NEWSLETTERS

A Sampling of SPRING SYSTEMS' Newsletters

This page has been visited: times since [08/28/98]

DATELINE: 07/29/99 The current subject has to do with HANDHELD ARRAY/MATRIX CRUNCHING and the RESURRECTION of DOS in a new MEDIUM. INTRODUCTION By way of background, since the late 70's we've been users and fans of APL (A Programming Language). The compactness (you can write a multiple regression program using a single line of code) and sheer power (especially as a prototyping tool for Math and Statistics) of this language is enough to give a knowledgable user a 'RUSH', and has given rise to a finite but fanatical following, bordering on CULT status. The downside for APL is that it requires a special character set, is interpreted [slower for things like explicit loops], takes a bit of effort to learn, and is hard to read without extensive comments. These challenges have kept the number of 'TRUE BELIEVERS' at levels that indicate that 'APL is not for everyone.', compared to other programming languages. Since 'COMPACTNESS IS GOOD' is a leading theme with APL, we've also wondered why someone has not implemented HARDWARE COMPACTNESS and a handheld calculator device run by an APL engine. The likely answer is that there are too few 'TRUE BELIEVERS' to make this economically viable. However, all this has changed in the past few years. HANDHELD PC's While there are still no hardwired APL calculators, both Palmtops and Handheld PC's have recently come upon the scene. In theory, both have enough memory and computing power to run APL. Sadly, there is no APL system written for these platforms (although there is a J [close relative to APL] system). DOS IS BAAACK What to do? Well, it turns out that there is a DOS under Intel 80188/80186 emulator that runs on various HANDHELD machines. It's called XTCE (http://www.pyram-id.demon.co.uk/XTCE.html) by Ian Dean, and is available as shareware. Once you have this installed on your HANDHELD, you can also install one of the many demo/shareware APL systems available for INTEL/DOS machines. We tried IAPL, which while compact, is very slow relative to other implementations (Sharp APL, APL Plus, APL2, TRYAPL, RATAPL). All of these implementations work, though only IAPL does the special APL character set on the CGI system emulated by XTCE (the others work only in ASCII mode). Nevertheless, the dream of HANDHELD APL has been realized. TOO PRICEY? Whoops, there's one more problem. These handheld marvels cost between $600 to $1000 NEW ($150 - $600 for Palmtops), plus there is the modest cost of the shareware software ($50 or less). This is a little bit out of the price ballpark of a $150 hand calculator. We're not quite there yet. THE MAGIC of USED/DISCONTINUED While it's true that NEW handheld hardware is PRICEY, there is a discontinued item called the Philips Velo1, that does everything you need (especially with an 8MB dynamic memory add-on) and is a potential bargain. You can place auction bids for this item on eBAY.com, and pay somewhere between $130 and $180 for it. WAIT, THERE'S MORE The Velo1 can also do a lot of neat things outside of DOS, under the Win-CE operating system. But as long as you have the DOS emulator, guess what else you can do? You can install and run the entire TURBO Spring-Stat for DOS system. A full fledged stat system in the PALM of YOUR HAND. WOW!! cul8r pavi DATELINE: 03/19/98 The current subject has to do with SCIENTIFIC METHOD and REASONING. INTRODUCTION We recently read an article in the American Marketing Association publication Marketing News. (circa March 1998). In essence, the article cited the desirability of tying advertising effort to sales, and then proceeded to give a number of 'HOW TO' rules for how to create advertising that effectively does this. The basic thesis was that 'CREATIVE' advertising without sponsor identification and a sales pitch was ineffective. On an intuitive level, we had no BIG problem with this. It seemed reasonable that in order to persuade consumers to buy, the ad had to identify the sponsor and sell the virtues of his product. The whole thing seemed well within the parameters of the time honored Hierarchy of Effects model. Nevertheless, there WAS this little nagging thing about the fact that we had read in the Journal of Marketing Research (Lodish et al. circa May 1995) an article that suggested that tracking advertising effects via UPC scanning of sales data was challenging because the sales impact of advertising copy differences was minimal and therefore hard to measure. The data suggest that sales response is primarily to copy that's recently changed and/or is geared to increasing penetration and changing rather than maintaining attitudes. This study also referenced an earlier study in the Journal of Advertising Research (Haley et al. circa April 1991). The Haley study indicated that for ANY pre-testing effort that could be correlated with results of UPC code tracking of buying behavior along with controlled media and creative input, the indications were that independent of AD SPENDING, sales gains were related to consumer perceptions that the ad is 'NEW/DIFFERENT' and to a degree REWARDING TO WATCH. As diagnosed by pre-testing techniques designed to spot this sort of thing, and subsequently correlated with UPC behavioral sales data, LIKEABLE (CREATIVE??) ADS seemed MOST EFFECTIVE in CONTRIBUTING to SALES. WHAT'S THIS GOT TO DO WITH SCIENTIFIC METHOD There are a few things that IRRITATE. First of all, the article in the Marketing News sets up a DEPENDENT VARIABLE incremental sales, but then makes no effort to relate the suggestions they make for 'EFFECTIVE COPY' to this dependent variable. In fact, they give NO EVIDENCE WHATSOEVER that their suggestions for 'EFFECTIVE COPY' have anything to do with measures of EFFECTIVENESS BASED ON SALES. If anything, empirical findings (see above) suggest that the basic premise of: 'HARD SELL COPY = SALES' is contradicted by the evidence, which indicates that 'LIKEABLE' ads SELL. While we don't deny the possibility that HARD SELL ads could be LIKEABLE, we will defer this possibility to your own aesthetic judgment. Even though there were problems and weaknesses in both the Lodish and Haley studies (i.e. [1] They only used data voluntarily contributed by AGENCIES and PRE-TESTING SERVICES... not the full universe of possible data. [2] Their measures from different sources were often of the apples and oranges variety commonly found in META-research.) and they are by no means the last word on the subject, at least they made the attempt to relate differences in ad creative style as measured in pre-tests, to a SALES based DEPENDENT VARIABLE. WITHOUT A CLUE The scary thing is that American Companies spend BILLIONS of dollars on advertising but for many, the techniques used to track its effectivenness seem to be based primarily on anectdotal (QUALITATIVE RESEARCH) evidence, secondary measures (i.e. stuff that's easy to measure but has no relation to SALES), and self serving retro-fitted case histories. As a result, there is still no reliable research that links 'Communication' or 'Recall' to SALES effective copy, or PROVES beyond a doubt that BABIES and/or DOGS 'Work as effective advertising'. True, there is UPC scanning and various services available that can track results unambiguously. Unfortunately, users of these services insist on designs to 'PROVE THEMSELVES RIGHT' rather than on designs to determine if ADVERTISING or its individual components work. IN THE DARK AGES Because we don't KNOW or DON'T WANT TO KNOW, the state of knowledge on what works in advertising is PRE GALILEO (i.e. in the DARK AGES). Galileo understood what was NECESSARY TO SET UP AN EXPERIMENT to prove/disprove a hypothesis in physics. Unfortunately, advertising practioners in THIS century don't have a comparable understanding of EXPERIMENTAL DESIGN (or if they do, a willingness to use it or to share the results) to demonstrate that their stuff 'WORKS in terms of SALES'. In other words, the experimental design techniques currently used to measure advertising effectiveness ARE 400 YEARS OR MORE BEHIND THE TIMES. There are of course exceptions, but by and large PSEUDO and JUNK SCIENCE prevails. What's more, the LOGICAL structure of ads ( both COMMERCIAL and POLITICAL) does not stand up to the RIGOR of ARISTOTELIAN LOGIC, so we can rightfully say that the level of current media discourse is more primitive than arguments in the B.C. (2000 or more years ago) ERA. NOT TECHNOLOGY'S FAULT It's NOT that the TECHNOLOGY isn't there. It's that like VAMPIRES, no ADVERTISING AGENCY or COPY TESTING FIRM wants to be exposed to the SUNLIGHT of OBJECTIVE EXAMINATION of EFFECTIVENESS as it relates to INCREMENTAL SALES. The likely reason is that advertising quality (as opposed to ad spending) will be shown to have LESS IMPACT than ADVERTISED. Even UPC tracker firms like NIELSEN and INFORMATION RESOURCES, though they encourage it, don't seem to insist that their clients use SOUND EXPERIMENTAL DESIGNS. Better a BAD design than a lost FEE. IT'S YOUR CALL So there it IS. Technogy is NOT the problem. It can be and is SUBVERTED to serve Corporate Politics. Meanwhile, there is NO OBJECTIVE info on WHAT WORKS IN ADVERTISING. cyl8r pavi DATELINE: 01/23/98 The current subject has to do with PREDICTIONS and PROGNOSTICATIONS. INTRODUCTION The possibility of being able to PREDICT THE FUTURE has tantalized us since we drew bison on cave walls and relied on the local Shaman to tell us if more would be there when we got hungry. Since those days, we have both OVER and UNDER estimated our ability to PREDICT. This is partly because people tend to REMEMBER when a prediction comes TRUE, but tend to forget misses. What's more, truly reliable prediction is taken for granted, and many don't recognize reliable predictions as legitimate forecasts. PROPHETS WITHOUT HONOR We tend to honor people who may have gotten something right by way of a lucky guess, and at the same time ignore the fact that they are wrong much more often. We also tend to ignore REAL predictive abilities that are taken for granted. A prediction is a prediction only if it's iffy. WEATHER For instance, under modest time and geographic area constraints we can reliably predict the weather. With the aid of satellites, computers, and a wealth of new information regarding geophysics and its effects on weather, we can now predict weather roughly 12 hours in advance. Even 24 hour reliable weather forecasts are currently beyond our capabilities. This suggests that we still have much to learn about how weather works on a local basis and with short time parameters. CLIMATE Strangely enough, if the scope is global, the information is fuzzier, and the time frame is 20 to 50 years, as is the case regarding GLOBAL WARMING, we are not nearly as humble about our limitations in understanding and our subsequent ability to predict. Perhaps it's because GLOBAL WARMING gurus are less accountable if they make a bad call than your local weatherman. But all of that is a different subject. WHY FORECASTING FASCINATES The fascination with PREDICTION comes partly because we tend to think of it as a thrust into the UNKNOWN.

A more general definition of forecasting is simply to be able to foretell future events within circumscribed parameters. For instance, even before the Babylonians we have been able to predict the phases of the moon, eclipses, the movement of the planets, and the seasonal positions of the stars. We can also predict what will happen if we jump off a 39th floor balcony, smash a hammer on our thumb, or hold our finger in a flame. UNDERSTANDING FACILITATES PREDICTION It appears that if we understand how something works and can thus predict future outcomes from present actions (i.e. we understand cause and effect for a particular situation SEE: Proving CAUSE and EFFECT), this is regarded as no big deal. The fascination with prediction is only there if we don't understand the underlying process. The ability to PREDICT is in fact a measure of our UNDERSTANDING of how something works. Theories and conjectures, no matter how erudite or well meaning, must ultimately be subject to the test of PREDICTION. This will probably come as a major shock to Social Engineers/Scientists, Politicians, Lawyers, Media gurus, and various assorted 'Advocates and Activists'. There are (arbitrarily) four types of prediction: COMPLETE STAB IN THE DARK WITHOUT A CLUE This is best left to politicians and various activists, whose predictions are probably Agenda or Idealogically driven. They are accompanied by a call for an immediate political solution to avert whatever disaster they are predicting (usually a few years off). Whether the predictions hold water or not is irrelevant since politicians/activists are betting on the fact that you are not likely to hold them accountable for mistakes in prognostication. The predictions themselves are so vaguely stated that they are difficult to test against actual outcomes. If this kind of thing interests you, talk to your Congressperson or President:FORECASTING via the White House LOOK INWARD AS OPPOSED TO FORWARD This includes various divination techniques such as I-Ching, Tarot etc. These are projective devices that let you look at yourself under the guise of PREDICTING the FUTURE. At least the recipient of the forecast brings something of his/her own experiential data to the table in this type of situation, and the projective stimulus may trigger some real insights. Again, the predictions are vague and difficult to test against actual outcomes. If this kind of thing interests you, check out the following resource: FORECASTING via Occult/Tarot EXTRAPOLATING FROM PAST DATA This is essentially 'Naive Forecasting' in that we collect historical time series data, fit a specific statistical trend model to it, and make future projections with the implicit assumption that past conditions driving the process will continue into the future. There is no attempt made to understand or causally model the underlying process of the time series, hence the term 'naive'. It is by no means 'naive' in terms of the expertise it demands from the user, although this type of forecasting can be learned with time. The predictions are amenable to testing against actual outcomes using measures such as Standard Error of Estimate and R squared. If this interests you check out the TIME SERIES and NON LINEAR REGRESSION modules in the SPRING STAT MODULE SUMMARY: FORECASTING via Statistical Modeling MODELING THE PROCESS AND PREDICTING VIA THE MODEL This covers everything from NEWTONIAN PHYSICS for everyday events to RELATIVITY and QUANTUM MECHANICS for more esoteric phenomena. It also covers less formal processes such as being able to predict what happens when you put your finger in a flame. The key element here is that we think we understand the underlying process well enough to model it mathematically and subject the model output to the relatively stringent criterion of PREDICTIVE VALIDITY. The predictions are amenable to testing against actual outcomes using measures such as Standard Error of Estimate and R squared, as well as common sense in many instances. If you're interested, check out a variant of this approach which attempts to model MARKETING phenomena (hardly as well understood as NEWTONIAN PHYSICS) to within acceptable tolerances, and to forecast the outcomes of various marketing mix scenarios using the resulting model: FORECASTING via Mathematical Modeling All of this is still a 'Work in Progress'. If you've managed to stick with us this far, drop in to our WEB site: SPRING SYSTEMS Analytic Resourcery. cyal8r pavi DATELINE: 12/19/97 The current subject has to do with FREE stuff and what it will cost you. INTRODUCTION This is really a continuation of the same themes as were touched upon in the 11-13 Newsletter concerning HIT COUNTERS. However, this time we'll touch on other freebies such as WEB PAGES and BANNER EXCHANGES. THE NEED TO FINANCE FREEBIES The only surefire ways to make money on the WEB involve either PORN, DIRECT SALES or ADVERTISING. The first two need not concern us in this context. If you are giving things away to surfers or other netizens for FREE, the only way to recoup your investment is to SELL ADVERTISING on the stuff you give away. It therefore follows that we should learn to live with the equation of 'FREE=ADS' , or the FREEBIES will start to disappear. The issue here is that not all FREEBIES and ADS/BANNERS are alike. BANNER EXCHANGES Exchanging banners strikes us as being a fairly good deal. It is essentially reciprocal in that you allow space for someone else, and vice versa. The only potential inequity is that your BANNER buddies may have a target audience that's totally useless to you, while the opposite may not be true. Assuming that reciprocal audience quality is not a problem, you still have to make a judgment about when too much clutter starts to hurt your page and becomes a liability. Other than that, BANNERS are an OK deal. HOME PAGES While BANNERS seem by and large equitable to us, a new phenomenon in AD formats has appeared recently. At least two fairly large sponsors of FREE WEBSITES (Tripod; Geocities) are using this format. We don't have a name for it, but it essentially consists of a second window either embedded within or superimposed upon the FREE HOME page. For some browsers such as AOL 3.0 's version of EXPLORER, the AD WINDOW literally clobbers the underlying (your pride and joy) home page. This slows down loading and will P... OFF visitors to your page until they figure out that they can hit the minimize or exit buttons at the upper right of the frame to get rid of the AD window . This type of advertising is far more invasive than banners, and starts to provoke the examination of alternatives to FREE. It's also not a complete bargain for the AD sponsors. Noticing their AD in the context of being P.... OFF, does not envelop the ad in the best of emotional auras. And so it goes. Associations form in strange ways. cyal8r pavi DATELINE: 11/13/97 The current subject has to do with our experience with FREE HIT COUNTERS for our WEB SITE. INTRODUCTION It's free... count on it! This may very well be a slogan used to advertise WEB SITE HIT COUNTERS. And it might be very misleading. A BIT OF HISTORY It all started on 5/96. At this point we installed a 'plain vanilla' hit counter on our home page, provided by the people who sponsor our web site. Later we also installed the same counter on some of the other pages in our site. All copies of the counter reflected the hits from the home page.. they did not count each page independently. THE PROBLEM Everything worked just fine for a long time until 10/13/97. On this date, some evil 'counter dracula' succeeded in hacking into the counter mechanism. The counter and its copies were all reset to ZERO [0]. THE SOLUTION We really wanted SEVERAL new counters. A few (for mutual backup) for the home page as well as independent counters for the pages where we had previously had copies of the single counter. The first thing we did was to link to the META CRAWLER search engine. We indicated that we were seeking 'free hit counters'. Voila! The search engine came up with over 30 sites that in turn provided links to sites that offered counters. This is where 'free' became a joke (unless of course your time is free and unlimited). About half of the listed sites were no longer reachable. Of those that were reachable, roughly another half indicated that due to 'band width' constraints, they could not add new applicants. Another variant was that they could add applicants, but only for a one month 'trial period'. There were of course 'for pay' versions of the counter that were still available. Of the remaining usable sites, there were three varieties: PLAIN VANILLA Your basic counter without other amenities. Strangely, some of the simplest of these counters were the hardest to install. Some required a substantial waiting period before 'kicking-in'. PLAIN VANILLA WITH BIG UGLY BANNER Same functionality as #1, but big ugly banners to keep the service 'free'. Some also demanded detailed information obviously intended for sale to 'e-mail lists'. They wanted to know just about everything about you short of your mother's hat size. In exchange, you got a PLAIN VANILLA counter with a BIG UGLY banner. MODEST BANNER; MEGA-FUNCTIONALITY In this case, there were also banners, but of a much more modest size and scope than those in #2. In terms of functionality, however, these were much MORE than just plain counters. Each had a full report by time (Day, Day of Week, Month), Browser, Operating system, repeat vs. one-time visits etc. They were also easy to install (just copy some HTML code). We selected two #3's for our home page, and a combination of #1's and #2's for the other pages. Since none of the counters were implemented at the same time, and they count each page independently, they do not agree with each other. Over time, the disagreements while not disappearing, will at least become predictable. There seems to be INVERSE relationship between the functionality of a counter and the size and garishness of the banner. Clearly there are many shades of 'FREE'. Visit our home page as well as some of our other pages (SUBJECT INDEX; RECIPROCAL LINKS; WHAT's NEW; VIEW GEOBOOK) to see these counters. cyal8r pavi

o Return to: 'SPRING SYSTEMS HOME PAGE'