Iterative Technology Approaches to a Media Distribution Strategy

As a small to medium-sized media company with a large catalog of digital media, finalizing a technology strategy to allow wide online distribution is a critically important task. It can also seem a daunting one. Vendors and potential partners are likely knocking on the door, each bringing a unique view of the technology problem and how their service or product can solve these needs. Consumer pressure also abounds, as each passing week sees new tools and portals for online media consumption. Certainly, Content Delivery Network (CDN) vendors, bandwidth providers, and full-services media-oriented solutions are in the mix. How should a small to mid-sized media distributor proceed?

A Sea of Solutions

The recent rise in the number of services available to the consumer, the semi-professional artist, and even to the media company directly has been significant. There are currently dozens of solutions that can meet various needs within the media distribution space. There are also any number of internally-deployed systems and infrastructures to handle the distribution tasks. We have observed a number of common pitfalls when medium-sized media outfits tackle this landscape, especially when the unit functions under a larger corporate ownership umbrella:

  • Corporate-wide, large-scale, all-or-nothing partnerships typically, especially in the highly competitive environment facing today’s media enterprises, leave no room for market testing and fail much more often than they are successful.
  • Group-by-group expirementation with consumer tools, while in most cases not harmful, usually is not quantifiable in terms of return. Creating a MySpace page, because it seems like the thing to do, and then collecting friends is not a standalone distribution strategy.
  • Internal technology efforts become mired in the internal workings of a large corporation, one which is not necessarily a technology corporation.

Flexible Experimentation is Key

The number of large-scale distribution programs that have launched, and subsequently failed, in recent years and months is often staggering. It is our belief that for the majority of small to medium-sized media companies, an effective strategy rests on the ability to experiment within campaigns. In a lesson taken from modern software engineering practices, an iterative strategy can often prove highly effective for a number of reasons:

  • By reducing internal cycle time and overhead, you can get in front of niche consumers with innovative new promotional vehicles quickly, and build a reputation by doing so.
  • By limiting the initial outreach to less-critical media properties, you can gain valuable insight on what works and what doesn’t, without risking the mission-critical properties.
  • By spending less time on organization-wide analysis and coordination, you can focus on deploying the solution quickly and spending that time to analyze and understand its impact, strengthening your overall strategy with each new piece of knowledge.
  • When it is appropriate to roll the strategy into a unified whole, potentially for use across all media properties, the insight and data obtained from the smaller-scale initiatives will be invaluable.

Given this, how can you adopt these types of flexible practices?


The challenge, then, becomes one of balancing the substantial capability made available by consumer-oriented services with the need to measure return and iteratively form the overall strategy through a series of learning exercises. To this end, we have found that often smaller technology firms with existing technology bases can offer a compelling solution during this phase of strategy development. Computing and data-serving platforms such as those available from Amazon (EC2 and S3) can greatly increase the robustness and speed of deployment for these solutions, allowing them to be offered as “for real” services for a fraction of the cost of large-scale CDN partnerships. The resources freed up by not involving a corporation-wide internal IT effort or large-scale partnership can then be put to use in gathering and understanding analytical data from these next-generation campaigns.

Over time, the value a media company can derive from a series of low-cost campaign deployments, in terms of strategic refinement, hands-on experience with the new technologies and user-interaction models, and avoiding costly and embarrassing mistakes, far outweighs that available from traditional closed-loop analysis.

Removing Friction from Online Transactions

This piece is about Transactional Friction in digital content sales. Future documents will address digital content business models more directly…
In the world of digital content sales, where optimal price points might be very low, a good business model depends on the lowest possible transactional friction. The two primary contributors to friction are transaction costs and the opportunity cost of TIME and general annoyance experienced by the user. If a near frictionless transaction environment for digital content can be developed, it promises to usher in a new era of digital content business models.

Imagine paying three cents to read a technical whitepaper or paying 10 cents to listen to the Grateful Dead, Cornell ’77, in high resolution audio. Grateful Dead

Most current digital content transaction models make this impossible. The solution has typically been to “resort to advertising” by translating page views into CPM and click-through revenue. While the ad model works as a monetization path for many types of digital content, it can be fairly limiting, especially for works that are highly attractive to a relatively small audience or in cases where the revenue from the CPM or click-through does not adequately capture the value inherent in the digital work. Ad Models do also have a significant annoyance factor associated with them, so they are not entirely frictionless for the content consumer.

Any transaction system seeks to avoid high abandonment rates since a high abandonment rate will lead to much higher subscriber/user acquisition costs which will impact long-term cash flows and can be real killers when a company is racing to acquire new customers (as in high network effect environments).

More detail on the friction creators:

    Transaction Cost:

Transaction fees for credit card and online payment system transactions are typically comprised of a fixed cost per transaction plus a percent of the total transaction value. (Google Checkout has, at least temporarily, removed this friction by offering free transaction processing through 2008 as a part of the Checkout’s launch). The fixed cost per transaction sets a natural bottom for digital content pricing.

    User Time/Annoyance Cost:

Users have a finite amount of time so a transaction system that fulfills a transaction with minimal time impact on the user should be favored. There is also a cost of user annoyance or frustration, which can lead to high abandonment rates, so systems must make the transaction as simple as possible for the user.

Here are a few of the digital content transaction models we have studied along with some pros and cons. While these are mostly web focused, they can apply to mobile and gaming environments we well.

The Stored Account Model: Users sign up and store their financial information on the vendor’s system. Future transactions do not require users to re-enter their financial information.

Examples: Amazon, iTunes Store, Fresh Direct fresh direct

Pros: Future transactions are time-efficient for the user since they only need to enter a user name and password, or possibly re-enter a subset of their purchase information for a transaction to occur. The advantage to the vendor is that they now have a “subscriber” that they can monitor and to whom they can deliver targeted recommendations over time. Once a user has put in the effort to sign up for system, they may be disinclined to spend time signing up with a competitor offering a similar service.

Cons: – This model presents a disincentive to execute the first transaction since additional user effort is required to establish an account. – The vendor still incurs a transaction cost each time a transaction is executed. – This model also requires that users reach a trust threshold with the vendor since the user’s personal information will be stored on the vendor’s system.

Stored Value Model: The user funds an account such that there is one “real” transaction where the user’s money is transferred to the vendor account. Further transactions all occur completely within the system until the stored value needs to be refilled. The transactional efficiency of this mode can be calculated as follows: Initial transaction cost/number of internal transactions conducted = new per transaction cost.

Linden MC

Examples: Amiestreet, Second Life

Pros: After the initial funding transaction, there are effectively no transaction costs for purchases within the system. Users also have a switching cost since they will have to go through some amount of effort to get their money out of a system’s “currency”. There is some incentive for the user to exhaust the account which will drive future user interaction with the site.

Cons: Users are hesitant to pre-fund an account unless a refund path seems to be low friction. Users must also be convinced that they will find enough content of value to exhaust the pre-payment amount. Please also consider the fate of BitPass who attempted to build a comprehensive stored value system and ceasd operations in January 2007.

A La Carte Purchase Model: User’s can visit a site and enter their financial information and make a purchase without having to join the site as a member.

Examples: This is a typically an option on most commerce sites.

Pros: Buyers may be more inclined to engage in the first purchase since the time required to make the purchase is limited. This model is incredibly efficient when online payment provider systems (like PayPal or Google Checkout) are used.

Cons: This model does not provide any future incentive for the user to return since there are no future user efficiencies. The user will have to go through the same procedure again and again. Very little user affinity is generated with this model.

Subscriptions Model: Users pay a recurring amount for ongoing access to an ever expanding body of content.

Examples: Rhapsody, Yahoo! Music, HBO

Pros: There is little transaction cost friction especially when quarterly or annual renewal models are employed. Revenue is less variable. Subscription pricing also takes advantage of bundle pricing models. Sites can also introduce effective switching costs (for example Rhapsody forces you to call to cancel your account which introduces a disincentive to cancel… a nasty form of lock-in).

Cons: User setup time and decision making may be more complex since users typically have a series of “packages” to choose from. Users are also making a higher level decision about a body of content, rather than an individual content element. This can lead to higher levels of abandonment. There is also the “Homer Simpson” problem. (Homer was known to put restaurants offering all you can eat specials out of business by consuming every scrap of food. In the Internet media context, if all users were streaming music from Rhapsody for every hour of every month, data costs would quickly erode margins.) Compensation models for content providers may be complex or introduce fixed costs for the vendor. Users will require new content to be available on the system which may cause fixed costs (like storage) to go up steadily over time.

Transaction Wrapping: Users who already pay for thing A can get thing B through thing A’s transaction system.

Examples: Trial Pay, Mobile Provider vending (ringtones etc.), On Demand

Pros: These models eliminate many transactional setup costs by the user. One example is, Instead of having to pay On Demand for every pay-per-view purchase, the purchase just gets tacked on to the user’s monthly cable bill. This increases the potential for impulse buying since almost NO additional work is required for the user.

Cons: this model requires a possibly complex relationship with another vendor. The primary vendor may impose a relatively high tariff per transaction, exposing the

A Cruxy Anecdote:

We have very much enjoyed developing transactional models for Second Life since it offers users a variety of low-friction ways to purchase digital content. Users can fund their avatar’s account and users can go around the Second Life virtual world buying digital content. Content can be purchased by paying a kiosk, another avatar or through real-world commerce web sites. We anticipate that users will be able to use Second Life to buy more and more “real-world” goods since this low-friction environment makes it an attractive transaction environment.

Second Life introduces another risk though since they have a market exchange for their “in-world” currency, the Linden.

Social Commerce at Ad:Tech 2006

The “Social Commerce” panel I spoke on Tuesday at Ad:Tech has been receiving praise as one of the best of the conference. I think it is deserved, though not due to any self-importance or coolness of the companies represented. The panel was great because everyone came ready to share their thoughts, learned lessons, and advice on how to engage social networks with tools for commerce – how to mashup the worlds of socialization with the traditionally closed silos of financial transactions.

Much respect to Stephen DiMarco of Compete for employing a format which helped limit the all-to-familiar powerpoint abuse issues at most events like these. Each person was given one question to answer using one slide and about ten minutes of time. Everyone abided by the rules, resulting in useful information being shared with the few hundred strong audience.

According to the Ad:Tech blog post on the panel, I said this at one point:

“Most social sites lean heavily on the audience to build the content and evangelize. To do this, he emphasizes exposing consumers to a deeper amount of content to encourage additional engagement.” Sounds good to me! I tend to go into auto-pilot mode when I get onstage, always saying interesting and sometimes profound things, but not really remembering any of it later.

Jeffrey Taylor, founder of, and most recently EONS, a sort of MySpace for the 50+ boomer crowd. Jeffrey seemed most excited to talk about how EONS is now doing infomercial style television ads, and how great the response has been. Apparently, its not as expensive as you think, and with his older target userbase, it completely makes sense.

David Andre spoke about the interesting work Mall Networks is doing engaging with known brands such as NASCAR. Every major brand wants their own commerce site, social networking site, and more recently photo and video sharing site, and these guys are right in the middle of that trend, stopping corporate america from poorly reinventing the wheel, while building customer loyalty in all the right ways.

Geoff Donaker of Yelp demonstrated that if you give people something to do, around a topic they care about, they will go at it like gangbusters. It was also interesting to learn that Yelp started as a social workflow application – connecting people with needs to those who might be able to fulfill the goal (i.e. “I need a babysitter tonight – anyone?”) – would have been a great app! Now Yelp is squarely focused on user generated reviews of local services – restaurants, doctors, bars, salons – (“Where’s the yakatori joint in Manhattan?”) Think Citysearch, but turned on its head so the reviews are the most important aspect of the site.

I had great time meeting all of these gentleman, and was honored to speak alongside. Best of luck to you all.

Writing Code in Three Dimensions

A funny post on BoingBoing regarding yet another thing Hollywood consistently gets wrong – writing computer code. I agree with most of the original post, and always laugh at how programming is portrayed in such “great” films as The Net, Hackers, and Swordfish, but there is one point on which I disagree:

“4. Code is not three dimensional
Remember in “hackers” when the gibson is depicted as a three dimensional city that the hackers must navigate through? Bullshit! We may use a dash of color in our shell to make things a bit clearer, but last I checked my terminal app doesn’t require OpenGL. I’m working here, bitches – I’m not playing quake.”

Obviously, this guy hasn’t been programming inside of Second Life. It truly is a new experience to be writing object-oriented code that is actually interacting with three-dimensional rendered objects. It is definitely not “playing quake”, but the skills I honed during my youth playing FPS games have definitely come in handy.

In Your Sim Writing Some Code

You can learn more about writing code “hollywood style” for virtual words through the Second Life Scripting page.