HowMashupsFail

From MashupCamp

Anil Dash leads this session on "how mashups fail."

or "When you've got people tracking mashup failures, you know you've arrived."

Goal: Discuss...

- What are the things that can go wrong? (they will go wrong)
- What are the problems that arise in real world situations and how do you deal with them?

What is failure?

- Bad user experience
- Security problem
- etc.

Common problems:

- Cheap Gas: abandonment of the site, or mothballing by developers
- Browser issues: many work only in Firefox 1.5
- Bad idea: putting anything on a Google map isn't by definition useful
- Overabundance: if 20 people create the same app, nobody gets enough users to real critical mass
- Success is penalized (your service gets cut off when...)
- Infrastructure: poor performance and/or reliability
- Lawyers! (any legal barrier, even internal to your company)
- Pollution: data cleansing
- No user manual / you must reverse engineer everything
- User education: mashups are new
- Hard to find useful Mashups

It is expensive to host an API, and eventually data providers will charge for us, or use advertising to make some money.

Why do mashups fail?

- We're doing everything as a hobby (we don't pay for data, we don't do usability testing, we don't document features for users). This creates low quality.
- We're racing around trying to build new things, and the barrier is so low that many mashups just aren't well conceived
- Those that ARE well conceived, can succeed. 90% will fail. It's not any different from web sites, but because the barrier has been lowered to create an application that uses external functionality or data now we have mashups...
- It's possible that current APIs are low quality, causing mashup quality to suffer and leading to lots of "me too" applications. For example, what will happen when ESRI releases a mapping API that's better than Google maps?

Examples of successful mashups that are going mainstream:

- frappr.com
- zillow.com

We're just starting the path to maturity. Today, this possibility is new and it's novel to just mashup two applications. As we progress, things will mature and more mashups will represent fully conceived ideas.

On the importance of creating an API:

- The government should be encouraged to release more data in mashup-friendly formats (RSS, APIs, etc.). It could save lots of $ to make it easy to get access to data without needing to integrate at the database level.
- People don't understand how important it is to release an API or syndicate data
- Putting up your API is going to become just as important to put up as a web site. If fact, payday loans it could be MORE useful in some cases to relesae an API than to have a web site.

Outcomes:

- What is success? Make something that has an impact, that gets to normal users.
- Help people understand the importance of creating an API; it will continue to get more important as more users interact with your data on distributed services
- If you want your mashup to succeed, don't forget the basics (usability, quality, reliability, usefulness)
- Please think about backwards compatibility if you publish an API. Know when you're shooting yourself in the foot. Provide advance warning if you *must* break your API, and explain how developers need to change things.

Conclusing quote: "How did God create the earth in seven days? He didn't have any legacy data."