Predictive Modeling in application software industry…
SoftwareAsia Inc.
The company and its new business model.
SoftwareAsia (SWA) is an I.T service provider, focusing on delivering technology driven solutions and services to a wide-spectrum of industry. SWA has been in the industry for more than six years with offices in the Philippines.The company provides flexible service offerings that leverage technology from mainstream to cutting edge solutions. With proven methodologies, tools and best practices we help ensure the successful implementation of technologies in complex project environments. The products offered by the company to its many clients are softwares for payroll system, financial management information systems, customer relationship management system, document management system, fixed asset management system, supplies and inventory management system, corporate portals and so on.
The company has been performing quite well ,though initially it faced some ROI problems , essentially because of difficulty in finding a comprehensive client base. The difficulty was over come by the founding team of the company establishing alliances with a software training group where the founders offered recruitment services for the graduates of the school.
Recently, the company has been thinking of using a new revenue model ,called Software as a Service (SaaS) in the application software industry in the country. The SaaS model, also often referred to as the rental model for software services ,is different from the traditional model where revenue is earned by selling software as a product in which customer makes a one time payment for permanent ownership of the software. The SaaS model offers the software to the customer in the form of a service . The customer does not own the software but simply use it and pay for the usage periodically based on the volume of usage. However whether SaaS revenue model is acceptable and needed in the Philippines is yet to be established and SWA is keen to evaluate this proposition.
The Business Challenge for SaaS Revenue Model
Recently, the Government of Philippine has launched its SME agenda with the objective to enhance their operations and make them more competitive by providing them with the necessary financial, marketing and technological assistance. SMEs are the economic backbone of the country and 99.6 % of the total number of registered firms in the Philippines. SWA feels that the SaaS model is especially suitable for the SME businesses because of its low cost of ownership and high and robust quality ,not given to breaking down, of the software product. Since the SMEs have comparatively less economic resources at their disposal as compared to large corporations ,thy are usually not willing to take the risk of owning a software product that has a small lifetime , as it requires frequent upgrades or willing to make upfront investments in software product without testing how well it will meet their requirements. Hence SMEs might want to try out the software on the rental basis. This would also enable them to switch their vendor in case they do not feel satisfied with the product. Further SaaS would even reduce the cost of ownership by eliminating the need for additional expenditure after the software procurement such as professional fees, infrastructure, hardware, training costs etc.
Value proposition to the application software industry
SaaS is a new concept in the application software industry. Adoption of this model would lead to the development of products that promise low cost of ownership, platform independence and secure transaction environment to users. Low cost of ownership emanates from the fact that users of the software will not have own the software anymore. They will just have to pay the rent for its usage. This will increase the bargaining power of the customers as they will not be tied to the software for life. They will have the option to switch to other vendors if they are not satisfied with the quality of product/service being provided by the vendor. This will put an increased pressure on the vendors to offer quality products at low cost. Since the infrastructure and product development cost would be borne by the vendors they will continuously try and improve their method of product development in order to curtail costs and economies of scale. Platform independence implies that the software will not be dependent on a particular technology for its deployment. This will save the customers from the need for extra investment in technical infrastructure and supporting software to enable them to access applications. The secure transaction environment is necessary as the data repository would be hosted by the vendors.
All these features ,namely the provision of low cost ,high quality, highly secure, platform independent application software, promising anytime anywhere access and developed using efficient means of production ,would add to the maturity of the application software industry. This will change the way revenue is earned in the industry .Time to market, accessibility and quality of service would also be critical factors in this approach.
Though convinced of the importance and relevance of the SaaS model in the Philippine application software industry ,before adopting the model within their own operations SWA Inc. decided to conduct a market analysis involving the acceptance of the model to its proposed target market. Thus they conducted a market research exercise where the primary sources of data would comprise surveys and interviews. The survey would help to determine consumer satisfaction, needs analysis , defining the effective customer segments and subsequently making an inference about the target market. The secondary sources of data would include study of relevant literature on SaaS and application software industry, documentation on internet ,research reports from global research agencies and so on.
For the primary survey data SWA designed a questionnaire to be administered to a sampling frame of SMEs which had the following components:
(a) a demographic section for SME company
(b) level of satisfaction on various features of software they have purchased and used
( c) their specific software needs
(d) awareness about the concept of using software on a rental basis
(e) extent of willingness to use such software on rental basis ,etc.
Questionnaire
Objective: to understand the market acceptability of SaaS
Please rate the following software attributes on your level of satisfaction with your existing software ,on 5-point scale:
1= not satisfied
2= neutral
3 = somewhat satisfied
4= satisfied
5= very satisfied
(Price of software)
(I) (a)How satisfied are you on the cost of procurement of your basic existing software.
(b) How satisfied are you on the high cost of availing professional services
( c) How satisfied are you on the high cost of application maintenance and support
(d) How satisfied are you on the high cost of training associates to use software application
(e) How satisfied are you on the high cost of IT infrastructure and hardware
(Quality of software)
(II) (f) How satisfied are you with the software application meeting the
specification
(g) How satisfied are you that the results of the software application are
accurate
(h) How satisfied are you that the results of software application are
consistent with for same values of input
( I )How satisfied are you with the security of data and information
offered by the application
(Usability of software)
(III) (j) How satisfied are you that the application can be accessed anywhere
anytime
(k) How satisfied are you that the program is able to generate desired reports
(l) How satisfied are you that the reports are well formatted and easy to
comprehend
(m) How satisfied are you that the software application can be easily customized to suit your requirements
(n) How satisfied are you that the software can handle complex transactions
(o) How satisfied are you that the software can be deployed on any technological platform
(p) How satisfied are you that the software offers easy navigation capability
(q) How satisfied are you that the software can be easily deployed
(Delivery of software)
(IV) (r )How satisfied are you that the vendor’s technical support personnel are prompt and knowledgeable
(s) How satisfied are you that the vendor adheres to the schedule in terms
of delivery of application
(V) Which of the following applications are you most likely to procure within the
The next one-year.
(1) Financial Management packages
(2) Payroll Systems
(3) Supplies and inventory management systems
(4) CRM systems
(VI) How aware are you about the concept of using software on the rental basis
(1) not aware
(2) heard about it
(3) somewhat aware
(4) well aware
(V) How willing are you to use software application on a pay-per-use basis
Willing Not willing
(VI) How much would be your budget for such a service offer
<10,000.month
10,000 to 15,000 per month
15,000 to 20,000 per month
20,000 to 25,000 per month
25,000 to 30,000 per month
(VII) Your company’s annual turnover
50 M to 100 M
100 M to 200 M
200 M to 300 M
300 M to 400 M
400M to 500 M
(VIII) No of people in your company …………………
(IX) Years of operation of your company ……………..
(X) Is your company export oriented yes/no
Empirical Research
The empirical research was conducted to obtain primary data from the target market comprising SMEs in the National Capital Region , to understand their software needs ,level of satisfaction on the various software features they have been using so far ,awareness of using software on rental basis and willingness to use the SaaS model.
To obtain a representative sample ,a multistage sampling procedure was used whereby the NCR region was looked upon comprising 50 different sub-regions. A random sample of 10 sub-regions were selected. Within each sub-region, thus selected, cluster sampling was used to select 10 SMEs. For the cluster sampling within each sub-region , SME clusters were looked upon within the sub-region and from the clusters any 10 SME were selected. From the sample of SMEs thus selected, 112 in number, the questionnaire was requested to be answered by CIO, chief information officer . For the survey , a researcher took an appointment with CIO and asked him/her the questions in the survey. So, the survey was actually filled by the researcher which ensured the completeness and authenticity of the process.
Thus a total of 112 completed questionnaires were obtained ,which led to a margin of error of 9 % under the confidence level of 95 %.
Towards Predictive Modeling based on the data
Testing the Reliability /Consistency of the different factors measured.
In the questionnaire the respondent SMEs were asked to give their levels of satisfaction on different constructs such as :
Price of software
Quality of software
Usability of software
Delivery of software .
In addition ,of course, the SME is asked to list their likely application softwares, awareness, proposed budget for SaaS service offer , annual turnover, size of company ,years of operation and willingness to buy.
The above constructs, often called latent constructs, have certain indicators associated with them, which are the questions under each construct.
For instance the Price construct is measured by cost of procurement, cost of availing professional services, cost of application maintenance, cost of training and high cost of infrastructure and hardware.
Similarly the other constructs have other indicator variables associated with them, as given in the questionnaire.
Before carrying out data analysis on the constructs ,first their reliability and consistency factors were evaluated. This was done by checking if the indicator variables under each construct measured the same feature. In other words , SWA checked high highly the items under each construct were interrelated. Since under Price construct there were five items their inter relatedness needed to be measured. For this, bi-variate correlation was not sufficient because this feature only checked correlation two by two.
To check the inter relatedness, there fore Cronbach’s Alpha estimate was used ,which measured the overall relatedness of all items under a construct. The Cronbach’s alpha should work out to > 0.8. If this does not hold , certain items are removed from the construct so that ultimately the alpha is greater than 0.8.
The Final Cronbach’s alpha values for the different constructs and the associated indicator variables ,which consistently inter related with one another ,are as follows:
Construct Cronbach’s Alpha Indicator variable associated
==========================================================
Price 0.8005 cost of procurement
Professional services
Maintenance &support
Training
Quality 0.8763 meeting specification
Accurate
Security
Usability 0.8173 desired reports
Easy to comprehend
Customized
Technological platform
Navigation capability
Easily deployed
Delivery 0.7373 prompt
Terms of delivery
Significance of satisfaction:
For the four constructs, the level of satisfaction was:
Significant for the first three ( taking the cut off value as 4) but not significant for delivery.
Analysis of variance (ANOVA):
In this analysis ,the company wanted to determine the distinct SME profile, if any, for the two levels of question 5. willing/not willing to use SaaS model for softwares. Thus , for ANOVA, the grouping variable was taken as willingness to use software application on pay-per-use basis and the dependent variables were :
---------------------------------------------------------------------------------------------------------
Dependent variable F-value significance
---------------------------------------------------------------------------------------------------------
Price 3.075 0.082
Quality 38.509 0.000
Usability 11.543 0.001
Delivery 0.002 0.968
Aware 0.730 0.395
Budget 8.844 0.004
Function 0.017 0.896
Turnover 1.910 0.170
Years of operation 232.281 0.000
Number of employees 21.422 0.000
Export 0.535 0.466
Thus there are two categories of SMEs, those willing to use and those not willing.
These two categories differ in their level of satisfaction on quality and usability. They also differ in their budget for soft wares, years of operation and size in terms of number of employees . For the other variables there is no significant differences in the two categories.
Linear Discriminant Analysis (LDA) and Predictive Modeling
Linear Discriminant Analysis was used to obtain the distinct profile of the two SME segments, those willing to use SaaS model and those ,not willing.
Here the equation used for the predictive modeling is :
D= b0 +b1X1 + b2X2 +b3X3 ….
D= 1, if SME is willing
0, if SME not willing.
The independent variables are :
Price, Quality, Usablity, Delivery, Budget, Turnover, No of employees, years of operation , awareness, function where software may be used .
The discriminant analysis tries to come out with a function which would discriminate /separate between the two segments in such a way that their within group variation is small but between groups variation is large.
Thus the Wilk’s lambda ,measuring within group variation to between group variation ,is as small as possible , where as eigen value measuring between group variation to total variation ,is as large as possible.
Summary of canonical discriminant function
Eigenvalue % of variance canonical correlation Wilk’s Lambda Chi-sq signif
2.946 100.0 0.864 0.253 142.76 .000
Structure Matrix
Independent Variable canonical correlation
==========================================
Years of operation .835
Quality -.384
No of employees .266
Usability -.212
Budget -.168
Price -.100
Turnover -.077
Aware .047
Delivery .012
Function -.001
In the above table we consider only the top 5 variables .
Thus the SMEs willing to use SaaS model are those with longer number of years of operation . Thus they have been operating for many years. Thus they are older companies.
They are also larger companies with larger number of employees.
They are not satisfied with quality and usability and have smaller budgets for Software use.
Hence if SWA would like to promote a marketing campaign for SaaS model, they should target companies with larger number of years of operation, older companies ,larger number of employees but perhaps having a small budget for softwares.
These should also be the companies which are not satisfied with price, quality and usability aspects of their existing softwares.
If the above equation for discriminant is used then the predictive power of the equation will be about 93.7 %; in the sense that 93.7% of the time the prediction of willingness/ unwillingness would match the actual happening.
Using CHAID for predicting customers willing to buy.
As mentioned and explained previously, CHAID is a combination of heuristic as well as statistical method which examines relationships between many categorical predictor variables and a categorical, usually nominal, target variable. It applies the Chi-square test on independence, also called Contingency table ,between the target variable and each of the predictor independent variable using the multi-way cross tab table . The null hypothesis H0 : the two variables are independent .
This iterative process works with repeated application of Chi square test between target variable Y and each one of the different predictor variables. The predictor variable which gives the smallest p-value provides the basis for first partition from the root node. Thereafter the tree ‘grows’ following the same iterative process of partitioning by the Chi-square testing.
The process of identifying the predictor variable with the smallest p-value is called the Bon Ferroni approximation.
Using CHAID for SaaS model for software application, the target variable used was, once again as in discriminant analysis , willingness to use .
The decision tress emerges with a few nodes only, because of the small size sample.
The index percentage levels for the three terminal nodes are :
Node index percentage
3 326.8 %
2 242.4 %
1 8.7 %
Hence the software company should target the nodes 2 and 3, which make up about 30.36 % of the market, with node 3 giving a response rate of 96.30 % and node 2 giving a response rate of 71.43 %.
These two segments conform to companies with years of operation 9 and higher , implying the marketing campaign should be targeted to SMEs operating for more than 9 years.