Streamline your b2b sales forecasting in UAE with airSlate SignNow

How to create outlook signature

thank you very much so let's go to the topic okay it's working so big rounds now before I just wondered current companies and it seven years ago I was working with Microsoft for 14 years and my last job was since excellence manager for multi-country tomato sauce so what does it mean it was the one of many possibility was that I was responsible for a revenue of certain group of countries from Kazakhstan then by the countries then smaller countries in this region that all X part runs and here Cyprus and Malta and what does it mean in practice is that everyone we had this internal life meetings when we discuss with countries so what's your forecast are you sure I'm going to do it why are you sure we're sure this is going to happen I'm not sure it's going to happen never I would be cooking the colleagues all smart people in my room but what could you do you cannot overrule a country because they know their business see and it was like a month after month all that pain was for me higher or deeper affirms and started in medicine machine learning is doing breast cancer case studies things like this so the question was really what can we do with the machine running in the fifth of business-to-business sales forecasting and that's my motivation coming from the businesses for very specific eating situations which I was daily responsible for for and the situation is actually described nicely by this short sentence that where we were actually on experience information overload but moment starvation so all that we have everything but we just don't know anything and the famous situation from and so so after life of the Microsoft I was starting searching for technical part very much easily but then I was more ambitious I didn't want to have this solution which will tell the forecast into instead of people but I will try to this smart people here to learn something so we kind of construct this high-level picture of the research so we have this B to B cells experts the CRM system and based on that we would build different classification machine learning models watch how they perform which one is better which one is more able take closer to people in terms of acceptability or maybe prior knowledge or maybe they have just measure the performance which they to metrics or do the IOC analyzes so based on that the key issue here it actually actually merging the University of organization Sciences birthday Neil Howe recommendation learn and computer science because what we put into the center of our research is actually explanations of our forecasts and wise we see that because when somebody from let's say Croatia would take Marco knows what is good is going to happen I would say well you're selling solution P it's red are you sure you can be so sure in your statement it's going to happen because the data does not support it and then next month day understand well we cannot tell any for Marco it's going to work we need to figure out something else other words we are just say well it's not going to happen now maybe in heart appear so this way we kind of created that brings data like combine the data from the Star Trek's when he was asked for opinion he would always give them or prefer them on a street back best tested based on data not biased with human expectations and then here we still have humans for making decisions and then we look this back we have few signs people are running the organization eventually gets better on 46 so during his PhD studies which I completed two years ago a lot of the research papers were published the most prominent was here in terms of theirs in difficult situation is publications just recently this year's spring a book take a spring even played like latest book machine and human learning showcases our case as a way to help organizations to learn from the data and works in Croatia and just three months of for master cooperation society of research published one of the research papers which is sub-province with infants that area and also I have just Oh making commercial will be commercially viable I put it as a service since empty so sells new technology calm which you can very subscribes for for one on trial and seat hogs work in the practice it's a skeleton version I can explain later ether questions so first we want to be very efficient with the data assessment so we've seen in many presentations today that the data is very important that everybody says that a cart an acquisition later in 2014 as well actually in practice I don't have the problem which is called big data problem I have a program which is called scarcity the problem there is no data I mean they are datin CRM but when you put into machine learning models they don't learn anything it's close to zero zero two point five which is close to the controlling why because the structure within the CRM is not focusing on the right which are then applied work for there for them machine learning so what does it feature to say this one a nice take the business to build the business and says dataset and we calculate the distance between cases so the lighter the better and we see we have this we have this black line which means that which just pros that concept before networking right because those are comparing two same cases and needs to be zero so there is no difference between cases then we have some stripes here which are showing us the light this from the sales perspective they've done some activities and they really went for the floor with the plans because the data cases are different so with this when you understand the budget you need two seconds to stand what's the quality of your medicine internal organs and differences this black dot here just means that somebody crews push my manager hey you need to update your serum you will just coffee in cases when I've been to present this picture well yes I was pushed I needed to do but it's shown here and it's not really good for the moisture garlic processing they are the next steps of assessing that which model is working capacity so it used random forest we based decision trees as support vector machines and neural networks and in classification accuracy and also in area under the curve random forest was among the top from most people understand this one but videos also they will see because we were happy cannot claim that our data distributed normal and for that you see is much more precise measurement quality of the data then then other matrix so I have intentional actions that super big machine is working poorly it's not really working permit or day but you need to have much more experience how to actually put the parameters into that to achieve compatibilism so it's easy for starters so for beginners tool to miss it it's a good method but here when you're comparing different models you need to be cautious about do you really can trust this victim so what we have here just for the illustration purpose we have the set when we collect five different features whether the cells is complex whether who is seller to be set to existing client which solution we are offering to the kind and what authority of the person we are talking and the client side and for example authority if I say Allah means somebody's just gathering information has no impact within a viable organization to purchase somebody quits hi it's actually the post signing the contract and say yes we have budget for them and we sample it who is doing it has the same responsibility from the for tech product side but he needs to ask for the money and just to give you the explanation of that and that's critical in practice you need to have workshops to explain over and over what doesn't this labels mean because they take to meet more iterations tool then using correcting the brackets so if we are just taking one letter exists we have here the crew sites on site and to read site the fan site it just shows graphically how much this attribute contributes to the positive outcome of themselves and also here how much contributes to them to the negative outcome of things for example if we are saying to existing plans which is yes then it looks like it's much more probable that we are going to be in the middle but it's in some level of cases or probability which means that we are not getting the deal or the we are selling to that existing plan while if we are thinking there let's say you know know it's slightly more has tension - but of course we always need new client so we need to win some so we have some portion here as well and what does it mean now in practice on some specific case because this is one of you we have a specific case where the complexity of sale was moderate and the seller account manager was selling to existing client solution P and he was talking with authority me and this deal was won and we won was 82 points and now we can see the most important element of this story here was the fact that this is the existing client and then that seller was a.m. and then the authority was me and then the competitive [Music] so yes 49 so this is kind of what if analyzes what makes sense for me to work and before managers make decision they can also play they do if we change salesperson what does it mean things like this and then from that perspective perspective you have a more educated guess about nine steps per minute okay so this is now ready I mean were completed for two years and last four years I was really focusing on on putting death into life for people to use it I have to say this graphical explanation is over killer for people so and for the moment it's not yet available but can be easily I mean it is added for some clients but they require the training so what we do now if you go for this site you reduce them and then you put your opportunity and then you can do it as well it's going to work immediately then you have your opportunities arranged by the maturity score means propensity when you have this picture then you can have a view table below so what does it really make sense for me to focus to close the deals and for example I would say okay then most food store its upset yes but but this is not realistic this is what it's not realistic I'm not going to the best but then which is the Big Bang when the competition currently is very unknown if be achieved with sense activity they tell us hey you not have competitions but be smart in putting the offer together so bishops know and this gives us 19.7 point in our probability will propensity score so as a seller I go for his fruits so which so at least effort I have a more album and then there are let's say who's really buying motivation when they say that needs to understand why I is really customer that's that's not real Christian it's really not the question and if you can put more she can confirm is clear to him then in this case would be two point eight point so based on that we can move on or the opportunities for the rest of the month so vision from all meat and high-end this is what we are listening to and then when we put that into the current performance all we have is opportunities and then you say yes there were one they call it will be six tries so in orange we can plan here to date it will be that achievements your date and then in green we just add those numbers when the Machine manic models would say this is highly close Apollo and you can realistically come to that then it's added to the graph so you can see based on your partner targets monkey plans plans and so on how you are going to finish the here so here you have to be a part the plans for the full year so we let include three swings in sight which use it try it it was out of the box without parametrization more time to focus on as we have then with such which is automatic on wave on that and then it was performing much closer to others but still not being the best for let's say for the most random forest redundant 250 400 so trees because you put it that and this will be the quiet and there was no huge impact on such a variation yes by the way the explanation model works on respectively which more than you take as a machine will reform so that's a huge benefit so you can one day use the pieces you that one day I use random forest I don't know because the new data causes them new base works better then next day I hope I see same picture but behind is my I basically see that's the beauty of that model development profession of each one there was perfection so currently you I like in watching I have a Genesis book see the problem is you start which is actually in the public domain because one of the issues is business-to-business there is no sex off during the research I gave one every day somebody three four people today was from India and from Vero Beach we already numbers that and then they do they what they do with that for further research and other is to take their from CRM we don't do that but we after we make the prediction power assessment so if you want to be my time with your day that you need we have the pacification is procures it is 70% which is the minimum standard for social sciences if we are not achieving that that's no goal and then this model it's not delivering on the promise income standard this zero concern there is a third way would be upload with Excel Provenza but we can okay thank you [Music] [Applause]