How To Be A Platform: Making Tough Partnership Choices
Disclaimer: I work as a partner of Twitter/Gnip and Datasift. I’ve consumed data from both, and Klout distributes data through both.
On Friday, news came out that Twitter was electing to not renew its contract for data resyndication with Datasift.
Years ago, Twitter quietly made available a product called The Firehose, a near-realtime stream of each and every public tweet, to select partners, including some early social media analytics companies, as well as (in time), two syndication/resale partners, Gnip and Datasift.
The resale partners offered an important platform and economic advantage: Twitter didn’t have to staff and spend time on negotiating, marketing, scaling, and tooling the distribution of data for the enterprise customers and startups capable and desiring to innovate with the data. Meanwhile, Twitter would maintain its REST APIs and a smaller-scale set of streaming APIs, building and scaling a different sort of platform that’s been dissected many times before, but focused primarily on consumer applications.
Twitter bought Gnip about a year ago. According to the latest Twitter analyst call, Gnip represented over 90% of the Twitter data reseller market. This actually illustrates one advantage of platforms in that by offloading the scaling of the data business, Twitter can later come back and acquire and incorporate that success.
Gnip and Datasift competed fiercely over an artificially engineered scarce market for Twitter data. They built distinct product sets from each other, illustrating the malleability and ingenuity of both companies. Gnip’s model was geared to serve technologically rich enterprises, brands, and a number of (now big name) startups. Products ranged from percentage streams (50% of tweets at any moment, 10%), to a robust programmatic rules engine (their Powertrack product) at varying price points. Most startups would blink at these prices. Gnip’s revenue model was charging for the technology to deliver tweets, various add-ons and enrichments (including Klout), and Twitter’s required license fee for the tweets themselves.
Datasift gradually moved from pure distribution to a value-added, and even “volume reduced” model, as well as offering a pay-as-you-go self serve front end; their advantage was requiring less engineering work to filter, process, and store data. It could be said that Datasift was the more savvy platform provider, in that it was integrated with multiple technologies up and down the whole stack. To me, this model started to look a lot more like an analytics provider themselves, not a data distributor. They were adding signficant value on multiple touch points, which Datasift could reap income from. As their filtering improved, the volume of licensed tweets may go down, leading to potentially less revenue to Twitter.
Expected Platform Behavior
This weekend, 3Scale CEO (and friend), Steve Willmott wrote about Twitter’s mistake in ending this program:
Building a platform ecosystem that genuinely works requires three types of parties — the platform provider (Twitter in this case), the recipients of the value (customers) and third parties (that innovate in and around the platform for the benefit of the customer). Anytime one of these is shut out, there is potential value left on the table — in Twitter’s case, it’s likely to be a great deal of value.
In theory, this is the foundation of platform organization, and is expected platform behavior: you encourage your third party relationships to create value for their customers for the benefit of you, your customers, and your customer’s customers. Everybody wins. But reality works very differently, and Twitter’s been in a situation like this before (as Steve points out — it’s a good writeup).
Every treatise and advisement on platforms leans heavily on perfect systems operating as defined. They assume that you’re offering the right platform products, you’ve chosen (or you’ve been chosen by) the right partners, and they are delivering the right products to the right customers. What if one or more of these things isn’t true?
If you’re not offering the right products, you should be able to change. If the products your offering are producing the wrong platform effects, you should be able to change. There’s various prescribed methods of doing this — sunsetting, replacing products, versioning, changing the terms.
Likewise, if you’ve chosen a few partners because of their unique technological capacity — and you have to believe that Twitter had very few options for so-called Firehose partners at first — are you then beholden to never adapt and change those relationships? If your platform or company grows up, and the industry itself has grown up around you, are you still required to head down the same road, with the same partners?
I think if we take a look at Twitter’s partnership with IBM, we may see more reasoning for the ending of a relationship with Datasift than we do in looking at the Gnip acquisition. IBM’s products, including Watson, enables a very similar (although not as accessible) analytical, NLP, and predictive ability to leverage tweets.
Continuing that train of thought — what if a platform partner was, by many measurements, a bad partner? I’m not saying that this is the case here, but let’s posit a situation where the licensed use of Twitter data is “distribution”, and a partner signs an agreement saying so, and then goes and builds “analytics.” Or any number of other poor platform partner choices, like incorrect marketing materials, unlicensed promises, or other violations or encouraged violations of terms to downstream customers. You could negotiate changes to accomodate this in a new contract, or you could cut ties with the partner, considering that while there is a platform ecosystem benefit in them being there — and there certainly still is — they may be doing a greater damage to that ecosystem.
In an ideal world, Twitter would have selected a new partner to put in place, and as I said, I think IBM fills that role. Their blog post doesn’t really call that out, and perhaps that isn’t helping. But there is, after a few recalibrations of platform strategy, a general malaise and negative attitude toward platform adjustments, even when they make the best business and relationship sense.
We can’t have a knee-jerk reaction that says “platform good”, “open good”, “retraction bad”, even when these principles are what we consider the underpinnings of a good platform. Reality doesn’t work that way. Reality has terms of service, business development reps and CEOs, competition, revenue. Reality can’t just sit there because taking the right action doesn’t fit the model of platform strategy.
I get that we’re in an early period for APIs, data, and platforms. We want the systems to succeed so we all succeed. But taking a moment to step back and understand the dynamics that contribute to healthy business relationships might underscore the rationale for this kind of action.
I think we’ll find that Twitter made the right choice for Twitter, and it’s very likely the right choice for the ecosystem. There will be a period of adjustment, and the transition will not go as smoothly as it could have — but their hand may have been forced by being faced with spending another year or more in a bad situation.
