MedPro - Self-Service Fuzzy Matching - Inspire 2017

The insurance industry is full of data that comes in all shapes and sizes. When this data needs to be matched or duplicates need to be removed, Fuzzy Matching in Alteryx is your best friend. However, Fuzzy Matching is... well... fuzzy and doesn't lend itself to flexible, automated solutions. Come see how MedPro Group — Berkshire Hathaway's dedicated healthcare liability solution — uses the power of Alteryx macros and the Alteryx Server to take the Fuzzy Match tool and transform it into a comprehensive matching and deduping system that anyone in the company can leverage for all types of business problems. For new and experienced Fuzzy Match users, this session is a great opportunity to conquer fuzzy matching, learn how to expand the capabilities of Fuzzy Match tool, and experience the Alteryx Server from the company who earned the Alteryx Excellence Award for Best Use of Server in 2016. You'll also learn the power of Alteryx batch macros and how all these skills can be leveraged to solve a variety of business problems.



Video Transcription


Jim Kunce:
Hello everyone. As I've said my name is Jim, happy to be here with you today. Thanks for joining us. Dalton and I are two of 25 actuaries at MedPro group. While we're a small team, I'm really proud to say that Alteryx has been part of what we've been doing for 10 years now. Checked it the other day, our first module had a, was in April 2017.

Dalton Moss:
July 26.

Jim Kunce:
We've been Alteryx users for a long time. Then last year, the team won an award, the Essence award for best use of server. Then as we saw announced today Diagon who's on our team at MedPro was one of the new aces. We're really happy to be here be part of the contributing to the continued knowledge and community that's Alteryx. If we go to the agenda, I'm going to kick it off and talk a little bit about who MedPro group is, hope probably nobody in the room has ever heard of us, talk a little bit about some of the challenges that we face. Then I'm going to hand it over to Dalton who's going to talk about how we have used Fuzzy Matching to address some of those challenges and then how him and the team have improved upon the fuzzy matching tool and created the fuzzy match master.

Then he's going to get into a really deep dive into how the logic of that Fuzzy Match Master works. One task will allow you to do is learn how to modify it and create customization so that you can use it for your business problems. Lastly, he'll show you how we've implemented that on the server as well.

As you can see, MedPro group is part of Warren Buffett's Berkshire Hathaway company. We specialize in medical malpractice. At the highest level, we're an insurance company and insurance is a promise to help and pay when things go wrong. In the United States, there's around 75,000 accusations of wrongdoing every year against healthcare providers and about 15,000 payments to injured patients. That sounds scary but the vast, vast majority of healthcare providers in the Unites States are highly trained, they're very careful and they want to provide the best healthcare possible. As we all know, sometimes bad things happen to good people and that's when you need insurance.

On the other hand though, there is a very small percentage that we all want to avoid that is sloppy, lazy, even deceitful. Amazingly enough, it's that 2% of high risk doctors that comprise 50% of those 75,000 medical malpractice payments. In order to keep cost down for our clients, we want to keep those high risk doctors out of our company as clients and that's part of how we use then Alteryx to help us do that.

Our market is the roughly 8.5 million healthcare providers in the Unites States. As we engage with that market, internal and external information can come into any of our processes at any point in time. Keeping track of those high risk individuals among the 8.5 million is like keeping track of a needle in an ever-changing hay stack. What we try to do is complete that process and have data that is synced up and clean across all that because every piece of information in there comes together and we use that to help us manage risk and find opportunities for growth.

Before Alteryx, the team was really struggling with keeping our data clean. Here's a couple of examples. First of all is list matching. We have a bunch of sales folks who go around the country and meet with doctors. They go to conventions like this. One example might be we have a sales agent who comes to the annual chiropractors' convention in Las Vegas. They talk to a bunch of doctors, they sign their names and they come by and visit our booth. We want to record that they've interacted and learned about MedPro. Then we go back and we want to put that into our CRM system. Unfortunately, the doctors don't always include the same information, they don't always fill out all the fields. As we saw all the number of records in our CRM database, 8.4 million is huge. That job never really got done. We weren't able to really keep good track of that.

On the other hand from a systems perspective, we may have claim systems, billing systems, underwriting systems, all that are different. Oftentimes, we can have the same name or a different name for the same account across all those systems. Being able to keep track of all those systems and duplicates was a real challenge as well. As we all know, garbage in, garbage out, and that's when we realized Houston, we've got a problem here.

Whether the problem is list matching or system deduping, the answer was really the same, use fuzzy matching, define the matches and either take them out of your system or record that information that you needed to. With that, I'm now going to hand it over to Dalton to go through the aspects of fuzzy matching.

Dalton Moss:
Thank you, Jim. All right. Click pull to help us understand [inaudible 00:06:04] a little bit. Raise your hand if you have done fuzzy matching in Alteryx before. Awesome. Raise your hand if you have done a batch macro in Alteryx before. All right, still [inaudible 00:0:17] hands. Awesome.

As Jim said, fuzzy matching was the obvious solution to both of those key business problems trying to integrate all of that data from that 8.5 million doctor market to our own systems. We're not going to be teaching fuzzy matching today but you are going to get some insights into how fuzzy matching works. We're going to focus on showing you how we've taken fuzzy matching as it exist in Alteryx and ramped it up a level and implemented into the way that we do business.

Just to give you a basic high level approach to when you think about doing fuzzy matching, you start with knowing your data and what fields you want to match on. To use the chiropractic association example that Jim use, we might get the doctor's name, first name, last name, we might get the address, or maybe the type of chiropractors or maybe we're attending a doctors' group and so there's different types of doctors, we might get that information, what fields do you want to match on. After that, Alteryx, drag drop, configure fuzzy match tool, boom, done.

Pick all the different match styles, there's a lot you can do in that tool. Now what if you want to do more than one fuzzy match tool? There's a couple of extra steps. You have to label as I've done with this formula tools there, what fields or some way to identify the fuzzy match that you are doing and rank them. Two examples I have on the left, I have zip code, last, first name, and address. That's a better fuzzy match combination, then on the right, last name, first name and address.

I'm going to rank with the formula tool, I'm going to label them as such, and I'm going to call the one on the left one, the one of the right two, and when I bring them together I can now step four, pick the best match. If I have records a and b and a come through on both, I don't really care that it came through on the left not the right but the right side might give you match results that the left side didn't. That's why we need both of them.

To go on a little bit more detail on why we need more than one fuzzy match tool like why can't we just include all the fields we want to match on in one tool. There's a couple quirks about fuzzy matching in Alteryx that you may be familiar with. The first is missing data. If you have [noles 00:08:40] or empties in your data, sorry, those are completely ignored. If one, any one of those four fields that I'm using on that left example has a missing data, that record's gone. You can uncheck that option, change the default but then you're basically telling Alteryx I want to match nothing to something. You're not going to get any results that way either.

Let's pretend that zip code was empty in some of my data, matches might come through on the right example. That's one reason why you need multiple fuzzy matches. The other is just bad matched. If I have a field like two last names either pretty close but not exactly the same because we transferred the information from the list at our booth wrong, fuzzy matching, it'll say, "Oop! That one field doesn't match well enough." Kick the combination out. Every field has to match stronger than the threshold for that field that you setup in the fuzzy match tool. Those two structural issues are why we want to use many fuzzy match tools but then when we do that, we introduce a new problem, combinations.

I got a little data table here where you see space theme going on here with Neil Armstrong and Leslie Peltier. In this data set with just four fields I can generate all those possible combinations for ways to configure a fuzzy match tool on the right. That's about 13 of them for just four fields. Obviously, using all four fields would be preferred but look I got missing data in rows two and three. Those weaker combinations with fewer fields are going to enable Neil Armstrong and Neil Armstrength to possibly match or the two Leslie Peltier records to possibly match, something that would not have happened with the strongest match combination.

When you introduce combinations, you then have to ask yourself, "How many of these combinations do I really need? I don't want to configure 13 fuzzy match tools." Then you also have to ask yourself, how are these more important? It's really easy to see that the one with four fields is the most important but what about all those ones with only two fields? How do you [inaudible 00:10:53] those?

We really wanted to get around those two issues I mentioned with the normal fuzzy match tool so what we decided to do is we're going to look at all the doctor list that we've ever gotten. We get several a week. We're going to look at all the different fields that we've received. We're going to configure a bunch of fuzzy match tools and throw then into all under one workflow so that we don't have to build a new workflow customized for every list every time. We're going to try to self-serve our doctor list matching.

We went ahead an did that and this is what we had. Introducing the MedPro doctor list scrubber. A whopping 48 fuzzy match tools, 48 joints, and Alteryx 11.0 cannot even open this thing. We had to download 10.6, open it up, snap this picture, throw in the slide for you guys. Speaking of the slides, these will be available to you after the presentation. Don't worry, you'll be getting these.

This wasn't going to work. There are actually a lot of errors we found out in this workflow. We were missing combinations. We had to decide for ourselves what combinations we wanted and, frankly, we missed a couple of key ones and we also included a bunch that we didn't really need. We wanted something that was easier to maintain, something that was more flexible and something that was more self-served than our first brute force approach. That's where we created the Fuzzy Match Master.

Last summer, I put this macro together to basically replace that monster of a workflow that you saw before. The Fuzzy Match Master is a Alteryx macro which is a workflow inside of a workflow. It was really meant to do what Jean Stoker said morning, simplicity and sophistication. How can I get the sophistication of that big workflow with the simplicity of a single tool.

Just to briefly explain the macro before we actually go in and show it to you. A macro in Alteryx as I said is a workflow inside of a single tool. A lot of the tools that you use in Alteryx are actually macros. The data cleansing tool, that's a macro. There's different types of macros. One of those is called a batch macro. Batch macro basically says, here's my workflow and I want to change the formula tool to say something else and have this list of different somethings I wanted to say. It'll batch loop through each of those different configuration, tweaks that you give it and run that workflow a bunch of times and then package all those results and give them to you, it's like a union basically.

The Fuzzy Match Master is a standard macro with a couple of batch macros inside of it to loop through all the combinations that you are about. It works with anywhere you can put a fuzzy match tool, you can drag and drop a Fuzzy Match Master as well. How it works is you input the fields and the match styles just like you would a normal fuzzy match tool. There's a couple of new options you get to have. Then, it's going to generate the combinations for you, it's going to rank them for you. It's going to allow you to include and exclude ones and then it's going to give you an output that looks almost exactly like the normal fuzzy match tool.

Now let's go to an example workflow that I have setup for everyone. It's a little challenging to get out of PowerPoint and into Alteryx. Jim mentioned the doctor list matching. I got a dummy doctor list in this input and a little extract of our CRM in the right output. If you've ever done merge mode in fuzzy matching before, you know like add a label with the column called source for example, where the records came from and then union them together so same thing I'd normally do for fuzzy matching. Give it a Record ID, fuzzy matching uses Record IDs to say Record ID 1 and 2 matched or 1 and 5 matched, so that's why we need that tool.

On the right here, I have seven fuzzy match tools. If I look at the... I label each of them, I union them together, I pick the best one, and if I look at my final results, I get 14 matches. I'm going to join the data back on and see that I get my results and going to see the data compares side by side. That entire tool container, this big tool container is completely replicated in this tool container, which makes use of the Fuzzy Match Master.

I'll pull up the configuration here and just walk you through it. It requires a Record ID field called Record ID. There's a checkbox that allows you to only see the combinations. It doesn't even have to do a fuzzy matching, it can just tell you for starters, these are the combinations that I'm going to make for you based on the fields that you enter. You enter the fields a little lower so I'll scroll down here. I got 15 spots in here for you to enter fields. You got to configure this one with first name using a match style names with nicknames, last name with a style of name, etc. then I can also set my overall match threshold just like a normal fuzzy match tool.

Now, we get a little more sophisticated, minimum number of fields in each match. I could have a combination that's just last name or just first name. That wouldn't really do any good in most cases so I'm going to make that there has to be at least two fields in each combination. Then you can select the fields that you want to make sure they're always there. In case of doctor names and doc matching them to our CRM system, if I don't have the doctor's last name, Yay! I matched these two addresses, Bob Smith and Betty Smith. No, not the same person. Then another option we have is to prevent fields that should not co-exist in a match. We do a lot of middle initial versus middle name matching. I don't want a combination that has both of those because that's like doing some of the same work twice. You configure all of this and the macro will tell you down in the C output the different combinations that it will run with those parameters.

If I change let's say the minimum number needed to three, numbers five, six, and seven there would go away. Let's say I don't like one of these combinations. Maybe I don't like last name and address because address may not mean that much without a state or a zip code for example so last address, I don't like that one. I'm going to take and I'm going to copy that output, C output, and I'm going to right-click on my canvass and paste. I go to text input and I'm going to just get rid of all of my combinations except for last address. Then I'm going to connect that to my C input on the Fuzzy Match Master and say exclude combinations in the C input. If I re-run this workflow now, we had seven combinations coming through before. Now, we only have six. You can tweak your combinations to make it more efficient of the process.

Then the output as I mentioned before is pretty much exactly the same as the normal fuzzy match tool with my two Record IDs. It also gives you the match score and then two new fields, rank and type. Rank is a way that you can sort and view your results, see how many matched you've got at the different levels, and the type is my label, how I describe which of those fuzzy matched, all of those seven different combinations on our right, which one of those it came from. It actually even labels. It the match core is a hundred percent, it'll say exact in the type for you so you know that, "Hey, this is better than a fuzzy with the same fields."

That's the Fuzzy Match Master in action. What we're going to do now is take you into the macro itself so that you can actually see the inside and then download this macro from the public gallery and begin deploying it within you company. It's already out there and available for download. We'll show you at the end how to find it. For those of you who don't know you can right-click on a macro and click open macro and it'll take you right inside of it. Here we have the Fuzzy Match Master with my super cool font, had fun making this thing.

It's a kind of a big macro it seems like. I broke it up into different boxes for you. It's actually not too complicated. All the fields that you enter are coming in these interface tools over here, the field to match style. Those with wireless connections just come into this text input right here. You can see the data starts in my workflow. It's going to do a little bit of cleanup to get the fields in order for the way that it needs it, and then we enter the screen box which generates all the possible combinations.

I made use if a macro that was available on the Alteryx community called the combinations macro. I did not make this. The source for that is right in this comment box so you know where you can find that to download as well. What it's going to do is if i have A, B, C, it's going to make ABC, AB, AC, BC, ABC, it's going to make all those combinations. We just do that with field names this time around. What comes out of that is going to be a list of all the fields and what combinations they were a part of. I can then join that back to my field names and I know what my combinations are.

Now, we enter this orange box. The orange box is where those different configurations like the minimum number of fields you need, which ones are required, which ones you don't want to co-exist, all of that is being excluded in this red box. Basic tools, filters, summarize, join, pretty simple. Now in this purple box is where the magic and the more technical stuff of the Fuzzy Match Master has. How many of you have worked at all with XML within Alteryx? Fewer hands as expected. XML is the language that basically every Alteryx software you have is just a text file of XML and the Alteryx engine knows how to interpret that to run the data and do the magic that Alteryx does.

What we're actually doing is built into this text input is the XML of the fuzzy match tool, most of it anyway. We copied this XML and these are the default fuzzy match styles, name, nicknames, address, etc. we just call them exactly that. We actually tweaked names to nicknames because the way that it's setup as a default actually wouldn't work the way you expect it so we took out or tweaked that a little to make it better for our purposes. Then we rank it to make it more efficient. I won't explain that here but the higher the rank like a one is given to exact, that means if the field... if I don't match on that field then I don't have to worry about processing all the other ones because that field didn't match so I'm done.

What the macro will do now is take that XML, join it to the fields and the combinations and have XML for every single combination that you're running through the tool. It finishes cleaning that up concatenate it here. This next orange box what it does is it takes that C input and includes and excludes are intern, [Skylar Hoss 00:23:35] also known as the lowly intern on the Alteryx community. He set this up so that you don't... it doesn't matter what order your fields are/ I could have typed zip, first or first, zip, it would have excluded it either way so you don't have to worry about that.

The C input is taken care of there, order doesn't matter, and finally, I can sort to make sure I'm doing the combinations most fields first, fewer fields last, and feed that into the batch macro dynamic fuzzy match, which is nothing more than a fuzzy match tool, some messages, and a couple of things to prevent duplicates and make it run faster. That XML from that text input made its way all the way into this control parameter and is literally replacing part of the XML of this fuzzy match tool, from the style tag to the n style tag, and then, for all the different fields that you have, puts those all in there. It's basically making a new fuzzy match tool every single time it runs a combination it loops through. This one little workflow is doing the vast majority of the work.

After that it just cleans up your results, makes them look pretty, and only gives you the best one. If a record appears in multiple combinations you're only going to get it once. Jim mentioned that you can customize this. It's a macro so you can tear this apart, you can tweak, you can add new features. We have ideas for things that we want to do with it, feel free, go for it. One way that we like to customize it is by adding new match styles. Let's say I'm back in my normal fuzzy match tool. I come up with a cool new configuration for zip code, let's say. Actually zip code is pretty tight, let's do last name. Let's say all of this, the configuration in the fuzzy match tool, I recommend you checking out the Alteryx community training videos, they explain this a little bit better in terms of all the different options that you have. Let's say I want a different key generation and I want to use both the Jaro and Levinstein algorithms. This is my new type, match style.

I can grab the XML for that new custom thing which you can see right here. If you copy and paste this XML into that text input and the Fuzzy Match Master and then add that new title whatever you want to call it to your drop downs and those interface tools, boom, you have another option in your Fuzzy Match Master and you can have as many options as you want. It's allowing you to do more fuzzy matching than you could with the normal tool.

You're probably thinking, "Okay, this seems pretty cool, pretty comprehensive but does it really work? Does it really make a difference and solve the fuzzy matching problems that we wanted it to solve?" The answer to that question is absolutely it does. Before, you were like a snail. You needed multiple fuzzy match tools and you have to configure all of them, and you were making mistakes along the way, we were making mistakes along the way. Now we're a cheetah, one tool, drag and drop, one tool, and it's a customizable tool too. Before, we have to maintain this monster workflow and then Alteryx updated and we couldn't maintain it anymore, now, it's one tool. You can actually, that C input, you can make that an Excel spreadsheet and then you can change your combinations that you want to include and exclude without having to re-publish your work, change your workflows, or re-publish your apps that you use this in.

That is why the Fuzzy Match Master is the star for self service because we've actually been able to use this. I took a brand new data system that we had and I spent a day getting all the data organized but spent five minutes dropping the fuzzy matching in and lo and behold I had a comprehensive fuzzy matching process in only a couple of minutes, view the results and we were able to deduplify that system from there. To guess some numbers, we had close to 50,000 duplicates organization like hospital facility duplicates identified in the data set that I just mentioned and the Fuzzy Match Master found those for us.

Then, for the doctor list, with that scrubber app that we've produced in just nine months after a hundred and twenty different runs of this app, we've added or changed close to 35,000 pieces of information in our CRM system. It's productionized and working on our server environment and has allowed us to go from this to this, much, much cleaner, much easier to maintain. We have one Fuzzy Match Master for individuals, one for groups, and an input tool there so we can change the combinations without having to re-publish it. This app is all on our public gallery. Since we were the best use case of excellence award for our server last year, we'll take you out and show you our server.

This is the MedPro gallery, looks a little different than the public gallery. We did some code tweaks to make it work for us. If I scroll down here, we have our Scrubber 2.0. we setup the interface so that anyone in our company can add a file spreadsheet from marketing or agent campaign that they're doing. Thirteen different fields that we've seen in our list. They can pick and choose which ones they've had and then match against our CRM system. Maybe they want to do group zone, maybe they want to do exact only, they have a lot of different options. Then they can even scrub against another Excel file or another text file or Alteryx file. They can do two files or a file against our CRM.

This app allows any non-Alteryx user to take advantage of the Fuzzy Match Master. Before, we only were running 48 combinations and they were labeled. Some of them are labeled incorrectly, now we're running a hundred combinations and getting more comprehensive results in our data. Twenty five percent of our employees are using or gallery with another decent percentage using it through APIs. Like I said, this app's been used over 120 times just in the last three quarters.

By this point, I hope you guys, a, have learned a little bit about fuzzy matching, and if you're new to fuzzy matching, there's a lot of great resources out on the Alteryx community to help you get up to speed, to understand why we did some of the things we did in the Fuzzy Match Master. I hope you're also excited to actually take the Fuzzy Match Master home with you. We have shared it out in the Alteryx gallery for you and all you have to do is search "Fuzzy Match Master" with the quotes, click on the icon, click Download, and save the file. For those of you who have never been on to the gallery before, we'll just show you real quick what that looks like. Here's the public gallery, macro's in the home page because I just uploaded it this morning after adding some new documentation. If you click up in the search bar here "fuzzy match master", enter, there it is. Details about it, click Download.

Remember there's macros inside of it so it's actually going to come as an Alteryx zip file. You can just double-click on that, open it in Alteryx and it'll unpackage those workflows and work automatically for you right there. All right. That is the Fuzzy Match Master. At this time, we are done with out presentation on self-service fuzzy matching taking fuzzy matching from Alteryx leveling it up with a batch process and implementing it on our gallery as well as in the designer. We encourage you to fill out this session survey. Here are the instructions. It's all in the mobile app, just pick this session and scroll down a little bit to the survey spot.

At this time, Jim and I will take any questions that you guys have. It's 1:34 so we got 10 minutes before the next group. We have someone with the microphone there in the back so you just raise your hand and you get question answered.

Crowd question:
Hi, my name is Amar. My question is related to your mass data management using this fuzzy matching. How frequently you'll get the data from the sources? How frequently you update you CRM system?

Dalton Moss:
Good question. The CRM system is one of many systems that we update. The CRM actually gets a lot of data feeds. Our data is updated quarterly but we can run the app that we showed you whenever we have new information come in at any point of the cycle that Jim mentioned. If we get a list tomorrow we can run this and then implement it into our process. We'll run a whole system dedup about once a quarter using fuzzy matching in Alteryx and then update our CRM quarterly from that. The only reason why we do it quarterly is just the time it takes to look at the fuzzy matching results and see which ones we actually want to implement and change. We could do it as often as we wanted to with the power of Alteryx.

Crowd question:
Thank you, one more question. Do you do the electronic death match using this?

Dalton Moss:
Sorry?

Crowd question:
Do you do the death match? Suppose you have an individual and there can be claims that are coming from the doctor, you have to do a list of your employees or your enrollees within the group, you want to do the electronic death match before even you process their claim or something. Do you have anything like that in this?

Dalton Moss:
I'm sorry, I didn't understand your question.

Crowd question:
Sorry. Do you do any electronic death match processes in your whole [crosstalk 00:34:38]?

Dalton Moss:
Electronic?

Crowd question:
Death match.

Dalton Moss:
Death? Death Match?

Crowd question:
Yeah.

Dalton Moss:
We do have a record or system that keeps track of retired, disabled, and passed away doctors that links into our CRM directly and labels the record so that we can match them that way.

Crowd question:
Okay, thank you.

Crowd question:
I'm curious to know how you handle doctors, physicians, or providers that have multiple addresses or whose addresses change over time. How do you recognize that they're the same individual?

Dalton Moss:
Sure. In our system, we have one row per address. If we match on one line that account number is on multiple lines and we're able to identify the doctor through that manner. Let's say a doctor moved from one county to another county within a state. Those weaker combinations are going to see that doctor's first name and last name and state match. The address doesn't really match anymore but because of those weaker combinations we say, "Oh! The state still matches, or the phone number still matches, or the email still matches, or one of those other fields that are going to give us a better idea that, "Oh! That's who the doctor is", and then we can say, "Wow! The address isn't the same anymore", and merge those two records using the new data as our new supplemental records to feed into the system.

We just hope that those weaker combinations have enough in common between the new state of the doctor and the old state of the doctor so that we can then update the record accordingly with just using the stronger combinations we wouldn't have been able to identify because the information was changed.

Any other questions about fuzzy matching or the Fuzzy Match Master?

Crowd question:
What changes did you make to gallery and why?

Dalton Moss:
Sure. Jim mentioned that Patrick Diagon is an Alteryx ace. He led us tweaking the look of our gallery a bit. One of the changes that we made is to put this more square approach rather than the rectangular approach and get rid of the author name and the type of thing that it is because we only have apps on ours not macros. We didn't really need this. We didn't really care who the author was because you can see it by clicking into it. To get more apps on one page, that is we motivated to this approach. It's a little bit more consolidated.

Aside from that aesthetic look, we made a couple of changes to make it easier for the admins to have control. I as an admin on our gallery can edit, I can download other people's workflows even if they didn't check the setting, "Hey, I'm going to allow people to download this workflow." Admins can now do that without that setting be in check. There's a couple of things to make a more admin-friendly small tweaks to the Java and CSS files that deed this. That's all, really just for aesthetic purposes.

Crowd question:
Thank you.

Crowd question:
[inaudible 00:38:10] was deduping your CRM systems. Can you share some other [inaudible 00:38:18]?

Dalton Moss:
Sure. For fuzzy matching in general or this macro?

Crowd question:
I don't know [inaudible 00:38:23].

Dalton Moss:
The question was to, what are some of the used cases for fuzzy matching or the Fuzzy Match Master aside from just deduping our CRM system? Insurance companies have a lot of different systems. Claim system, policy holder management system, CRM systems. We're only going to be deduping a lot of those systems. The list matching taking external data that comes in any point of that cycle, engagement, service, marketing sales, taking external information and trying to integrate it into our existing systems is the other broad use case for fuzzy matching.

We've also used it to test new data vendors. For example, we want to buy this new doctor data. How good is that data stack up against another data vendor that we're investing or our current data that we're already purchasing. We can fuzzy match to our system or to the other vendor's data and compare them and make a better decision on what data vendor that we should go with going forward. That's the third main way that we've used fuzzy matching. Got a few more minutes for anymore questions or comments.

Crowd question:
[inaudible 00:39:46]

Dalton Moss:
Twenty five percent of our employees are using the gallery apps and have never seen the Alteryx designer before. We make it very easy for them. We do a lot of work to think about our interface tools that we want to use. For example, this report is as simple as just entering a file of MPI numbers and a couple of drop-downs. Then we train everyone with a video. We actually do the screen capture video to say, "Hey, here's this new Alteryx app. This is how you fill it out. Here's a couple of examples of ways it would use it. Now they even need to know all the data processing, fuzzy matching algorithms, [inaudible 00:40:44], address, they don't need to know all of that. They just need to know why is this out here, what can I use it for, and how do I fill it out and any other pertinent information? We've really tried to take the technical piece out of it for our gallery users and focus on technical discussions with our Alteryx designers that are actually building some of these apps and macros.

Jim Kunce:
The only part I would add to that would be that we've tried to disseminate the use of the designer across the functions as well. We have people in IT, data warehouse, finance. Claims, I don't know if they're using it yet or not.

Dalton Moss:
Not yet.

Jim Kunce:
We've got those folks. They're seeing the power of Alteryx and they're designing the apps for their teams to use. It's not just coming up from one function to the other. It's actually growing up from the inside.

Dalton Moss:
That helps give the context. Someone in finance is better able to explain the data that they're using and what they're doing with it better than we could even though we may be better at Alteryx itself. Find the right users in your team and use the interface tools to their full potential.

Crowd question:
This is not a question. I just want to appreciate the amount of work that you guys have done. It's an amazing job. Kudos to you.

Dalton Moss:
Thank you very much.

Jim Kunce:
Give it up for Dalton. He has done a great job.

Dalton Moss:
Thank you. Go ahead and download the Fuzzy Match Master from the gallery. There's plenty of documentation in it to hopefully get you started, customizing it, using it for yourself. We hope you enjoy the presentation. Don't forget about the surveys. Enjoy the rest of the conference. Thanks for coming.

^Top

TESTEZ LA
PUISSANCE
D’ALTERYX.

Pour Démarrer