Enhance your vendor negotiations with sales automation services in Vendor negotiations

How to create outlook signature

[Music] thank you hi my name is Jonas I'm a principal engineer in the selling partner Services organization I have been in Amazon for 10 years today I will talk about the automated negotiation system my team builds I will discuss first time negotiation problem in general followed by our high level architecture then I'll talk about one of the components of negotiation offer Encounter of Our Generation and the challenging problems we are tackling in that space now let's build some context Amazon has millions of selling partners for those who choose to sell Amazon instead of customers on the Amazon Marketplace we negotiate the terms of sales we call this pet selling Partners vendors Amazon has more than you know hundreds of thousands of vendors that supply products negotiation may include hundreds of terms as well from product cost to more complex terms like how to deal with damages and return items and you know where and how products are delivered payment terms hundreds of terms figuring out the best agreement that most parties can find acceptable is complex even for a single product we are sending now imagine negotiating efficiently to procure hundreds of millions of products from hundreds of thousands of vendors now my team owns systems that help both vendors and internal negotiators to complete this important business operation effectively before I go to describing how we approach to solving this problem let's talk about negotiation in general the goal of negotiation is reaching an argument that is acceptable for both parties the first thing that comes to your mind maybe the human element being tough or friendly to improve the negotiation outcome while it's true both parties try to maximize their gains it is important to build long-term strategic relationship that benefits the end customers so in reality negotiation requires preparation and lots of data for example you need to set goals you have to understand the goals of the other party and you have to identify opportunities for negotiations such as products where profitability can be improved and this requires operational and financial data from several Amazon teams that we have to integrate with negotiation often involves creating value for Mutual gain not just sharing what exists but what do we mean by Mutual gain let's say we're selling coffee mugs for ten dollars and we're buying them from our suppliers for eight dollar that means we make two dollar profit per month not counting other cost us like shipping storage marketing and infrastructure now let's say we want to lower our cost to seven dollar per month this could be because we are planning to reduce prices for our customers or or because we want to improve our marriage and too much or profit Target the simplest option is to ask the vendor to reduce their cost to 7 which would give us an extra one dollar profit per month however this is not a win-win scenario but what if we found out through data or through the negotiation process that suppliers typically spend three dollars per month to ship the item to Amazon if Amazon agrees to buy in bulk and have them ship product to a single fulfillment center they can reduce their fried cost down to just a Dollar by doing that we will be creating a new value so what you can expect from a negotiation system that it would identify opportunities like the one I talked about I'm simulating the financial impact of bulk buying and find out if it would be beneficial for both parties so the system we're building has a dual purpose we support human negotiators here at Amazon and we also automate negotiations in some contexts to support human negotiators the system can automatically identify negotiation opportunities and alert human negotiators this can range from simple things like missing profitability targets to more involved with systems like identifying the best supply chain option for a vendor the system can also help in setting goals by benchmarking across similar products another example is generating the optimal proposal to send to a vendor now the system can also capture historical data boost for reporting and training purposes in selected scenarios the system can fully automate the backend Force negotiation from identifying opportunities and creating an initial ask to evaluate encounter offers and approving agreements and creating the required contracts let's go over the key components of the high level system we have negotiations can be started by Amazon or by the vendor let's see what happens when it's started by a vendor a vendor can make a request either to sell a new item or change an existing term like product cost a request this can come through API or through the UI and are received by one of those ingestion systems the negotiation management system receives this requests and orchestrates the workflow required to make a decision the first thing we do is evaluate the long-term Financial impact of The Proposal as you can imagine evaluation is the complex system that requires forecasting and understanding macro Trends based on the evaluation result we can decide to accept a proposal or provide a counter offer it's also possible that we will sideline this for a human negotiator to take a look in some cases now if we decide to provide a counter offer the offer generation system will use data such as you know the cabin offer operational financial performance I talked about earlier science models to come up with a counter offer now the back end force can continue until we reach an agreement or the negotiation is abundant we will store the contract Downstream systems like purchasing can use the contract to take actions after that we constantly monitor via signals like the customer shipments financial data operational data like fulfillment to identify new opportunities for negotiation that Amazon will initiate now the opportunities we generate are used in two ways we use them to generate offers for negotiation we will also evaluate all opportunities holistically to set goals and start new negotiation and plan yearly now each of these boxes I have shown you here the Encompass multiple services in total we are talking about hundreds of services to accomplish the stars now that we looked at the high level systems let's dive into the challenge of one specific area the automated offer generation problem the goal of the automated of Our Generation system is to come up with the best negotiation proposal we can offer to a vendor that also maximizes our long-term profitability it used both for Amazon initiative negotiation as well as to generate counter offers when the system generates this proposals it's time to maximize the probability of acceptance as well as the long-term financial goals hence it shouldn't make egregious askers that will damage trust or negatively impact selection or profitability the ask set generates also have to be explainable and one way of doing that is like by including clear justification for example Computing the forward-looking vendor benefit of our proposal from their point of view hopes the system is also expected to use various levels and incentives we have available that will make sense for the given scenario we have to do all this while meeting the scaling demand and the size of the negotiation and the number of terms that we have as you can guess the offer generation problem I described it above includes several machine learning problems we are working on some of this includes coming up with the reasonable cost for all products Amazon selling so we use product attributes similar products past sales history to train a model that will give us a range on what is a typical or Fair cost for a product but having a fair cost range is not sufficient by itself we have to estimate the probability of acceptance we build probability of acceptance model based on past negotiation history product and manual attributes many other data points a related topic is like Financial simulation we have to evaluate the long-term impact of whatever proposal we generate while we arrive with simulation requires models for forecasting demand as well as our cost structure in addition to the above like we did with cross-term optimization there are multiple ways of achieving the same objectives on the engineering side we have to scale the system both to the business size as well as make it extensible to New use cases for example we have to deal with how do we capture high quality training data to drive our ml systems how do we model a negotiation so that it can easily support new terms as we are adding a new negotiation term every now and then if you are interested in helping us solve the challenges I discussed in this video please check out our career page shown here thank you [Music]