All Collections
Sniffie: Customer's Help Collections
Price optimization - AI
Tweaking preset AI settings in Sniffie's pricing workflows
Tweaking preset AI settings in Sniffie's pricing workflows

This article provides information on AI settings to fit your needs, when creating AI-based pricing workflows in Sniffie

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

Sniffie's pricing automation platform has lots of AI modules that you can use to help you make better pricing modules. Although our data team has selected the best configurations for you and those are applied as a preset, you are able to tweak the AI settings yourself if you want to. This article will help you to understand how the AI settings works.


Where AI settings are tweaked

First, you have to have at least one AI module enabled in your account before doing this. Whether this module is enabled depends on your pricing module package. You can contact our customer success team to inquire about enabling AI modules to your account.

Second, you should create an AI pricing workflow to one or more products or product groups. You can read about creating a pricing workflow in this article.

When you have selected a pricing workflow, a modal opens where you are able to rename your strategy. There is also a section "AI settings", which is closed by default. You can open this to edit AI settings for this strategy (shown below).

Settings for profit optimization

Profit optimization AI workflows have the following setting options available for you.

  1. Starting price: the price that defines the starting point for optimization. It is used to calculate the maximum and minimum boundaries where the price can be optimized for.

    The available options are:

    1. price (current price of the product)

    2. recommended retail price

    3. original price

    If you have the price monitoring module enabled, you also have the option for the starting price to be one of the market prices.

  2. Maximum percentual price increase from starting price: The maximum upper boundary where the price can end up. For example, a value of 25 means that the maximum price the optimization can end up is starting price * 1.25.

  3. Maximum percentual price decrease from start price: The minimum lower boundary where the price can end up. For example, a value of 25 means that the minimum price the optimization can end up is starting price * 0.75. Please note that in profit optimization mode, price will never be set to a level that will provide negative profits (i.e. lower than the costs of the product) even if the maximum percentual price decrease would allow this.

  4. Maximum one time price change in %: The maximum amount a price can change from the price when the AI runs. For example, when the price of a product is 100 when AI is optimizing, a value of 20 means that the price may change to a value between 80-120 (20% below and above current price, which is 100).

  5. Minimum one time price change in %: The minimum amount a price can change from the price when the AI runs. For example, if the price of a product is 100 and this value is 1%, the price needs to change at least 1% for the AI to change the price. If the models suggest a 0.10 price change, this change will not be implemented.

  6. Market behavior: Defines how much the AI prefers sales from the recent days compared to older sales data. This is affected by seasonality.

    The available options:

    1. Stagnant: all sales from any date (e.g. 10 years ago) are equally important in AI models. This means that the consumer behaviour has not changed at all.

    2. Stable: market is relatively stable and sales from two years ago still have a relatively large impact when deciding optimal price points in the market.

    3. Balanced: the most common option, when more recent sales are still preferred, but the sales data from the last year has a significant impact still in the decision making. Data from two years ago has less impact, but still is taken into account in the consideration. This option will work very nicely with seasonal products.

    4. Volatile: Consumer behavior is changing fast, and only sales data within the last 6 months is considered important in the AI decision making. Older data has very little impact on the decision making.

    5. In turmoil: Last week's sales have the most impact. Data from 3 months or older has no impact at all.

  7. Which prices to include in the optimization: determines if AI will take into account all price types (normal and campaign prices), or normal prices only. If normal prices are selected, also sales that occur with coupons are ignored.

  8. Skip extrapolated dates: since we can only accurately know the price of a product when it did not sell after an integration is live in our system, we extrapolate historical prices from receipt data.

    This means that if an integration goes live today but receipt data is available for the past 30 days, we apply that data to the integration to make a determination about price performance. Enabling Skip extrapolated dates would prevent that determination from being made. There are some scenarios where this is useful.

    For example, lets say we are looking the 30 day historical receipt data for a product called Product A. The data shows us that 30 days ago Product A was sold for a price of € 100, there have been no further sales of Product A since then, and the price has not changed at all in this time. Based on the historical receipt data we extrapolate that the price of the product has remained € 100, and at this price there has been no sales.

    This may not be accurate if the price of Product A had changed during that period, but that change is not visible in the receipt data, since no sales occurred. Similarly, if Product A had not been available for purchase during the past 29 days, but came back on sale today, the system would have no way of knowing this as this is not reflected in the receipt data.

    Extrapolated dates are considered significantly less important in the AI decision making as is, but if you want to get rid of the extrapolation effect fully, you should enable this option.

Setting options for markdown optimization

Markdown optimization AI workflows have mostly the same options as profit optimization. However, these extra setting options are also available for markdown optimization.

  1. Markdown target: defines the target of markdown process. Available options

    1. Clearance: will try to achieve the desired stock level just on the date that is defined in the strategy. Does not consider storage or discard costs an important factor in the decision making.

    2. Best possible profit: will consider storage or discard costs of the product as well as the target of the optimization. This means that it may not achieve the desired stock level if the discard / storage costs are not that significant.

  2. Return rate %: Defines the return rate for the products in percentage. This has an impact on the perceived stock of the product, so the AI will adjust prices to account for the higher amount of products that are currently in stock. For example, if the inventory of the product at the start of markdown optimization is 10 and return rate is 50%, the AI understands that it needs to sell 15 units during the optimization period.

  3. Desired inventory at the end date: desired stock level at the end of the optimization period. If you want to try to get rid of all stock, set this to 0. Note that if you have huge amounts of inventory, the AI may not be able to achieve a low stock by changing price if the sales volumes of the products are low at all prices.

    For example, if you have 300 units of available stock, give the AI 1 month to get rid of stock, and the price of the product is currently 100 and estimated daily sales volume is around 0.5 units and with a minimum allowed price of 50 would be 0.75 units, there is no way that the AI can achieve 300 units of sale, but rather will only sell 22 units within the 30 day period at the lowest possible price.

  4. No price increase allowed: If this is on, once the optimization starts, the AI will not set the price of the product to a higher level even if it means that the optimization goal is achieved faster. This means that the direction of price in this optimization is a decreasing one.

Did this answer your question?