Last weekend brought the intriguing rumor that Salesforce is in late stage talks to acquire data integration vendor Informatica. This followed the previous week’s rumor that Google-parent Alphabet is consideringan offer for Salesforce-competitor HubSpot, and came the same week as a slew of partnership announcements tied to Snowflake’s Marketing Data Cloud Forum.
You don’t need a crazy wall to see how these events are connected. Cloud database vendors including Google and Snowflake are expanding into marketing applications. This poses an obvious threat to customer cloud vendors like Salesforce, in good part because it expands the ability of IT and data teams to build their own solutions rather than buying them. Viewed in this light, Informatica helps Salesforce to compete by strengthening its ties with IT and data teams. Of course, this is the same benefit that Salesforce sought with its Tableau, Slack and Mulesoft acquisitions, although you could argue those were all primarily tools for business users while Informatica is used primarily within IT and data teams.
A stronger strategic argument is that an Informatica that is properly integrated with Salesforce Data Cloud would help those teams build more powerful databases within Salesforce. This would make Data Cloud a more viable alternative to Google Cloud or Snowflake as the company’s primary customer data store. In Salesforce’s wildest dreams, Data Cloud might even replace data warehouses.
Let’s put aside the question of whether Salesforce could or should become an enterprise data warehouse supplier. The immediate question is whether it can retain its position as a primary platform for customer data and customer-facing applications, or that role will be taken over by cloud platform vendors like Google Cloud, Amazon Web Services, and Microsoft Azure and cloud database specialists like Snowflake and Databricks.
I don’t know the answer. But what does seem clear is that customer systems will run on platforms, whether from Google, Salesforce, Snowflake, or someone else.
The platform approach isn’t particularly new but it’s not the way things have always been. Remember that most customer systems – even today – are stand-alone, self-contained products that maintain their own databases. These are the dreaded silos we keep hearing about. The trick is to connect these silos, initially at the data level by copying their data into a central warehouse or Customer Data Platform, and ultimately at the decision and delivery levels through shared journey orchestration and messaging.
Platforms solve the data problem, since all systems run on a common data store. They don’t necessarily unify the decision and delivery layers, since those functions are handled by applications built on the platform. Those boundaries are not clear: platform vendors sometimes add decision and delivery functions. Indeed, the risk that the platform vendor will incorporate your application is one that application developers must accept as the price for accessing the platform vendor’s customers. Data clean rooms are a good example: once independent applications, they are now offered as core platform services. Machine learning and artificial intelligence are following the same path.
The transition from silos to platforms is far from complete, but the industry direction is clear. The entry of cloud and cloud data platforms as alternatives to customer clouds is likely to accelerate the change. As we’ve already seen, platforms are a mixed blessing for application developers: they gain access to a larger audience but risk of being undercut by a platform product extension. (The good news is those extensions often involve acquiring a leading application.)
Application vendors also benefit from the platform providing data management functions that the application vendor would otherwise have to build for themselves. This simplifies application development, freeing resources to improve the primary application functions. Unfortunately, it also makes it easier for other companies to build competing applications. This lower barrier to entry (and to continued survival) is one reason the number of marketing applications has increased so sharply in the past decade.
The only way for application developers to escape this "app trap" is to themselves become platforms. Many aspire to this and some, including HubSpot, have had some success moving down that path. But most are too small to support a robust app marketplace and certification program or to attract a critical mass of application vendors.
And what does all this mean for marketers and other business users? Since the platform model isn’t new, we can already draw on experience. It’s a mixed bag: buyers have a greater choice of applications for any particular task, which is mostly good but does create a burden of having to choose. Buyers are tightly bound to whichever platform they select, since switching platforms means switching the related applications as well. This creates a quasi-monopoly situation, with the potential for higher prices and poorer service that comes with it. Raise your hand if you’ve seen that already. This will only become worse as platforms become more common, making it harder for independent application vendors to survive and, thus, harder for companies to assemble their own ‘best of breed’ architecture from separate point solutions. In other words, you may miss those silos when they’re gone.
It doesn’t have to be this way. The alternative to a platform architecture is a true composable architecture, where different applications are written to common standards that are not controlled by any particular platform. The ecommerce world has managed this to some extent, with ‘headless’ solutions and standards enforced by the MACH alliance. Standard data access languages like SQL are another example. That the customer data world will come up with similar standards seems doubtful, although there’s always hope.
It’s also possible that integration platforms will let different systems interoperate even without shared standards. That’s certainly their goal but so far results are limited by the effort to accurately capture all the nuances of one system and present them to another without losing anything in translation. AI might well change things by making this translation easier. But AI might change a lot of things, so let’s not count on any of them quite yet.
For the immediate future, then, we can expect a handful of platform vendors to control customer data at a growing number of companies. Those companies will be captives of the platforms, but it will be a fairly enjoyable captivity, with many pleasing applications to choose from and few obstacles to working as they please. Every so often they’ll pull back a curtain and see the bars on the windows. Some will be so unhappy that they'll escape. But most will accept the limits of their chosen platform and get on with their work.
No comments:
Post a Comment