Free Evaluation

Analytical Advantage wants to make sure you're getting your money's worth. The free evaluation is designed to give you a sense of whether or not your data can achieve what the intended goal is. With that, there are a few constraints to the free evaluation, no paperwork needed.

  1. Max Variables. The maximum number of variables for the free evaluation is 10, inclusive of the target variable.
     

  2. Max Records. The maximum number of records are 1,000.  This should be randomly sampled to reflect the dataset. 
     

  3. Masked Data. Because this is a valuable and no paperwork is done, all variable names should be masked as well as any data elements that you may not want exposed. Absolutely no PII information. Data should be safe to transfer via email.  
     

  4. Datasets. No more than one dataset. 
     

  5. File Format. File format must be Excel, CSV, or TXT with the appropriate documentation on field type. No direct database connectivity. 

 

Reducing Your Cost

With careful inspection and some thoughtful data aggregation, there are material savings available.

Number of Variables

The number of variables are how many fields or columns are in your data. There are several ways to help minimize the cost while simultaneously decreasing the turnaround time. 

  1. Random Variables. Remove variables that are random or enumerated. Variables that are IDs or systematically assigned are variables that are unlikely to be predictive because of their inherent definition of being random or enumerated. 
     

  2. Redundancy. Remove variables that are redundant. Often enough dates are redundant, some systems don't have a discrepancy factor between a 'Book Date' and a 'Created Date.' Looking at all variables that are similar, based on your domain knowledge, you will know whether or not it will be added-value. 
     

  3. Missing Data. If a variable has missing data, there is no point in paying for an an analysis around it. There is no refund on bad variables. While a part of modeling is understanding if missing data actually yields some level of predicability if it's inherent to customer's choice, in general, if the majority of data is missing, it will not be value-added. 

Number of Records

The number of records are how many rows or line items are in your data. The more data the better for predictability and insights, however, not all data is applicable to your model.

  1. Time Horizon. Reduce your dataset by a time period that is truly reflective of your business. As an example, if your business has been around for the last 10 years, but you changed your business model or products in the last 4 years, you can remove 6 years of data if it's unlikely to be reflective of the future. Nevertheless, be cautious and thoughtful in removing data.
     

  2. Products. Remove products that you no longer offer or that is not reflective of the business segment you are trying to predict. If you are trying to predict the default rate of used car loans, you'll save some money by not included your new car business. Removing any kind of products that are test or trial products will also help reduce cost. Another example is location, exclude international data if the model is a US based model. 
     

  3. Missing Data. Do not excluding missing from records. This may reduce your cost, but has a material impact to the predictability of your model. Variables with high missing data is different than records that have missing data.

Number of Datasets

Reduce your cost by having your IT team merge datasets prior to transfer. As an example, if there are three datasets in which 2 of the datasets can be easily merged without losing data, merge the two datasets with the primary dataset and save some money! Be cautious to not create duplications or truncate your data in the process.

Always think about the goal that is trying to be achieved. Most database admins will be aware of risks, and if you're concerned, we've got it under control.

The first dataset is at no incremental cost.

  1. Merge. Merge your datasets into one if there are common elements to join. As noted above, be cautious on the joins. If you have 4 datasets that can be merged into two, you're saving the cost of 2 datasets.
     

  2. Database Connections. There is a 5% discount when datasets are in an approved database environments where direct connectivity is available.   

 

Terms and Conditions

Note that these are general terms and conditions and may vary from client to client and project to project.

  1. Pricing as defined is set by the number of variables, rows, datasets, and models. The price for each has a discounted rate for each incremental respective asset. One dataset is included. The starting cost per variable and its respective discounts are available in the table below.
     

  2. Additional cost may be applied if new data is presented after the modeling processes has started. 
     

  3. Forty percent of payment is due once the scope is defined. The remaining 60% is due upon completion of model and team sign-off. 
     

  4. Based on the data provided, models can yield no predictability. This does not constitute a refund.
     

  5. The modeling process and limits in which the model can be used will be discussed, Analytical Advantage is not responsible for using the model outside of its intended use and cannot be held liable for any outcome or utilization of the model.
     

  6. Data that has PII information cannot be transferred via email or non-secure measures, if client chooses to send data over unsecured means, the client is liable for any breach of the data. 
     

  7. This product does not include any future scoring, updates, or validation of models.

Discount Table
* Note that all discounts are compounded.
 

© 2020 Analytical Advantage, LLC

All Rights Reserved