There are many transportation terminals in the United States that are connected to refineries through oil pipelines, railways or water terminals, and there are huge oil storage tank areas. The oil at the gas station is usually pulled by trucks from these terminals. Every morning, there are many tankers coming and going. It is these oil trucks that are filled with oil and then imported into the underground storage tanks of the gas stations and sold to consumers. Fuel prices at gas stations in the United States are adjusted almost daily. For decades, fuel pricing has been the art of the door, relying mainly on human computing and intuition. With the rise of artificial intelligence and machine learning, their enormous potential will make fuel pricing more accurate and efficient. Currently, the testing phase is testing this approach using a new pricing tool. PriceAdvantage is a division of Skyline Products, a smart signage solution that was created and manufactured in the transportation and petroleum industry for more than 40 years. Its previous fuel pricing software used linear algorithms to predict price changes. This will affect the retailer's sales and profits. But unfortunately, this method is not perfect. Brendan Doner, Scientific Data Specialist at PriceAdvantage, said that if the price is reduced by two cents, the predictions obtained by this method will have about 30% errors, and the results are not very accurate. This fuel pricing software uses a top-down modeling approach. This method was based on economics, and both Doner and his colleagues thought it should work (but it was not). Based on these assumptions, a model is created, and the price given in the model is the approximate fuel price set. He cited an example. For example, in this model, retailers increased their gasoline prices to achieve profitability, making them 10 cents higher than previous pricing and 6 cents higher than the market price. Customers will say "This is crazy. Maybe I will get more profit today, but after three weeks from now, if everyone realizes that our pricing is high, we will lose all our customers." On the Other hand, the old model would suggest lowering fuel prices to increase sales, regardless of the dynamics of other locations. This model is based on the knowledge of economics textbooks - if you lower the fuel price, you will sell more fuel today, and Doner thinks this is old-fashioned. Another problem with linear computing is that competitors' responses to price changes cannot be taken into account. Experienced retailers get price data feedback through frequent on-site surveys or regular Oil Price Information Services (OPIS) to quickly understand price changes from competitors. As a result, they can react faster and prevent competitors from achieving the desired sales or profit growth. Doner said that the old model draws on the experience of real-life pricing and does not do a good job of adjusting pricing. To abandon the idea that "we think it should work", it is more like a bottom-up approach to see how it works in reality. What PriceAdvantage is pursuing is not creating an algorithm that allows retailers to blindly price or simply give a suggested fuel price. Instead, I hope that this economic model can provide data for fuel price analysts to help them better understand the ins and outs of the pricing. And artificial intelligence technology has become the core of PriceAdvantage's "new" fuel pricing software. Doner said that most people think that artificial intelligence is supported by neural networks. These huge models are like black boxes. It is not clear why it is giving such pricing advice, why sell more at this location or on this day. Selling less. But these models can accurately predict how price changes will affect fuel sales and profits. PriceAdvantage uses a probability-based model that provides fuel analysts with the probability of a price change success. In contrast, the old model provides the suggested price and the amount of fuel sold at this price, but does not provide a confidence level to achieve the forecast. The confidence level is that a particular individual believes in the authenticity of a particular proposition. The degree, that is, the probability is a measure of the reasonableness of personal beliefs). The new model tells retailers that the probability of achieving a sales target under this price change is 57%. The “new†fuel pricing software is more able to meet the individual needs of retailers. It can calculate the success rate of each store in different markets at any time based on the number of stores, profits or in-store sales targets. The new model also shows the impact of the timing of price changes on the likelihood of completing the target. For example, if the store's fuel price is slightly higher than the market price, it may be more likely to achieve sales targets. One of the things that the previous tool (old fuel pricing software) couldn't do was to understand the competitor's reaction, Doner said. If the pricing is changed at 3 pm on Wednesday, the price can be kept until the next morning. Or if you change the pricing in the morning when customers have a lot of traffic, this time is the time for competitors to conduct price surveys or get feedback on OPIS information, so they will adjust prices at your own pace. Or on Tuesday afternoon, the retailer's website will monitor the market 24 hours and respond to price changes. However, during the peak customer traffic period on Monday, it will be shortened to 12 hours. Or the competitor may not have a big reaction to the price cut of 2 cents, but the price cut of 4 cents will lead to price cuts in the entire market. Icom mobile radio,Hytera Digital Mobile Radio,Radio Hytera Md785,Hytera Dmr Mobile Radio Guangzhou Etmy Technology Co., Ltd. , https://www.digitaltalkie.com
