Collecting emails isn’t enough

“Amazon can do it. Why can’t you?”

Do you wonder if your customers think that?

I don’t. I’m fairly sure they think it all the time.

We all live under the shadow of the technical excellence of Amazon. They’ve spoiled people with incredible service, and the rest of us struggle to catch up.

Part of the reason Amazon can do what they do is they have loads of data on their customers. In this email I’ll outline one strategy that can help you keep up.

First, you need to be a fanatic about collecting email addresses from web visitors. Second, you need a Customer Data Platform (CDP) — or something like it — to unite all the information you have about your customers.

A CDP is a technology tool that can help you advance your business in two key ways: first, by collecting all your customer data into a single view, and second, by orchestrating custom campaigns based on that data. Some CDPs also activate the campaigns with on-site messages, or even with their own email service, while others make connections to outside activation services.

The important thing for this discussion is that the CDP allows you to create a single customer view that enables you to take appropriate actions for all your customers based on their unique circumstances.

The burden of legacy systems

Amazon has the advantage of having built their infrastructure in the modern world, from the ground up. They’re not shackled by lots of old systems written in out-dated computer languages. By contrast, many other businesses have customer data in disconnected silos — a store, an email service provider (ESP), a CRM system, analytics packages, a fulfillment database, etc.

Your customers don’t know or care about that. When they give you their information, they aren’t thinking “I gave my address to an ecommerce platform,” they’re thinking “I gave my address to this company.” And they expect you to treat them differently based on the data they provide. They expect you to have the technology chops to make all your systems work together.

In this edition of The Krehbiel Report, I’ll focus on the basic steps you need to take to get all your customer data in one place. I can’t cover everything, so I hope to prime the pump and get you thinking about what data is available and how you might use it to make your site, your marketing campaigns and your content efforts more useful to your customers and prospects. If you’re inspired and want to talk about it further, please contact me.

The connective tissue

To create a single customer view, you’re going to load your CDP with data from various sources and then cleanse and merge that data into a single record. In order to do that, you need to have a way to match the records. For example, if the store has a record for Jonathan Smith, and your CRM has one for Jon Smith, are those records for the same person or not?

This raises the question of deterministic vs. probabilistic matching. Deterministic matching looks for an unambiguous connection between two records. Probabilistic matching looks for a likely connection. Your choice between the two will depend on the use case. If you’re sending a bill, you need to be certain it’s going to the right person. If you’re deciding who should see an article about kayaking, it doesn’t matter that much.

A CDP allows you to tie static, typed data — like a transactional record from your store — to behavioral data, like which pages a person viewed on your website. This allows you to ask questions like, “What web pages do people who purchased our Paleo Diet Weekly prefer?”

You can do this because the CDP creates a profile for every web visitor and tracks what pages they view, then adds that data to all the other customer information you have from your other systems.

But the CDP isn’t magic. Just because you loaded all your customer data into the CDP doesn’t allow the CDP to track customers’ behavior on the site. For example, if I go into Walmart and buy a fishing pole, then go to the Walmart website and browse tackle boxes, Walmart has no way to match those two profiles. If Walmart uses a CDP (and I’m pretty sure they do), they will have two profiles for me — one from my in-store behavior, and one for my online behavior. They won’t be able to merge those two profiles unless and until I provide some information on which to match them. Usually that will be an email address, but other things could be used, like name and address, or a hash of a credit card.

For most companies in most situations, your main tool for identifying an anonymous web visitor is getting them to enter their email address, which then allows you to match the online behavioral profile to the back-end data from other systems. Once all that data is in the CDP, you can start treating your web visitors like customers you know, rather than anonymous strangers.

Not all emails are the same

Amazon links a customer record to a single email address, which is incredibly convenient. You might not have that luxury.

For example, some of my clients offer free e-mail newsletters as well as paid services. The same customer might use one address for newsletters and another for billing. Or the customer might have a work and a personal email address, and the work address might change with a new job.

The upshot is that you can’t merge all your data on “email address.” Your goal is to unite the data around a person, and that person might have multiple, and changing, email addresses.

There are different ways to collect emails, some of which are better than others. An email a customer uses to log in to a paid service is probably more reliable than an email a customer provides to download a free report. There are also ways to infer email addresses. For example, if an ESP sends an email with a link that includes a query parameter that provides that email address’ unique ID in the ESP, when the recipient clicks on that link, the ID can be written to the CDP, then the CDP can look up the email address. It’s a nice trick, but what happens if the recipient forwards the email to a friend, or posts the link on social media? Then you have lots of profiles with the wrong email address.

I recommend thinking of email addresses in different classes. For example:

• User-entered email: This will be somewhat reliable, but not entirely. Lots of people enter Mickey@disney.com if they can get away with it.
• Account email: This might be used for a login, or for delivery of a paid or free product. It’s pretty reliable.
• Verified email: Your ESP has been able to deliver to this, so you know it’s valid.
• Inferred email: Along the lines of what I mentioned above — an email address harvested from an ID in a query parameter.
• Best email: With the help of some machine learning, you might be able to decide which, of a customer’s five email addresses, is the one most likely to get an open or a click.

Summing up

You need to be a fanatic about collecting email addresses, because that will most likely be your best tool for unifying your customer data — but only when you can collect those email addresses and match them to profiles that include other data. For that, you’ll need a CDP. You’ll also have to be a fanatic about keeping all the various types of emails straight, and ensuring you merge your customer data on the right emails.

Let’s talk

I hope you found this article helpful. If have any questions or great ideas along these lines, please let me know, or post a comment below. I’d be happy to discuss your unique situation and give my advice on how you can tame the customer data monster.

And if you know someone who could benefit from this article, please pass it along.