Showing posts with label google. Show all posts
Showing posts with label google. Show all posts

Thursday, December 6, 2012

Why banks fail at analytics


Bankers love numbers and poking their heads in spreadsheets. That’s why we find it rather odd that so few banks are really good at analytics. And even fewer implement their analytic findings to boost profitability. How is that even possible in an era when data is so abundant and revolutionizing industry after industry, from books to airlines?

The latest MACAW Bulletin gives you not one, but six reasons to why banks fail at analytics. These are observations from working with banks in countries across five continents. And the findings are surprisingly consistent. Maybe your bank is the exception, and you really are great at analytics, but odds are you will recognize your bank in some of the following.

Locked in Old World thinking
Look at what Facebook, Google, Amazon and Apple are doing with IT, analytics and rethinking business models. They have revolutionized everything from web hosting and search engines to music, books, and even payments. They are capitalizing on information, analytics and performance optimization. Banks are watching from the sideline. In the end it will be too late to be proactive to change. 

No culture for innovation
Banks attract many clever people, but they tend to be economists, accountants and lawyers. When did you last attribute ‘innovation’ to any of these professions? Sadly we find few truly innovative marketers, business-wise engineers, or entrepreneurs working at banks. And when the occasional brightly shining star does come along, she tends to fight an uphill battle for gaining acceptance at management level to challenge conventions.

No statisticians
Among the numbers loving bankers, where are the statisticians? One might find the occasional number cruncher with a statistical degree working in risk management at a bank, but they are surprisingly few and even rarer are those allowed to participate in business development processes.

Legacy systems
Mature banks will groan when you say 'we need this data'. The data is spread across different systems, and what’s worse, they’re legacy systems. Often it is a patchwork of spaghetti that almost no one wants to go near, and even fewer know their dark, inner workings. But legacy systems are not all to blame. A lot of it has to do with attitude.

Other reasons why banks fail at analytics:
  • Wrong KPIs
  • Risk adversity

This blog post is an excerpt from MACAW Card Bulletin 54: Why banks fail at analytics. Purchase your subscription today to obtain the full article. Contact sales@macawresearch.com. 

Wednesday, June 24, 2009

Journalism quality

There is an ongoing global debate concerning the quality of journalism, or rather its decline. The reduction in printed media, online advertising models, reduced government funding, increased demands for speedy delivery, an increase in tabloidization, and even Google and Wikipedia have been mentioned as reasons for the perceived decline in journalism quality.

This certainly isn’t a new debate. It goes all the way back to the advent of online newspapers and probably all new media. However, an interesting development is that it’s not just journalists who are waving the red flag. An increasing number of non-journalist bloggers vent their frustrations. It would appear journalism quality might have reached an all-time low in the recession, with the crisis eating into advertising income and more cost cuts finding their way to the editorial boards.

Market researchers who turn to news articles as a source of information continually face the challenges of journalism quality. While the average news consumer might be annoyed by spelling mistakes, casual stories being treated like “the latest scoop” in terms of headlining (which can even be directly misleading to attract clicks), and celebrity gossip and sports news finding their way into other content categories, market researchers face other challenges.

For example, the widespread use of cut+paste journalism, which eventually leads to articles becoming less than a stub – in some instances an article was only a stub to begin with. Unfortunately, this not only removes what can be important contextual or additional information, but it also can lead to extensive hunting in search of the original source due to a lack of or incomplete quoting practices. As a policy, market researchers should always go to the original source, which can involve following a surprisingly long trail of bread crumbs which might abruptly result in a dead-end.

Another concern for market researchers is the use of Wikipedia and Google as sources by journalists. Wikipedia is not an objective, quality-assured source, despite how intriguing it is and practical for instant look-up for less professional occasions, such as when discussing at a party whether the CD or LaserDisc came first, or who won the 2004 Eurovision song contest. Google uses a number of algorithms for determining search engine results, some of which are based on external references. Thus sites that fulfill certain criteria will get a higher rating, and thus come up early in journalist’s search results. This can create biased results, which are transferred to media sites and magnified by journalists who reference the found pages in their articles, thus creating further bias towards their sources. Popularity does not by itself determine truth. Google doesn’t separate right from wrong, that’s the time-pressed journalist’s job.

This leads to the question “who is the author of a given article?” In a number of countries, journalist is not a protected title. Even if it is, news articles can be written by “unlicensed” news desk workers, who aren’t specifically trained in critical journalism. Some newspapers only list an article sponsor, for instance “This article was sponsored by Company X”. All this makes it difficult to trust a growing number of online media sites in a professional research context. Market researchers need to choose their sources carefully in today’s online media jungle.

We await the day when new incentives surface which will encourage the rise of journalism quality again. And, as the critical reader will have determined, this post can easily be picked apart with regards to elements required to qualify as high quality journalism. That is because it is not meant to be such, rather an informal blog post on a topic of interest to us – and hopefully to other market research professionals as well.

Sunday, May 31, 2009

Testing Wolfram Alpha

Wolfram Alpha is Stephen Wolfram’s much hyped search engine project. We decided to give it a little test run to see how it can be of use in market research, and wanted to share our experiences with you. We are probably not the only ones in the market intelligence community that are curious about Alpha: http://www.wolframalpha.com

It is important to distinguish Wolfram Alpha from Google and other text-based search engines. From the hype one might almost be led to believe Alpha has artificial intelligence (AI) and can compute an answer to almost anything. It’s definitely not an AI, but uses a series of engineered scripts to provide some interesting results in select areas. The scripting is based on Wolfram’s Mathematica software, and numbers is what Alpha is good at.

Alpha is not a universal search engine. It does not return information on for example “EU regulations” or “Adam Smith’s invisible hand”, giving you the quickly annoying null result message “Wolfram Alpha isn't sure what to do with your input”. Alpha excels at numeric data and performing calculations, though, and it can conjure up useful tables and charts in instants, but only based on facts it has stored in its databases. In some instances, Alpha utilizes external sources too, typically for frequently updated data such as stocks.

For instance, we entered “DnB NOR”, which is the largest financial institution in Norway, and Alpha returned latest trade data and returns, a chart with price history, performance comparisons and correlations with various indices in table and various chart formats, projections for future pricing and alpha, beta and R Square values for daily returns for DnB NOR compared to S&P 500. Wow. Impressive, assuming the information is correct. To verify some of the data, we went to the Oslo Stock Exchange website and checked the latest trade value and weekly, monthly and year-to-date returns. The latest trade value was identical, but the return values were a bit off, as in Alpha reported an additional +5% on 30 day-return and -10% on year-to-date return. Hmm. That’s a lot.

This led us to check out the Source data link provided on Alpha’s search result page. It gave a long list of finance-related sources ranging from academic books to sites such as Morning Star, with a disclaimer that not necessarily all sources were used in the results. This can make quoting a source for particular data difficult, as the source list seems to apply for the search result page as a whole rather than for individual elements on the page. The Oslo Stock Exchange was not on the list, by the way, but several sources for aggregated financial data were.

We decided to try one of the recommended searches Alpha provides on its front page, and thus entered “EU populations.” Alpha returned a summary with the total EU population, the country with the highest population and the lowest, a chart with historical developments for the total EU population (does not say whether it includes ascension member states), a list of top 5 and bottom 5 countries ranked by population size – which can be expanded to a full list with a few clicks, a distribution plot with unspecified data (what does it show?), along with some demographic data measures for the countries that have the highest, lowest and median values for the population, population density, population growth, life expectancy and median age. This part didn’t strike us as intuitively very practically organized. A few clicks at a More option for yielded some additional demographic variables, such as annual deaths and birthrate fractions.

Of high importance, we were unable to find references to which years the various population data stem from – in fact there seems to be hardly any meta data available. We could be comparing years-old figures for some parameters and countries with very recent data for others without knowing. The source list doesn’t provide us much help in this case either. This is a major issue that needs to be addressed if Alpha is to be used by market intelligence professionals.

Summarizing briefly, Wolfram Alpha currently can produce substantial amounts of data on select subject matters of quantitative nature. For market researchers looking for quick access to on-the-fly generated charts and tables on data otherwise obtained and structured via other sources, Alpha could be a low-cost solution. However, seeing the source references are vague and Alpha seems to lack very important meta data, it seems more like a tool for a) the general public with simple data requirements, b) students and specialists in certain fields with a need to perform mathematical and statistical computations without using desktop applications.

We’ll definitely keep tabs on Alpha and play around with it some more, but currently it doesn’t strike us as something that will revolutionize the market intelligence industry. If anything, it is a supplement to other search engines and tools. As far as search engines go, Wolfram Alpha is certainly not a replacement for Google, but then again, it doesn’t intend to be either. We look forward to seeing Alpha evolve.