A Framework to Conquer Your Data Quality Blues

Welcome to another X Force Data Summit presentation. Today, I've got Andre van Kempen, who works, for sales engineers in Amsterdam. He's a consultant there. And he's gonna talk about a framework for data quality. And I think we all wanna hear how to make our data quality better in our Salesforce works. So here's Andre.

Thank you.

Thank you for letting me speak at the summit. It's quite exciting, my very first virtual summit.

I'll skip most of this because you did the introduction.

Just a little bit of background before we, you know, dive into it.

So I have a background in IT and for the last seven years, I've been involved in Salesforce as consultant.

And currently working in Amsterdam for a consulting partner.

And you might recognize the logo, you're dreaming. That's what I do in my free time. So I'm a user group leader and also organize community conferences.

A bit like this, but then in person.

So for today, there's four can be coming to four chunks.

So first of I have some background of what's this whole data framework and quality.

We'll introduce a data steward then, yeah, why data quality matters. So this will help you explain it within your company or team, why we're actually doing this.

And then the actual framework. So that's five steps or five main areas that you can go through go through all the steps and hurdles you might come across.

And then also some more practical tips and tricks of what you can do after this presentation.

So for the first part, these might be some questions you get from your users, your managers, or even your customers.

For instance, you know, do we actually need to back up the data in Salesforce?

Usually in IT, you know, that's not even a question. But since Salesforce is cloud software, that's actually often kind of forgotten.

Yeah, other questions are do we adhere to the privacy and data retention rules? If not, how do we fix it?

And one that you can get from, for instance, your managers or users, why do all my reports show garbage? Like, what's the deal there?

And they'll come to you as an administrator, architect, consultant or a data steward.

And this is our data steward, Alex. And so, you know, the role is, yeah, Alex is responsible for the enterprise data in the organization and helps with these four main areas.

So identify the focus areas for your data management programme, Evaluate, you know, your data needs that your business has.

Assess KPIs, you know, and then plan and execute accordingly.

And of course, to then monitor the impact of all the processes on your data.

So this is, of of course, we're talking about data quality management.

But this is to show where it actually fits in with all the other major building blocks of information strategy.

And there's enough content available online, you know, per building block to fill several days of webinars and training sessions. Today, we'll zoom in on the data quality management and a bit the master data management.

But yeah, I do encourage you to actually dive into each of these separately and don't just focus on one, you know, one of the blocks.

Then on the second part, why the data quality matters.

So what you see here on the left hand side, that's often how companies start. So you have your operations department, who has their own data and tools, the sales has their own data and tools and also manufacturing and so on. So it's really siloed per department.

As your company grows, or team, company, you go more towards a enterprise view your data.

And it's better to, you know, with larger enterprises, they're better served by master data management once a platform, instead of separate tools.

And this enterprise view will be sort of the gold standard for each department who is pretty much like a subscriber to the whole MDM master data management platform.

So yeah, you know, MDM provides a consistent context for all your consolidating for consolidating of your data.

Now, one of the things that could also impact your data quality is legacy systems.

Every company has them. Even if you're, you know, five weeks old, there's already some legacy stuff happening using the form of Excel sheets. But yeah, you could easily have like seven, ten, thirty different data sources, Which then of course causes the different business units, like you know, sales, they're not talking to marketing, operations is not able to anticipate on the demand because from the sales teams, they're not getting the up to date information.

And the most important thing is that has impact on your customers.

So yeah, should ask yourself, how can you provide your customers with a seamless and personalized experience that yeah, how can you do that if your back end is not in the back end, you don't have the right access to the information.

So this is not just an internal exercise.

Then, besides the tooling, there's also a few different processes that can impact your data quality.

I won't go through all of them, But for all this together, you can choose your class everything, how complete is it, how accurate is it, all the data we have, how relevant is it.

So you can score that.

And if you do that just across the board, entire data set, you get a score. And of course, first you start with a benchmark and you'll see the score will all change constantly.

Because internal processors are changing, tooling is changing, because these days, which tool lasts longer than five years in your organization.

So and yeah, a lot of these processes, they, you know, over time, yeah, they'll help decay the quality of your data.

And sometimes it could also be the lack of appropriate skills within teams or your company.

So coming back to the data quality score, once you have that, measure it, you can improve it and you can work on that.

Which you'll do. It's an iterative process. So you kind of just say, okay, today we have bad quality, after next week and we have good quality data.

It's something that just evaluates over time.

And, of course, it is good to to map out which which process processes impact your data the most at the moment.

And then the actual framework.

So this will help you, you know, with the iterative increases of your your data quality.

As you can see, this might remind some people of the PDCA cycle.

So plan, do, check and act. There's five steps here.

Says data profiling on the top to start with. So you profile or analyze your data so you know where you're starting.

Then you start cleansing it.

So you identify the different anomalies within, you know, the profile or analyse data.

Then you know, you standardise and normalise your data. And I'll dive into all these separate topics in the next few slides.

Once you've done that, of course, you can start matching emerging.

And it doesn't stop there. Just keep going, monitor it, keep your score in mind or, you know, keep an eye on it, and then repeat.

So to start with the data profiling.

Yeah, don't try to boil the ocean. That's pretty much main takeaway from this slide.

Now perfection is really, really difficult to achieve.

You just want to improve over time.

So what you start doing is you first, you start grouping a sample of your records and then define what's, you know, what was good and what's bad, and then you just start grouping them.

Then once you've done that, you focus on prevention tactics for the point of entry of your data. So what are all the different channels and tools where your data comes in?

And also keep that in mind.

Then So the cleansing, once you have your good data and your bad data, then start with, you know, focus on formatting issues, for instance. So you have your dates or postal codes or legal values.

Yes, start with those ones to do the first cleansing. They're usually the easiest. Dates, postal codes, address details.

This is the part where your data steward, as you can see here, Alex, plays quite a big role involving the business and actually identifying all these rules like what do we actually class as bad data and good data?

Make those rules, stick to them. Of course, in the next cycle, you can you can improve those.

Then standardizing a bit more. So for instance, the the legal name of your customers business partners.

Often a company has a name, internally is known as something else.

And maybe if you have a look, you know, Chamber of Commerce details, there's a different name. Make sure you standardize that. Other easy things, the website domain or the country.

And doesn't matter if the data quality is good or bad, make sure you standardize this and then split it again to good and bad.

Then matching and merging.

This is the fourth step. This is where you decide who should survive the battle of the duplicates.

This is a question you you'll come they will constantly pop up and you have two duplicate records. Yeah. Which one wins?

So here's a few few steps.

Yeah. Identify them, the duplicates, determine what should be consolidated. So so you have an account, business account, all these fields, what which parts are actually gonna going to be involved in in the the consolidation process.

And build survivorship rules.

So indeed, that's the decision what which one which one survives.

Make sure you can automate these rules as well.

Because if you have really super complex rules and then only one person understands, it's never gonna fly.

If you can explain it to someone who can automate it for instance in Salesforce, they'll really help. Don't have to explain it. Just let the system do the work for you.

And one thing that's often overlooked, dependent data.

Say you have a company or a contact person, a customer and there's email communication, there's opportunities, cases, everything. What's going to happen with that? Make sure they actually get re parented.

If you merge two companies and all the new cases, opportunities, emails should go on the new winner.

Now there's two of course, when start matching, you know, there's two types of matching, deterministic and probabilistic.

Deterministic is ideal.

It's a hundred percent exact match with everything.

That's the goal. But yeah, sometimes you just have to go search for duplicates that are not a hundred percent duplicate. So keep that also in mind. It's often it's more in my experience, yeah, less than a hundred percent probability that something actually is a duplicate.

Then the monitoring part.

So you took you create a profile of the data at the beginning, just to understand, you know, your current state of the quality.

Then took some actions to normalize, standardize match emerge to improve it.

But yeah, let's not stop there. So make sure you keep monitoring the quality.

Yeah, I won't go through all of these details, but the recording of this will be available after today.

One thing I do want to highlight is make sure your data steward is not alone in this.

Because Alex looks like a powerful person, but managing your entire, you know, enterprise and all the different related companies and stakeholders is just not realistic.

So start building something that's called tribal stewardship. So have people from all different departments and areas in your business, you know, the key business users, involve them in this process.

And if they're not involved in, you know, the first iteration, involve them in the next one or the next ones after that.

Now, these are the five steps, but this is additional step. Doesn't really fit in its kind of overall more an overall statement.

Backing up.

Like I mentioned, yeah, Salesforce doesn't actually back up your data for you in a way that you can easily contact them and say, hey, I deleted this case, please send me the backup.

So you're responsible for that yourself. Also, and not only for large organizations or old ones, make sure you have some type of archiving strategy as part of your information strategy.

Yeah, because there will also, you know, your business rules change, your quality rules change over time. A chunk of data that's maybe ten, twenty years old or sometimes even three years, that's already old, might not be relevant anymore. You don't need it, but you don't want to delete it, archive it somewhere.

Then the last part after you've heard all this, what are you actually going to do?

First step could be build a dashboard. There's actually out of the box dashboards available or some tools that can help you with this. But for instance, what this dashboard shows is per object, for instance, accounts or opportunities, it shows of all these fields that are mentioned here, like your name or your address details, how many times is it actually filled out.

And of course, will be perfect if everything's always filled out a hundred percent. But we all know that that's utopia.

Then once you've done that, there's there's a lot a lot of different tools out there.

Also, on on the AppExchange, they can help you with your with your data management.

Yeah. So there's there's a few few links in here to to help you with that.

Then some additional resources, for instance, on Trailhead, it's an online learning platform. You have a few modules specifically about this topic.

There's a success community.

You can get in touch with Salesforce experts and other customers and partners.

And of course, there's also the face to face or in person events like user groups and community conferences.

Then the last slide, because you've been listening to this yeah. Because you subscribed to this presentation, I'm allowed to to give away a free workshop as it's a dedicated two hour virtual workshop for where we help you create a solution for your most challenging Salesforce problem and that could be in any area, but for instance, you know, setting up your master data management strategy, building a business case or roadmap for Salesforce implementation, improvements, help with adoption or performing health check, all those kind of things.

The link will be sent to you after this session.

And there we also have my contact details.

Yeah. So they're also on the X Force website. Yeah. I'll be happy just to go in more in-depth on some items or areas to help you with get started with your information strategy or data quality management.

And with that, I'd like to say thank you very much.

Thank you, Andre.

I had a couple questions here.

And you answered most of them with one of your slides there. You had a really good slide on the different tools you would use to clean up the data. I know I'm familiar with demand tools as an example.

Do you find that a lot of your customers are You started out talking about the problem of having silos. Do you find that a lot of your customers are ending up using Salesforce as a single source of truth and therefore concentrating their MDM efforts in there? Or is Salesforce still you know, sort of you know, on the sidelines just a sales system?

It is Well, what I see it is indeed a trend.

Let's call it a trend that a lot of companies say, okay, we now have Salesforce, let's use this across all the business units, often as the first tool, so to say that does that.

And you know, often they'll then say, if it's on a Salesforce, it didn't happen or doesn't exist.

Do hear that quite a bit. Yes.

Yeah.

Excuse me.

The data quality dashboards, know that there's the built in Salesforce ones, but some of those tools in there have scoring, other scoring metrics are a little more sophisticated than just is the data filled out. Can do things like look for, know, the, for instance, are the abbreviations, you know, are the addresses, you know, standardized and so forth.

Is that one of the, I would imagine that'd be something that you would want your MDM team to do too, right? Run these tools and get a dashboard based on, you know, how much cleanup is needed in your org in addition to just, you know, missing data.

Yeah, indeed there's, but yeah, this dashboard I showed is really just a starting point. Like if you have absolutely nothing, it's a good start.

Just to get a feel of, yeah, one of the easy things to fix.

But also there's tools like you often have integrations with, I think it's called webservices dot com or like a chamber of commerce integration, where you can indeed do that if say you start entering company data or contact data in Salesforce. Once you click Save, it will then contact the database and do a check for you and maybe even make suggestions.

Like, if you have a Chamber of Commerce reference number, is this the company you mean? And then you have a whole list of more data.

Yeah, tools like that out there.

Cool.

Yeah.

I think thanks. It was a good framework, you know, basic overview of the framework for data quality management and I appreciate your time and appreciate you presenting at the conference.

Thank you very much. Thank you.

‍