Welcome to OpenForum. Join our global community: Sign up, Login or Learn more

Profile

Recent contributions

(4)

Contribution list

Bloomberg's hugely profitable value capture formula derives from the ability to price discriminate by bundling all its products into a single annual price. This pricing mechanism is critical to Bloomberg's profitability due to both negligible marginal costs, and the difficulty of determining each user's willingness to pay.

Bloomberg's hugely profitable value capture formula derives from the ability to price discriminate by bundling all its products into a single annual price. This pricing mechanism is critical to Bloomberg's profitability due to both negligible marginal costs, and the difficulty of determining each user's willingness to pay.

Photo of Joe Holliday
1 0

Recent comments

(3) View all

Lior, sorry for the slow response here. It sounds like you understand this better than I do, and it's been awhile since I've looked at this directly, so this consider the below with that in mind!

Splunk makes a distinction between the data that can be rapidly queried, and data that is harder to reach, kept in "cold store." The cold store data is indeed highly efficient & compressed, and it's great to hold on to for compliance purposes, but it's difficult for analysts to quickly pivot through while they are in the middle of an investigation.

It seems to be up to each user how much to put in cold store and how much to keep easily searchable, so you can compress it after one month, three months, whatever makes sense to your organization.

As far as size overall, I'm sure that Splunk is highly efficient across the board, but I was just pointing out that the decision to full-text index logs requires a lot more storage space than if you apply a uniform schema up front. I think one of Splunk's key insights was that the cost of storage will keep plummeting, so it's worth it to just go ahead and full-text index it.

link

Joe commented on Skybox, Big Geospatial Data

Great choice for this post. I've worked with imagery a lot in the context of national security intelligence, and what' I've seen of Skybox's imagery highly impressive - best I've seen in some ways. Selling this to commercial customers seems like a slam dunk, but it's not as easy as it sounds at first.

Google Maps Enterprise recently shut down because it was so hard to build a profitable business model around what commercial customers wanted. Google made the decision to empower 3rd party developers to serve this market, and just charge these developers per API call to the maps database.

Skybox will figure this out, but it might take a bit. Patrick Dunagan (HBS '14) is at Skybox working with big commodity traders (Glencore, etc) to figure out how Skybox can add huge value to their trading operations.

Great to see Palantir up here as always! I will try to keep this short, but having spent some time with Palantir, I want to clarify something: Palantir doesn't "own" any data, they just make tools to help customers with their own data.

This may seem like a minor distinction, but it's actually pretty important, largely due to the sensitivity of the customer data Palantir deals with. The Palantir folks who work directly with customers are always carrying two laptops - one from Palantir and one from the customer's IT department - because they can't mix customers' proprietary data together. So the point about the positive feedback-loop, and the ability to refine algorithms better than others doesn't really hold in this case.

The other reason why "algorithms" aren't as central as many assume relates to Asaf's question about unstructured data. Palantir is founded on the belief that computers can't deal with data by themselves - they need smart human analysts working with it to figure stuff out. What Palantir does is dramatically amplify the power of human analysts.

Palantir certainly ingests structured databases where possible, and applies machine learning algorithms when plausible, Palantir really shines by putting unstructured data in front of teams of human analysts who can apply structure to it through the interface. In the case of video - the most annoyingly unstructured data out there, the analyst can easily create database objects - events, people, items - that are linked to the appropriate point in the appropriate video file.