Sunday, October 27, 2019

What Happens When Everyone Has a CDP?


October 22 was a landmark day for the CDP industry.  We reported three significant announcements:


Each announcement follows a string of similar previous developments.

SessionM/Mastercard: most CDP acquisitions have been made by companies that are not primarily marketing software vendors. Arm (microprocessor technology) bought Treasure Data; Dun & Bradstreet (B2B data sales) bought Lattice Engines; Kabbage (small business finance) bought Radius; Anaplan (business planning) bought Mintigo; Informatica (data management) bought Allsight; Equifax (credit bureau) bought Datalicious. The exception that proves the rule is Salesforce’s purchase of Datorama, which it uses for marketing performance measurement, not as a CDP.

I believe the reason for these deals is that the buyers want to offer services that depend on unified customer data, but find it’s easier and cheaper to buy the necessary technology than to develop it internally. Note that it's truly a build/buy choice: many of the buyers already have extensive customer data management operations, so they probably could have built the systems for themselves.  They simply realized that buying was the better option.

The implications of this are substantial.  Competitors of the acquiring firms will feel pressure to offer similar services that help their clients deploy customer data.  Such services can be important tools for retaining clients, since switching customer database providers is painful at best. For CDP vendors, these deals are a promising exit path from a crowded industry which will only become more competeitive (see below).  For marketers, these deals mean their companies gain new options for access to a CDP.  This is especially true for small and mid-size businesses that might lack the resources to buy and integrate a CDP on their own.

Teradata: major marketing software vendors have chosen to build their own CDPs rather than buying one. This list includes Salesforce, Adobe, Oracle, Microsoft, IBM, SAP (sort of) and SAS (although they don’t use the term).  Their decisions to build their own CDPs are a bit perplexing, given that most have made many other acquisitions to fill gaps in their product lines.  My best guess is they like to buy companies that give them a substantial position in a new market, and even the largest of the pure-play CDP vendors are too small to catch their fancy. It might also be that building a CDP looked simple to them, so they all decided there’s not much reason to purchase a CDP for technology alone. The time it has taken them to deliver proper CDPs suggests it may have been harder than they thought.

The marketing software vendors’ delay in delivering CDPs has given other vendors opportunities that might not have existed had the marketing software companies moved more quickly. But with the marketing software vendors products now finally reaching the market, that era is ending. This will make life more difficult for the independent CDP vendors.  I still expect many of the independents will survive by developing systems tailored to particular industries, regions, or client sizes.

Wunderman Thompson: many ad agencies have decided to partner with a CDP vendor rather than purchasing one outright. The analysis gets a little confusing here because the big ad holding companies have been purchasing data-based marketing agencies: Dentsu bought Merkle, IPG bought Acxiom, Publicis bought Epsilon. But those agencies have themselves generally resold marketing technology rather than building or buying their own. This is probably a good choice: although they have considerable skill working with customer data, they have limited software development capacity. So it makes more sense for them to rent technology from others.

Assuming they continue to work with other vendors' technologies, the agencies represent a market for CDP vendors that won’t go away. If anything, it’s likely to grow as more agencies offer customer data-based services. But agencies have special needs and are often very cost-sensitive. So only a handful of CDP vendors are likely to get much benefit.

These three lines of development all point in the same direction. The path leads to a world where unified, sharable customer data is available to nearly every organization: that is, a world where every company has a CDP.  Nirvana, you say?  Yes, possibly, for CDP users.  But remember that CDP might be a stand-alone system, part of a marketing software suite, embedded in operational systems, or provided as part of an agency’s service. So it's not necessarily great news for CDP sellers.  

The broad availability of CDP functions affects users in other ways. When CDP functionality was available only from specialist vendors, the choice of a CDP was based on finding the best system (or, more precisely, the system that best fit each buyer’s particular requirements). But when CDPs are baked into larger software and service offerings, the quality of the embedded CDP is one of many considerations in selecting a vendor. In fact, the CDP itself may be invisible, as buyers base their choice on which vendor can best meet their business needs. If the potential vendor can meet those needs, its CDP must be adequate. If the potential vendor isn’t a good fit, it really doesn’t matter whether the fault lies with their CDP or some other component.

Note that there will be exceptions to this new rule. Large enterprises are likely to assemble their own collection of best-of-breed components, including a stand-alone CDP to integrate data from disparate sources. Mid-size firms who don't want to commit to one comprehensive marketing suite may prefer broad-scope CDPs that combine centralized data, analytic and personalization functions while integrating with external delivery systems.

What doesn’t change are the needs for users to define their requirements, to accurately assess which vendors will meet them, and to deploy their choice effectively. These tasks are closely related: you can’t define requirements, assess alternatives, or deploy effectively without understanding what the CDP needs to do. So education and training of CDP users will remain important regardless of how CDPs are purchased or delivered.

Another way to look at it is this: CDP has been cruising through the Gartner Hype Cycle for the past four years. Each hype cycle stage implies a common customer question*.  Here's what we’ve seen for CDP:

  • at the start, when the technology is unfamiliar, the question was: What’s a CDP?
  • during the peak of inflated expectations, people had some understanding of the concept, so their question became: Do I need a CDP?
  • as CDP enters the final stages of disillusionment, enlightenment, and productivity, people accept that they probably need a CDP and ask the next logical question: How do I best use my CDP?
Of course, different markets and individual users are at different stages in their own CDP journey, so we still get all three questions. But it’s clear that the third question – often phrased in requests for use cases, best practices, and maturity models – is becoming the most important. I’ve no doubt that the industry will provide more answers.

____________________________________________________________________________
*This is my interpretation, not Gartner’s.

Monday, October 14, 2019

Privacy Regulations Will Lead to Advertising Innovation

Naysayers doubted that the European Union’s General Data Protection Regulation (GDPR) and ePrivacy rules, California Consumer Privacy Act (CCPA) and related privacy regulations would have any real impact on the flow of consumer data. But it’s now clear that they will, as European regulators show they will interpret the rules to require meaningful consent to data sharing, will treat cookies as personal identifiers, and will assess significant fines for data breaches. The imminent effective date of CCPA, which rather surprisingly survived without being watered down or preempted by weaker federal regulation, confirms that privacy can not longer be ignored by data collectors. Meanwhile, actions by Google, Microsoft, Apple and other major browser are putting still more barriers in the way of data business as usual.

None of this marks the end of the advertising industry or even of targeted online advertising. Advertising was a big and important business before the Internet existed and would remain one if the Internet went dark tomorrow. Pre-Internet advertising wasn’t as targeted or measurable as today’s display and social ads, but it did work. A total end to distribution of third party personal data would still leave Internet marketers able to target based on context, content, and in-session behaviors. This probably wouldn’t be quite as effective as individual-level targeting, although I don’t recall any studies that prove this.  It’s even possible that closer attention to ad targeting methods would reduce fraud enough to more than compensate for the loss of third party data.

In any case, the hunt is on for mass advertising audiences to replace audiences based on third party personal data. The recent merger of content-based advertising leaders Taboola and Outbrain was driven in part by the desire for greater scale in content ads (those click-bait teasers that lead you off-site at the bottom of many Web pages). The rebirth of out-of-home advertising (billboards, wall posters, in-store displays, and clever niche products such as elevator and gas pump ads) as a digital channel opens another mass medium. Above all, the various forms of individually targeted TV advertising promise new levels of personalization and tracking in a medium that already has near-universal reach. Other types of video, available through YouTube, Instagram, TikTok, Snapchat, Twitch, and much else, are delivering new mass.

Taken together, these emerging options promise a new Golden Age of Advertising Creativity, as marketers explore their potential and evolve the most effective approaches to each. It’s true that a vastly more fragmented media landscape will be harder for marketers to manage. But it also means that fears of a Google/Facebook duopoly controlling all access to new customers are clearly overblown. Regulatory pressures, changing consumer attitudes, and new competition should further deflate them every day.

What does all this mean for martech in general and Customer Data Platforms in particular? The most directly affected martech systems are Data Management Platforms (DMPs), whose core function is to manage the third party data-based customer profiles that are quickly becoming extinct. Marketers who had hoped the DMP would manage all their customer data had already lost much of their interest when they discovered how limited DMP capabilities really were.  Many DMPs have repositioned themselves as CDPs, although not all have the technical capabilities required to meet the RealCDP requirements.*

The implications for CDPs are more positive, although not quite as rosy as sometimes suggested. There’s a common argument, which I frequently make myself, that the loss of third party data makes first party data more important, and that’s good for CDPs because they primarily manage first party data. This is true on some level but shouldn’t be overstated. The third party data that populates DMPs is mostly used for acquisition marketing, while the first party data in CDPs is mostly for marketing to current or former customers. So switching focus from DMP to CDP would leave a significant gap in many acquisition marketing programs.

As we’ve just seen, marketers will plenty of other ways to reach new audiences even after today’s third party data-driven advertising is gone. But there is one way that CDPs can directly support acquisition programs: by making it easier for companies to share first party data with each other, creating what is usually called second party data.

Second party data can refer to any first party data that is directly shared with another company (the second party). The CDP supports this by creating an extract file of first party data that can be sent to the second party. If that’s all that happens, it’s no different from any other extract.

But often second party data is created by comparing the customer lists of both parties and finding shared customers. The trick is often that the two parties don’t want to expose information about customers they don’t share.

One way to achieve this is to send a copy of both customer lists to another company that both firms trust. That company compares the two lists, finds matches, and returns whatever information the parties have agreed to share. Alternatively, a common identifier such as email address might be run through a standard “hashing” algorithm that will yield a unique result for each input but not reveal the actual identity information. Both parties use the same algorithm so that all records with the same identifier yield the same hash code and are identified as a match. Records that don’t match will generate different hash codes which are meaningless to the other party. With this sort of processing, there’s no need for another trusted company to do the matching because customer identifiers are not actually shared. Each party keeps a record of the original identifier and the resulting hash code, so it can tie the codes back to its actual customer records.

Here’s a practical example. Suppose a hotel chain and airline agree to identify loyalty program members on each others' lists.  That is, the airline would learn which of its customers were members of the hotel loyalty program and the hotel would learn which of its customers were members of the airline program. They can do this with the type of matching just described, adding a "partner loyalty member" flag to appropriate customer profiles.  The airline could then make a special offer to hotel loyalty members and the hotel could make a special offer to  airline program members. Each party could also send acquisition promotions on behalf of its partner to its own customers who are not  members of the partner's loyalty program. So long as each party sends the promotions to its own members, the other party never learns anything about them.

The minimum role of the CDP in this sort of second party sharing is to generate a customer list with the required data. Some CDPs can also find direct matches (acting as the trusted third company) or generate and match the hashed identifiers. CDPs with these added functions are in a position to benefit as the loss of third party data makes second party data sharing more important to acquisition marketing programs.

In sum, privacy regulations won't kill targeted marketing.  They'll actually lead marketers to expand the methods they use, promoting a healthy diversity and weakening the much-disliked Google/Facebook duopoly.  Customer Data Platforms will benefit as first and second party data play larger roles in the marketing mix. 

_____________________________________________________________________________
*load all types of data, retain all original details, store the data indefinitely, build unified profiles, and expose the profiles to any external system.