Ever notice that every financial story looks like the same template with the names changed once a work shift so that the editors and reporters have something to do? Reading financial news articles is like a lot like watching grass grow. In practice Writing them often feels like you are operating a lawnmower. Grad information from A news bureau restricted feed terminal, rewrite it, place a release timestamp on it so that it comes one second after the embargo ends, hit go, then on to the next one. Yes that really is the way it works.

Under the hood, a goodly portion of the information that appears as so called "breaking news" gets picked primarily by junior reporters monitoring a 60 minute embargoed press announcement feeds containing statements from companies trying to insert their “spin” into the media.

It’s called “journalism”.

If there was no information in the source feeds, there were few stories written about the subject. This is why you only see columnists publishing once a week and entire market sectors getting only two stories a day, an opening piece and an end of shift piece.

In 1997, I concluded a negotiations with several market data suppliers for CBS MarketWatch that resulted in a rich environment of raw data in computer readable form becoming available to online news engines.

So naturally the next step was to mine this data. I was also in charge of building quantitiative analytics tools for CBS MarketWatch in addition to building professional analytics systems for fixed-income and equity-research for parent company Data Broadcasting Corporation, so my team and I wrote a series of computer programs that created automated news stories to do securities specific automatic story writing, what is today called "microcasting".

These programs read a real-time market data feed and created reports on every stock trading over $5.00 in the three largest US markets. I took a copy of the Associated Press style guide and followed every rule. They were good journalism. People liked reading them and the analytics I selected to perform the filtering detected meaningful information on thousands of stocks every day. The program covering mutual funds created over 3,000 fund specific stories per day and updated over 75% of them at least once intraday. This compared to two stories per day on mutual funds in the equivalent portion of the Reuters feed at the time. These automated newswire was highly popular because readers were receiving focused news about securities they owned.

My friends in the newsroom hated it. First they asked questions about who was writing these stories because they looked hand written. Not only did the programs use the style guides precisely, we had programmed random variants in the code so that the idioms of different types of writers such as male, female, young, old, formal, casual would be selected. The stories looked like an entire staff was typing furiously. Eventually the newsroom hit me with the real rub. I was crowding out the “real news” written by reporters and crafted by editors. We killed the program on that site.

But the concept did not die. I pleasantly ran into it again in 2000, as CTO at Houston-based Telescan where a very talented team was working on a program named DCipher that also did “automated content production" at the single security level. In this case, the output format was to be an analysis tear sheet. The rigid style guides and use of analytics to extract and interpret specifics were also an ideal packaging tool to address the emerging world of wireless devices and interactive time-sensitive broadband channels. The environment for such applications only ripens with time as higher speed wireless infrasturcture and devices arrive in the marketplace.

I believe Jean, this is paradigm change of the best kind.