping them alleviates this issue, and the cannibalization is automatically minimized)

  • Please note that his can be done also on strategy level

Optimize products individually in these cases:

  • You are ok with products having different price points from one another

  • You are using stock optimizer and want to get rid of stock as quickly as possible

  • Your products have a high amount of regular sales (>0.5 units per day)

Should you take into account sales with all prices or non-discounted sales only?

This is a question we get a lot, as our AI is able to optimize different price types separately. Generally it is an okay choice to optimize across all price types when your marketing activities run on a recurring schedule, and less than 40% of your sales occur as discounted sales. With the rest occurring as normally priced.

The AI automatically accounts for seasonal fluctuation, to mitigate seasonal effects on pricing. Therefore, extreme marketing activity during special periods like Black Friday are accounted for by default. The AI understands such fluctuations are about the period of the year itself, and not only price.

If your products generally only sell when discounted, then there may not be enough sales of normal-priced products for the AI to get anywhere. In these cases you should assess all price types individually.

Note: remember that marketing activities can be added as a cofactor in the optimization.

How do I know my optimization is performing as expected?

Look into the strategic pricing gateway, as well as your volume vs. price, volume vs. revenue and volume vs. profit graphs.

Our AI will optimize profit by default, unless you have given it permission to sacrifice a certain percentage of profit. What this means is that the AI will always move toward the peak of the volume vs. profit graph, and the strategic pricing gateway. Depending on your cost structure, optimum revenue and optimum profit can land in different places.

If the peak of the volume vs. profit graph is very steep, there is a clear winner in the price points and once the optimum is found, the price won't fluctuate as much. If the peak of volume vs. profit is broad, a larger amount of price points are optimal and prices may change more frequently between them.

How much history do I need for accurate future forecasting?

The more we can get, the better. There is no limit really. Especially with seasonality we want as much history as you are able to provide for us.

If you have less than 1 year worth of historical data, the model for future forecasting is incomplete. If this is the case it will be indicated in the forecasting modules.

Can I add something as a cofactor in AI optimization?

Yes. There is no limit on the amount of cofactors you can add. Cofactor data types can be either numeric, Boolean or discrete. Cofactors are then taken into account in all evaluations, and you can also simulate on them. If you wish to add more cofactors into your AI optimization please contact our team.

Did this answer your question?