51黑料不打烊

Transcript

Now let鈥檚 get started. Katja, take it away with personal scoring mastery with Marketo engage. Thank you so much Stephanie and welcome everybody. Good to see you all here. I鈥檓 going to take you through the ways that you can actually take your global lead scoring model and allow for local flexibility in there. Just to introduce myself briefly, my name is Katja Keesom. I work for Chapman Bright in the Netherlands as a principal marketing technology consultant. I鈥檝e been working there for about four years working with several multinational customers over Marketo, helping them implement and optimize their use of the platform. Before that, I spent a long time working for DHL Express and that is also where I did my first Marketo implementation including having to cater for around about 140 markets included in the lead scoring model. So that鈥檚 a part of where my background in this project came from. So let鈥檚 have a quick look at the basic principles of scoring before we dive into the details. Person scoring or lead scoring is essentially an agreement between marketing and sales and that鈥檚 an agreement on which leads which people in the database are actually useful to hand over to sales and that involves identifying what the data points are that identify a good lead but almost more importantly the relative worth of those data points compared to each other. And when you add all of your scores together, the agreement also covers a minimum total score that needs to be reached before a lead is handed over to sales and all of that of course is meant to have as high a conversion into actual pipeline as possible for those leads that you do hand over. Now it is a an important step to evaluate your model periodically is you can based on your experience set up a lead scoring model but actually monitoring and tracking which leads convert successfully, what their overall score was, which data points the successful leads had in common but also what were the bad leads that were not converting and what data points did you score on that they had in common helps you evolve your conversion rate over time. And based on your evaluation you can identify which parts of your lead scoring model need to be updated. Now how often do you need to evaluate I would say as a general rule of thumb try and do that every year but whether that鈥檚 useful or whether you actually need to do it quicker depends on the volume of leads that you hand over to sales because in order to do a proper evaluation you need to have a good volume on which you can base statistically significant conclusions. So we鈥檙e talking about scoring and the different data points but essentially scoring falls apart into two different halves typically in most lead scoring models that I鈥檝e seen. One part is a demographic score and that actually reflects how interesting a person is to your company. So in other words how close to the ideal customer profile you have in your company is to your profile you have is this person that is engaging with your content. The other half of that medal is actually the behavior score so that is how interested are they in you so how much are they engaging with your content and the addition of the two scores together is your total person score or lead score. I just mentioned in the previous slide that you need to set a threshold and you need to agree that with sales when you鈥檙e handing over a person to sales for follow-up. You can do that on any of these three scoring components or even a combination of them. The most common one you see is on the behavior score but I also have customers who are really interested only in very specific people in specific roles within their target accounts so they have a very clear minimum threshold for demographics before they hand over a lead and they will actually hand them over on the first interaction where they collect that information. So you need to have a look at where do I set my thresholds so that the leads are as relevant as possible to my sales teams. Now what are those typical data points or data topics that you鈥檙e scoring on when it comes to demographics those are usually values in fields in your database that describe information about the company that somebody works for or their role in that company. So typical ones are industry company size for the person what is their job role what is the department they work in what is their seniority but you can also go into very specific demographic characteristics related to the business that you are in. Some companies may want to score on the amount of buildings that a company owns or the size of their vehicle fleet so it depends entirely on your business model what the demographic data points are that you want to include. When it comes to behavior you will not find that so much on the person data fields or the company data fields in the database behavior comes out of the activity log of a person so that can range from email clicks to pages on your website that somebody visits registration or attendance of events both online and in person engagement with your social channels so all of the ways in which you can communicate with your customers and your prospect bases can lead to responses that cover the behavior scoring. Now when we did this exercise with DHL the first time we had a mix of people from the global team and people from country teams involved in that lead scoring exercise and the ones that came from a country or a region actually started protesting here because on a global level there were very clear choices to make on what industries are relevant and other demographic criteria and even in some cases behavior but not all of those to the global standards are equally applicable in local markets and in the end your lead scoring model hands over actual individual leads to sales development reps in a market and that鈥檚 where you want that conversion to be highest so that鈥檚 where the use case came up to say it鈥檚 nice to have a global model but I need to have leads that are relevant to my local market. So let鈥檚 investigate a little bit what local market means here. As I mentioned there鈥檚 industries in different countries different industries are relevant some countries have a very strong financial market otherwise other countries have a very strong healthcare or fashion market so there鈥檚 quite a big variance there but also something that you wouldn鈥檛 necessarily always consider that quickly communication channels can change quite differently and the relevance of communication channels can quite dramatically change depending on where you are from WeChat to Facebook to LinkedIn broadly speaking in different markets you will have other important channels that people can engage with you on but let鈥檚 also look at something other than a geographical difference because within your company you can have different brands or different product lines that you鈥檙e supporting and for those different brands and those different product lines it can be different departments different stakeholders that you want to engage with and who are actually involved in the buying process. So there鈥檚 all sorts of angles where a local variation can play a role in which leads are actually relevant to those sales people on the ground. So time for a little survey among the audience here. Do you have a lead scoring model in place in your Marketo instance and does it allow for any local flexibility? Please select the option that applies to you best. Let鈥檚 give that a little bit more time. I see there鈥檚 still people submitting their answers but let鈥檚 have a look at what we got. So it鈥檚 good to see that most people actually do have a lead scoring model in place but it鈥檚 quite a mixed bag on whether local flexibility is baked into the model as we see. Each market having their own model is also quite a low percentage. A challenge with that is that on your entire instance it鈥檚 going to be really difficult to compare how your different scourings are stacking up against each other. So there鈥檚 quite a bit of opportunity to see whether you can improve anything there. So what we did and what I present to you as a possible solution here is to create a scoring matrix. So with all those topics that we just discussed, in this case the example is a demographic matrix, the idea is that each of those topics you can assign a priority to and each local market can assign a different priority and for the different values within, let鈥檚 take as a demographic topic industry as an example, you can assign different values, in this case high, medium, and low values to the different industry values that you can identify. Now when you set up a list allowing for the countries to set a priority for a topic and to assign different values to the different scores to the different values in their list, there鈥檚 a couple of things to bear in mind. I would recommend to take a different scoring model or matrix for the demographics and for behavioral elements. That is because it is all hinging on the relative worth of your different components and when you compare demographics and behavior you are still comparing apples and oranges. So it鈥檚 really difficult to compare the different elements across demographic and behavior. Another question to answer is which topics that we鈥檝e selected for the lead scoring model are going to be equal for all markets and which of the topics do I really need to allow for that local flexibility on. Once you鈥檝e identified that you can have a look at how many of the topics do I have that need to be set up locally and based on that you need to think about how big does my matrix need to be. So how many priorities can I choose between, how many different value levels can I choose between. My recommendation would be to keep that within two by two sort of as a minimum and at the maximum five by five. Why? That again relates back to the relative worth. If you set up a matrix with nine different priorities it becomes pretty arbitrary whether a specific topic becomes a priority six or seven. So the more compact you keep your model the easier it becomes for your local market experts to actually distinguish one between the other and make clear choices. When it comes to putting the individual points in there you need to make sure that those points for the local scoring variables align with what you have in your global model. So based on where you set your threshold what your high and low global points are dictates what sort of levels should be in this matrix. And one really good way to test whether the score levels that you鈥檝e set up actually make sense is to take some common business scenarios. Take some buyer journeys that you see frequently and actually go through the steps the data points that you collect in that journey and the engagements that somebody has. Calculate the points that they would attract for that and whether they would meet the threshold to hand them over to sales or not quite yet or still be quite a long way away from being handed over to sales and discuss those scenarios with sales. Would these be indeed the people that you want to see handed over or not quite yet? Discussing it based on real life scenarios actually is a really great validation for your model. Now once you鈥檝e set up your matrix of course you need to collect the information from your local markets. You can do that with a matrix that looks more or less like this where you list out those local topics demographic and behavioral topics that need the local input and you ask for the local analysis where the first question is is the local market actually allowed to say this topic is not relevant in my market at all so I鈥檓 gonna exclude it from my scoring matrix. The next one is on the high level topics what is the local priority that I give to that topic? So it would be wise to actually set a global standard here that countries can make changes to. It always helps to have a reference point so the question is for the country give me the relative worth of industry versus company size here. Now when that is done you go to the next level so within the topic. I can see a question is this a scoring matrix based on a threshold of 100. The values that are in here are examples only so whatever ends up as scores in your matrix needs to align with your thresholds and with the global scoring model as a whole. So it can be 50 it can be 100 but you need to look at that consistency. Now going back to collecting the local input once you set or the market has set their priorities for the topics you go to the individual values in there and again I would set a global reference point and allow the local markets to adjust that. So you can go high, medium, low depending on the different variations you allow and why am I showing high, medium, and low here and not so much the individual points that you saw in the matrix in the previous slide. Again that all goes to I want the people in the local market to focus on the relative worth whether it鈥檚 six points or eight points that comes out of the combination of the priority and the score is not so relevant. The relevance is what is the relative worth of industry versus company size and what is the relative worth of technology versus healthcare. And in some cases for some topics it鈥檚 actually also necessary to get the local input on actual thresholds. So in the example here for company size if you have an exact number of employees you can actually let the local market change that value from 250 to 400 or any sort of local input any sort of local input that you would want on the detailed values. Now that you鈥檝e set up your model it鈥檚 time to actually implement it in Marketo. So how can you do that? What I would suggest is have two versions of your scoring program. I would build it out as one central program first but then separate it out into your global scoring model with all the topics that you鈥檝e identified that do not need any localization and then create a copy of the program but only leave the smart campaigns in there that handle the topics that can be set on a local level and that program is actually one standard program that you copy into all the folders where that local scoring needs to be implemented. Now I鈥檓 a big fan of naming conventions because they keep things nice and clear for everyone and showing the consistency and the relative the way the programs are related to each other. So you see here the two versions of the program have naming conventions for the smart campaigns to clearly show that there is a correlation between these two programs. So with the B5s and the D3s keeping your Bs for behavioral stuff the Ds for demographic show consistency and make it easy to understand what鈥檚 going on. If you don鈥檛 really know how to start with this there is actually an import function in Marketo where you can tap into the program library that 51黑料不打烊 provides for you and you can download a basic lead scoring program from there and optimize it based on the exercise that you鈥檝e done with sales. So that鈥檚 a great starting point to use if you don鈥檛 want to start with an empty program. While we鈥檙e on top tips I would also strongly suggest and I鈥檒l show you a little bit later why to set up all your tokens or all your all your scores as tokens in the program. So rather than in your smart campaigns put actual scores have one central repository again using good naming conventions to have a clear overview of all the different scores that are in there. So again you see the Bs for behavior and the Ds for demographic and you will recognize the high low and medium and the priority one two and three that you saw in the matrix. So again everything relates to each other and is very recognizable. Once you鈥檝e got that set up it鈥檚 time to start working on your individual smart campaigns and let鈥檚 talk about the smart campaigns. And let鈥檚 stay with the example of industry. The first thing you do is set up your smart list and for demographics as mentioned before that鈥檚 information you will find in fields and that gets populated either when that person is first created or you can have a data value change on an existing person where the field was empty before and is actually populated after. So you鈥檝e got the two triggers and a filter to work together with the person is created trigger to make sure that there is something in that field to score on. So make sure that your triggers and filters are set up to score the information on that field once. Once you鈥檝e done that you go into the flow and you see that in the flow there鈥檚 two activities or two flow steps for changing a score where the first one focuses on the demographic score and the second one focuses on the person score. And you can see in the video that changing the priority of a topic in this case industry is simply updating the value of your token and selecting the right one that pops up from there. And if you have changes compared to the global model of which actual values are scoring low, medium and high it鈥檚 simply a matter of taking that value and putting it in the list for the relevant choice. Now there鈥檚 a couple of really simple housekeeping tricks that you need to keep in mind. One is make sure that your choices are mutually exclusive. That value for technology can only be in high or medium and not in both. Similarly make the updates that you do both in the demographic and in the person score. The person score is an addition of behavior and demographic score so any change to the demographic score needs to be reflected in exactly the same way in the person score. Now once you鈥檙e done with that and you鈥檝e activated your smart campaigns your lead scoring model is active and you start building up a database of scored leads based on which you can evaluate that model. Now how did we fare with this model in DHL? I cannot give you statistics just yet because as I mentioned evaluating your model requires some time to build up a significant volume of data and we鈥檙e almost fully deployed in all 139 markets for DHL express but not quite so some markets are not live yet and several other markets have only been live for a short while. So I can鈥檛 show you any stats yet but from the onboarding process I can tell you that almost no markets went for the full global standard. Pretty much everybody made even some minor changes to the scoring model and some actually had a very clear idea of how they wanted to optimize that for their local market which in turn gave them trust in that scoring model that it would actually generate relevant leads for their markets. So the adoption is really high. One date on the statistics in a follow-up blog post or through any other means. Now the key takeaways that I would like you to take with you after this webinar is make sure that you have a scoring model that is consistent throughout your organization so that you can evaluate it across the board but don鈥檛 let that hinder you to allow for local flexibility. One size fits all is not going to work in a complex organization so build in local relevance but within a framework with that clear matrix so that the leads that go to a local sales development rep actually are relevant to that local market. And when you build it in Marketo create your overall program first and then separate it out into a local and a global version where the local version gets copied into your different market folders. Use your naming convention to keep it clear also for someone who goes in months if not years later how the two scoring models fit together and how your different scores are built up and evaluate your model periodically once a year or depending on your volumes more frequently or less frequently but do it on the global level but also on a local level so that you can see whether those local adaptations actually make a difference. So having heard all of this and I do realize it鈥檚 a lot of information in short amount of time but walking away and going back to your desk how prepared do you feel to implement some changes in best practices to allow for local flexibility in your scoring model? Let鈥檚 give it a little bit more time as it looks like lots of people are still thinking about that Let鈥檚 see what the result is there. The good news is that everybody feels at least a little bit prepared but there鈥檚 still a little bit of work to do for quite a few people. That鈥檚 a good thing then that we鈥檝e got some further resources for you. If you don鈥檛 have a scoring model in place yet at all and there were some based on the previous poll there鈥檚 a lead scoring exercise that you can download that you can use as a guide to sit down with your sales team to actually go through what should be the topics to include how do we put the different scoring values into that. So leverage that to have that first conversation then go into the program import library to get that example lead scoring program into your instance so that you can set up your scores and build out your smart campaigns in a relatively quick and easy way. I would say it鈥檚 important to start with the basics actually start collecting some data so that you can evaluate your initial thoughts on the lead scoring model. This is an iterative process so it is more than okay to go in and to optimize but it鈥檚 important to have a baseline. Now if you already have a scoring model in place the things you can do is have a check when was it last evaluated what was the outcome of that were any changes made to the model and have those changes actually led to higher conversion rates and monitor your conversion rates over time but also between the different markets. If you don鈥檛 have any local flexibility built in but your conversion is very different between your different markets there is an opportunity to see if that localization can actually boost your conversion rate. So always plan ahead as well to review and update your model periodically and building in that local flexibility when you see that there is a need for it. As a recap I鈥檝e got a few resources here the exercises the lead scoring program I think they鈥檙e also available in your resource section in the webinar screen and I鈥檝e also included some other inspirational sessions from previous summits but also from different blogs on also from different blogs on the marketing nation that will inspire you and give you ideas on how you can set up your lead scoring model. Thank you I鈥檓 looking through some of the really good questions that came in through the Q&A. I鈥檝e got one here I hope I haven鈥檛 missed your comment on this do you have a global CRM where the score is passed and if so how can the localized scoring models be compared globally or are they used only at local level? Now the whole concept of this is that you have one global model but that you allow for local adaptations within that model so even on a local level the scores are still fitting into that global model just several topics get a localized version of the weight so you still have the same thresholds at the same handover moments all through your organization so that鈥檚 how you can actually compare different markets to each other because it鈥檚 one model but several different markets can give different weight to different topics so where one market can have 20 points assigned to one topic another market can have that based on another topic. I hope that explains the concept there a bit better. How often do I recommend going back and re-evaluating and updating the model? Yes it is indeed more of an art than a science but as a general rule of thumb I would start with going annual if you have a whole lot of data coming in thick and fast you may want to do it more often but make sure that you have sufficient data that you can draw proper conclusions and that you鈥檝e got statistically significant outcomes. How would you incorporate a lead scoring model to identify what the next best product or service is for a current member? That鈥檚 actually an adjacent topic where I would go into setting up something like interest scoring or have two separate models running at the same time. Let鈥檚 see and a question are these clones of models across workspaces? Yes the local one is replicated is cloned across different workspaces the global one sits in one place where you can score across all workspaces and the local one actually gets cloned across where every workspace has its own version. Do we have any other prioritized ones? Not by the looks of it but I鈥檝e got an interesting one here. How would you work with sales team to determine the end goal KPI for the model? I would say that the end goal KPI you can set a target but it鈥檚 also a moving target because the goal should always be that with every evaluation that you do of the model you increase your conversion. So you have a baseline with your existing lead scoring model or with your existing lead handover process and you always want to optimize that by leveraging your lead scoring model. So I can鈥檛 give you a number but you need to work with how much room for improvement is there. I think that鈥檚 the key question to ask your sales team. Yeah my opinion on AI driven solutions to automatically calculate scoring for us I think that鈥檚 the big question indeed because the AI model is going to be a lot better and less biased in identifying statistically what works and be more predictive about it but there鈥檚 not a lot of experience in that just yet so you also need to train that model quite well. And I do see one more prioritized question if a contact is shared between two or more workspaces that鈥檚 an interesting one that depends on how the workspaces work in different product lines how different they are from each other that would be something to explore with the sales team or sales teams actually from the owners of the two different workspaces. There鈥檚 always people that come up with the difficult questions around the fringes. Similarly the question on when you have leads that will exist in different segments would you have different models for different segments that depends on how you define your segments. There鈥檚 a lot of creativity that you can apply there and I can鈥檛 give you a definitive answer on that one without understanding what your different segments are based on. Okay I just got the suggestion to take one more. If you make significant changes to your model over time would you always start fresh or do you update the current model? I wouldn鈥檛 start fresh and I would update the current model but document quite clearly when you make changes to the model. So I wouldn鈥檛 go in and make one change today and another change in two months time and another change in six months time. I would implement a set of changes but I wouldn鈥檛 actually reset your model because then you lose a lot of changes. Sometimes depending on the changes that you are making you can make corrections but I would only do that if you really confidently can manage the impact there. I think that is pretty much the end of the time that I鈥檝e been working on. I think that鈥檚 it for now. Thank you very much for your time. I think that is pretty much the end of the time that we have with you. So thank you very much for your time and for your engagement and your input and any question that you raised that we didn鈥檛 have a chance to answer keep an eye on the community because we will actually go there and provide an update with answers to all of the open questions that are left. So go there for your open questions. Make sure to actually click on the resources and download and watch what is relevant to inspire you and to actually get started tomorrow. Thank you so much everybody.