Unfortunately, there is no best Customer Data Plaform (CDP), just like there’s no best shoe.
You don’t buy “the best shoe.” You buy the best shoe for your foot, for a particular need: running, dancing, playing basketball, or looking good on a date. You’re going to need a different shoe for each situation. And you’re going to need to find a CDP that meets your business requirements.
The trick is to know what questions to ask yourself as you start this process, and that’s the point of this article. My goal is to prepare you to evaluate your own needs, which will equip you to speak intelligently to potential CDP vendors.
Forget your website for a minute
The Internet has changed the way we think about things to such an extent that we often assume our website is the starting point for all tech questions. That’s not always true.
For some businesses — like a restaurant, or a gym — web traffic might not be the first consideration, but they might still need to harmonize customer data from multiple sources.
That’s the main use case of a CDP – to collect data from several different sources and merge it into a “golden record” for each customer so that you can take appropriate action.
As you think about a CDP, ask yourself what customer data you have, where it’s stored, and what would you do with that data if you had it organized the way you wanted it. Possible sources might be …
- Prospect lists or sales leads
- Email data (lists, opens, clicks, unsubscribes)
- Purchase information
- Questions / customer support
- Membership / rewards programs
And potentially a lot more, depending on your business.
Make a list. Note where the data is stored, how often it’s updated, what you currently do with it, and what you hope to be able to do with it.
Now think about your website
Now that you’ve thought about all your other sources of customer data, think about your customer data again in the context of your web traffic. Imagine that you could sort your web visitors into four buckets.
- Customers who have self-identified (e.g., by a login)
- Anonymous visitors who are customers that haven’t yet self-identified
- Anonymous visitors who might become customers
- Anonymous visitors who will never become customers
Eventually you’ll come to think of your web traffic in many more categories, but this is a good start. Try to estimate how much of your total traffic might be attributable to each group. (You’ll probably be wrong.) Then, develop a plan for each of those groups. That plan will influence what sort of CDP you’re interested in.
For example, if the vast majority of your visitors are in group #4, you might not even need a CDP. You might be better off with a Data Management Platform (DMP). But if most of your visitors are in group #1, you probably want a CDP that specializes in customer-centric use cases.
Probabilistic vs. deterministic matching
It’s possible to make inferences about a web visitor based on behavior on your site, and even on other sites. But these inferences aren’t always correct. Is that a problem? Maybe. It depends on what you intend to do with that inference.
For example, if your data leads you to believe that a customer likes cats, and he doesn’t, what’s the downside? You might have displayed an ad that was irrelevant to his needs. Is that a big problem or a small problem?
On the other hand, if you have different rates for different customers (or types of customers), it might be a colossal mistake if you give Joe Suzanne’s renewal rate.
In some cases, you want a deterministic match. For example, let’s say Harry Pratt is one of your customers, and his subscription is up for renewal. Harry clicks on one of your emails and goes to your website, but he hasn’t logged in. You could guess that he’s probably Harry from the click on the email, but he might have forwarded that email to somebody else, so you’re not sure it’s Harry. Before you show Harry’s subscription information, you want a deterministic match – like a login – so you can merge the subscription record for Harry with the website session for Harry.
(Realize that a “deterministic” match is never quite as deterministic as you’d like it to be. Harry might share his login credentials with his mother.)
In other cases, you’re not all that worried about precision. That’s okay. It all depends on your use case.
Think through what you intend to do with your data and decide when you need to be deterministic, and when probabilistic is good enough. You’re going to need to make that distinction when you speak with prospective CDP vendors and evaluate the tools they use for each scenario.
Data cleansing
One of the basic use cases of a CDP is to merge data from various sources. But we all know that data can be messy. A customer might submit his home address in a format the post office doesn’t recognize. Someone might misspell yahoo, or gmail. Or even their own name.
How are you going to merge these records?
Is Gregory Krehbiel the same person as Greg Krehbiel? How do you know? Does it matter to your business strategy? That is, what’s the downside of having two records — one for Gregory and one for Greg — vs. the downside of mistakenly merging them?
When merging records, does the system know that gkrehbiel@gmail is the same as gkrehbiel@gmail.com?
When your data has a postal address that doesn’t match USPS requirements, do you need to correct it? Where is that correction made? Some CDPs offer such services while others do not.
Sometimes data cleanliness matters, and sometimes it doesn’t. You’re going to have to think through the various scenarios, identify when clean data does make a difference, and ask potential CDPs how they’d address those situations. You might need to use an outside vendor to clean up your data.
Advertising
If you sell advertising on your website, your ability to create segments of people with discreet interests can be very attractive to your advertisers. Ask your advertisers what they might like, and then do some creative thinking on your own. What sorts of segments might an advertiser want to target? What data do you have to collect to create those segments?
If you use advertising to drive traffic to your site, it might be helpful to separate users by which ads worked for them.
Think through these things and write out your use cases, then find out how a potential CDP might address those needs.
Paywalls
Paywalls are becoming very popular on content-rich sites. Some CDPs can manage a paywall directly, while others make a connection with the service that manages the paywall and transfer data back and forth.
This raises the “jack of all trades vs. best in breed” question — or what I sometimes call the Swiss Army knife issue.
A Swiss Army knife is very convenient because it has lots of useful tools that are there when you need them. But none of those tools are as good as a dedicated tool. The blade isn’t the best blade, and the screw driver isn’t the best screw driver.
Some CDPs have a lot of built-in functions — or they have tools that allow you to build that function — but they’re not likely to be quite as good as the service that’s dedicated to solving that one thing.
This is not an easy problem to solve. It affects your budget, the drain on your tech resources to have to manage multiple different services, internal training issues, the different contract lengths for each service, the fact that “integrations” are never as nice as advertised ….
My advice is to start with your use cases. Decide what you need to accomplish, and try to use as few services as possible to get the job done. Good enough is often good enough.
Leakage — how many people escape the CDP?
If you’re going to rely on your CDP for a paywall (or something similar), you need to know how many people are falling through the cracks.
CDPs usually rely on a snippet of JavaScript that sends information about a web visitor’s session to the CDP’s cloud-based database. That’s quick and efficient, but sometimes it doesn’t work. Incognito browsers, ad blockers and other privacy tech sometimes interferes with the ability of the CDP to profile every user. Depending on the scope of that leakage, it could cost you a lot of money.
Talk to CDP vendors about this problem, and come up with a way to audit the CDP’s ability to profile your users. If possible, use a server-based solution to see how many people are getting past the CDP. Another solution is to compare the CDP’s data with what you see in analytics.
But keep an eye on it. Don’t assume the CDP is profiling every visitor, and realize this is a problem that’s only going to get worse as people become more concerned about privacy.
Connections / integrations
In order to gather customer data from multiple sources and merge it into a single customer record, the CDP has to have several different ways to collect that data. How that data is collected can affect your ability to use it the way you want.
Take your list of customer data sources and see how many of them a potential CDP vendor already integrates with. They’ll tell you they can integrate with anything, but be very cautious relying on that sort of promise.
Here are a few things to consider.
- Real-time vs. batched. Real-time sounds better, but it’s not always the right choice. A real-time connection uses up resources that you might need for other things. Also, some data – such as subscriber information from a fulfillment system – may not be complete until reconciliation processes are run at the end of the day. In that case, a real-time connection isn’t helping you.
- Racing issues. If you try to rely on a connection that has to run before you render a page, you might run into issues where the processes are stepping on each other’s feet. This can slow down the user experience, or cause errors. If a use case requires real-time connections, have a long talk about exactly how that will work.
- Do they have an integration team? The existence of such a team shows a commitment to managing existing integrations and creating new ones.
- What’s the process for getting a new integration? How long does it take? How much input will you have in the process? Is there a cost?
- Do they monitor the health of your connections? What happens when a connection fails? Who is notified? How is the problem resolved?
Snapshot or historical data?
A CDP takes data from multiple sources and merges that data into a single view of the customer, which allows you to do a lot of cool things right now. But what if you want to know what happened last week, or last year?
For example, let’s say you have a magazine website where people can subscribe month to month. The short-timers are losing you money, so you want to compare the behavior of your long-timers with the behavior of your short-timers, to see how that might affect your offers, your content, engagement efforts, etc.
If the CDP only provides a “right now” snapshot of your users, you’re not going to be able to do that sort of analysis.
Reports and Analysis
Speaking of which, you want to make a list of the kinds of reports and analysis you’re going to want from the CDP, and make sure they can provide it.
There’s going to be some overlap between what you want to see from the CDP, and what you see from your analytics package. Find out how well they work together, and use each of them to keep the other honest.
The reports you will need will depend on your specific business, but adding a CDP to your tech stack will open up interesting opportunities. Show the CDP vendor your current Key Performance Indicators and ask how they can enhance that information, or give you new insights to help acquire or retain customers.
As a simple example, let’s go back to the magazine publisher. They have reports from their fulfillment system on how many customers they gain and lose each month, how long people stay subscribers, etc., but the fulfillment system can’t tell you that Firefox users are more likely to lapse than Chrome users, or that people who visit on their mobile device are less likely to buy ancillary products. Those sorts of insights from a CDP can help you patch up holes and create a better user experience.
Customer Journeys
Some CDPs have very slick systems for creating customer journeys, where, for example, a user qualifies for a particular journey after some triggering event, and then the system presents the next offer at the appropriate time. These programs can be very useful for on-boarding, retention, upselling, etc.
Origin story
CDPs are a big thing right now, and a lot of companies want to compete in that space. Some CDP vendors started off with a technology that did some CDP (or CDP-like) functions, and then they bolted on a few additional services to fill in the gaps, or purchased some other technology.
It’s good to know that back story. If the CDP was originally an email service provider, or a data warehouse, or an ad system, that’s still in their DNA, and it’s going to affect how they approach other issues.
Conclusion
It’s impossible to say what’s the best CDP, but with careful thinking and planning, you can find out what’s the best CDP for you. I hope this article was informative, and if I can help you evaluate your technology needs, please contact me, or take a look at this page: How I can help you with marketing and publishing technology and operations