software for data presentation, statistical analysis, marketing and prediction.

Free download:
LeoStatistic.zip or
(selfextracting winzip file)


  • Introduction
  • Data
  • Statistics
  • Results presentation
  • Samples
  • Popular statistics and data analysis
  • Market analysis for advertisement campaign.

    There is a frequently appeared practical statistical problem of market analysis for advertisement campaign. It can be easy solved by LeoStatistic software application. There are two major schemes incorporated in LeoStatistic these can be used for solution of the problem.

    Conditional distribution of parameters can show in details a histogram of response rates

    Ri = Nicond/Ni
    where Nicond - number of events falling in i-th interval in case of fulfillments of all set of conditions; Ni - all events for the interval without taking into account any other conditions. The response rate is a direct measure of tendency of any specific value of the tested parameter to increase or decrease a probability for positively reply (to be true for the set of conditions). The more deviate a response rate in the interval from Riav , average response rate for whole data set, the large its significance of the influence of the occurrence of that parameter inside the tested interval on the estimated probability of conditional parameter to have a desirable value.

    Observation of the conditional distribution for all parameters can give initial material to make a conclusion about most influencing parameters to include into final analysis.

    Scoring of the marketing campaign.

    Let's presume that we have a database table with results of marketing campaign. It could have column with variety of personal data of the potential and real clients, including, that is crucially important, their response on the advertisement offer. The problem can be formulate in following terms: at first to create a mathematical model described potential client and then apply the model for estimation of the probability for positive action of potential clients calculating scores for each of them to become an actual customer.

    To complete this task with help of LeoStatistic following steps have to be performed:

    Import historical data table of previous marketing campaign that includes as one of the parameter a response of the client. The most obvious values representing a response could be 0 for negative and 1 for positive (to buy or not to buy). The actual amount spent by this client will work too. Example of such table in form of text file can be download here

    Go to tab "Data" of the control panel and press button "Insert". Dialog "Insert" will appear:

    Insert score

    Check up the control "Score". In dropdown control do select parameter that represent response of the customer and set its border level overcoming which means positive response. User also can chose between creating more or less detail model by checking on or off a "simple algorithm" control. Obviously more detail algorithm will demonstrate better lift for the historical file but not necessary a more stable results one because it will incorporate not only most significant peculiarities. The big numbers of rules for score calculation could neutralize each other producing noise like output. So the best advise is to try both options and use your common sense. 

    Press "Insert" button.

    After some calculating time a report dialog will be shown:

    Lift curve and algorithm of score assingment

    At the chart two curves characterized a quality of produced model will be displayed. Blue one is so named "lift curve" that shows a dependence of part of records with positive response in portion of all records with given number of all responses with the best, largest scores. The higher this curves is lifting above diagonal line the better model works to select records with right responses. Red line presents the same picture in other form, giving response rate for part of records with best scores. Obviously for 100% of all data it will give an average response rate.

    The "recipe" edit control contains an algorithm by which scores are calculated. There are three alternative for each variables how to calculate their deposit into the score calculation. It could be by linear or square equation or just add/subtract a number if the value of the parameter is inside of the given interval. Finally there is a formula for normalization of produced scores between 0 and 100.

    Before closing report  dialog user can and should copy and paste results to save them in some other application. By clicking the button "Word report" the page in MS Word with all presenting information will be created.

    To use creating model for optimization of the marketing campaign the scoring operation has to be performed for the set of data that have exact the same format as used for creation of the model. Only one distinction between them must be that file to be scored has not to contain a response parameter as it was in historic file. The example of such file to be scored on the base of the model described above can be found here By editing in the control a "Take only % of best scores" user can create file with only given part of the records these have best scores. This feature permits to use this file as a mailing list immediately.

    It's worth to add several words about valid format of the scoring file. The first numbers replicating parameters of historical file in exact the same order must be put in the begin of each line separated with space, commas or semicommas. Rest part of the line could be used to place a ID of customer or even its full address. The output file will completely copy the initial scoring file except at the end of each line a two digit number of score will be added. The name of the output file will be the same a copy of scoring file with addition representing percent of best records is taken.

    Screenshots of the LeoStatistic software:
    click on picture to enlarge

    Building histograms
    Building histogram

    Distribution of two variables.
    Distribution of two variables.

    Approximation (constructor style interface).
    (constructor style interface).

    3D view.
    3D view.

    DOW trend.
    DOW trend.

    Signals revealing
    Signals revealing.

    Near neighbors method
    Near neighbors method.

    Harmonic analysis.
    Harmonic analysis.

    Fit with free format formula.
    Fit with free format formula.

    Curve fit of crystal growth rate.
    Curve fit of crystal growth rate.

    Get data from image file.
    Get data from image file.

    Data analysis  Crystal growth simulation  Internet robot  Photoshop and image analyzer  NetCDF editor  Calculator
    Software archive  Expert database  Photo album  Maverick thoughts  Open forum  Search for cheap sale 
    Home  Products  Partners  Service  Contact
    Copyright by LeoKrut