TrialMAP© Trial/Repeat New Product Simulation Model

---

SYNOPSIS

We tried to achieve a balance between having a large enough number of parameters to make the model 'realistic in its details', and keeping it simple enough so that it could be used instructively to play 'what if' games even if NO HARD DATA is available.

Parameter inputs are derived from a combination of Marketing Data (Market Size [Units], Competition, etc.), Consumer Research Data (Trial Rate, Maximum Trial Penetration, Repeat etc.), and Product Manager judgments (Marketing Investment, Awareness, Distribution, Pricing etc.). Reasonable DEFAULT parameters can be modified systematically even if NO HARD DATA is available, to play various 'what if games'.

TRIAL MAPŠ is suitable for both Students and Marketing Practitioners.

INTRODUCTION TO TRIAL MAPŠ

LOTS OF TRIAL/REPEAT NEW PRODUCT MODELS OUT THERE
During the 70's it seemed that Trial/Repeat models were almost as numerous as successful new product case histories... it appeared that everyone had one. Most of them were linked with Laboratory Test Market Simulators operated by different vendors, but there were also stand alone models generated by advertising agencies and also by academics.

IT'S HARD TO MAKE A REALLY BAD PREDICTION
No matter how the model was put together, it was hard to not be able to predict what would happen. There are two reasons for this:

1) MOST NEW PRODUCTS FAIL ANYWAY
Since 80% to 90% (if we don't count line extensions) of new products fail anyway, a model that predicts failure most of the time would be generally accurate no matter whether its underlying algorithms and assumptions were sound or not.

2) NEW PRODUCT YEAR END SHARE HAS NARROW RANGE
What's more, most of these models did a pretty decent job of predicting market share at the end of YEAR ONE. Again, it did not matter how the model was cobbled together. Market shares for new products tend to fall in a range of 0 to 10% in most cases, and between 0 and 5% in at least half the cases (unless we define the market as a small low volume, sub-segment with few or no competitors). The range of prediction is therefore virtually never going to be 0 to 100%.

To predict a share of somewhere between 2.5% and 5% was thus not going to be too far off the mark no matter what the underlying product or its model was.

AD AGENCY MODELS (AWARENESS)
For instance, models written by ad agencies tended to make Awareness a very critical component of success. If enough money was not spent on ads, 'said these models', the new product would have no chance of surviving, because Awareness would not be adequate, and this in turn would inhibit TRIAL. And of course, if there is no TRIAL, then REPEAT is largely irrelevant.

CONSUMER PANEL MODELS (DEPTH OF REPEAT)
For those vendors who also sold panel data or who specialized in lab markets or extended product use tests, it was therefore also not surprising that their models demanded inputs in areas that their consumer research arms provided. Thus, a purveyor of panel data for new products put heavy emphasis on a measure of DEPTH of REPEAT and the product's SHARE among TRIERS as being predictive of a new product's failure or success. Estimates of DEPTH of REPEAT could of course only be obtained from panel data.

R&D LIKES SAMPLING and/or REPEAT MODELS
These kinds of models also tended to be very popular among R&D types, because their underlying assumption was always that the product they had painstakingly developed, was 'so great it would sell itself', and rather than needing ad generated awareness or distribution generated by the sales force, all it needed was sampling. After that, sampled triers would be so satisfied that the product would sail on REPEAT purchases alone.

SALES FORCE/BROKER MODELS
(DISTRIBUTION + COUPONS + P.O.P. ACTIVITY)
Finally there is the sales or broker contingent, who are absolutely certain that all the product needs is adequate DISTRIBUTION and shelf space, and some in-store promotion in order to succeed. They know from experience that once a product's sales volume on a store by store basis falls below a certain minimum, it loses shelf space and is ultimately no longer carried by the store. That's why they put a lot of emphasis on 'introductory pricing' and 'spiffs' and 'couponing' and 'in store displays' as being critical drivers of a new product's TRIAL and ultimate success. To keep track of this kind of information, store audit data generated via NFI or Information Resources (IRI) or in some cases, custom distribution studies is needed.

RECENT DEVELOPMENTS
The relatively few recent developments in this field (all academic) tended toward models where additional CO-VARIATES were entered into the model, since UPC SCANNING technology allowed for richer data sets than were previously available. Whether this resulted in greater PREDICTIVE ACCURACY is still an open question.

Probably the best synthesis of all the preceding modeling efforts is one described by Anne Martensen in her working paper 18 (SEE: Bibliography). In her paper she indicates that the computational gruntwork used in her modeling efforts has been taken care of by a PC based computer program. However, this program was nowhere to be found.

AN UNMET NEED?
In the new millennium, we have found that for just about every application need there are at least some free or shareware programs available on the WEB. Strangely, this did not seem to be the case when it came to Trial/Repeat modeling software for New Products.

The closest approximation to such software was the availability of some MACROS (XLS files) that ran on MS-Excel[tm]. If you do not have MS-Excel[tm], you're out of luck.

As far as we can tell, there are no free or shareware standalone programs for this purpose out there on the WEB today. That's not to say they don't exist. However, existing software tends to be very 'proprietary' (you need to be a client of the sponsoring agency or you need to use the sponsoring vendor's Lab Test Market Simulator or you need to use the sponsoring vendor's panel data etc.) and also very 'pricey' when you consider the total package. Even so, it's cheaper by far than a test market.

OVERVIEW

INTENDED MODEL USE

MODEL SKEW

LESS EMPHASIS ON SAMPLING
Since we assume that AWARENESS and DISTRIBUTION are the main TRIAL builders, we tend to put less emphasis on situations where the product is assumed to 'be so good that it will sell itself', and therefore a greater or exclusive emphasis might be put on sampling or in-store incentives. The underlying assumption in such situations, is that once a consumer has sampled such a product, satisfaction will be so great that a high rate of REPEAT is assured. If that comes to pass, the product will continue to sell primarily based on its REPEAT momentum, which is in turn a function of product satisfaction better than that of competing brands.

LESS EMPHASIS ON HIGH TICKET HARD GOODS
We also tend to put less emphasis on relatively expensive hard lines markets (Personal computers, Lawn Mowers, Appliances etc.) where over a period of years, CUMULATIVE TRIAL is really the only variable that's relevant. The model is flexible enough to accommodate this situation by specifying a ZERO or very low repeat rate. However the main focus of our model design was not primarily tailored for such situations, as might be the case for the BASS MODEL (SEE: Bibliography).

FINANCIAL CONSIDERATIONS
Since the model estimates both unit and dollar sales, there is also a provision for checking how this revenue compares to what's invested in the new product venture. We restrict our scope to MARKETING INVESTMENT, since this will be the primary expenditure once the product is launched.

TIME FRAME
The typical time frame for the model is ONE YEAR. More often than not, the revenue for this year will not be sufficient to cover the MARKETING INVESTMENT. One can continue to run the model for a longer time frame, where the total MARKETING INVESTMENT will average to less per year than in a one year simulation. This will reflect the fact that MARKETING INVESTMENTS for a new product launch are typically FRONT LOADED.

TRACK RECORD
We readily admit that our approach to modeling begins with what amounts to a clean slate or TABULA RASA. Most Ad Agencies or Panel Research Firms or Vendors of Laboratory Test Market Simulations have something we don't, namely an extensive library of HISTORICAL TEST MARKET EXPERIENCE, by product category.

While we have some very general 'rules of thumb' covering the relationships between results of survey research and model inputs, as well as between at least some of the variables representing model inputs, we cannot offer the depth of TRACK RECORD DATA that vendors of PROPRIETARY SYSTEMS offer.This difference is reflected in the price that PROPRIETARY MODEL VENDORS charge for their products.

If you'd like to examine a very small sampling of the GRAPHICS available in each of our software products, JUMP to our GRAPHICS SAMPLER.

If you want to 'Try via DOWNLOAD', by all means link to our: FREE DEMO SOFTWARE.