First of all I do not have a mathematics degree only a B.S. in finance so please take that into account when writing an answer. Generally what type of mathematics is involved here? And specifically what statistical formulas can be used in a scenario like this?

Recently Target was able to predict that a teen girl was pregnant by analyzing the items she had purchased, and sent her the appropriate coupons for her current condition. I would like to know broadly how were they able to do this, and specifically what types of mathematical formulas they used/ can be used to do this. This link will describe the specific situation.

  • 6
    $\begingroup$ Did you bother with the tag excerpts that people worked so hard to have here? [model-theory] is one of the furthest things from statistics and "modeling" that I can imagine. $\endgroup$ – Asaf Karagila Mar 2 '12 at 0:06
  • 4
    $\begingroup$ The specifics are probably private corporate information. Also, it's a probabilistic gambit, not a logical deduction on the part of Target's system (FYI). $\endgroup$ – anon Mar 2 '12 at 0:09
  • 2
    $\begingroup$ Sounds like an exercise in data mining to me, and would probably fall under the heading of "market basket analysis": en.wikipedia.org/wiki/Market_basket_analysis $\endgroup$ – ItsNotObvious Mar 2 '12 at 0:10
  • 2
    $\begingroup$ Data mining, machine learning, many things. I think this question should be moved to stackoverflow. Otherwise, the question is too broad & should be closed. $\endgroup$ – user2468 Mar 2 '12 at 1:46
  • 2
    $\begingroup$ I suspect its a form of insider knowledge that target used here. The store is clearly the father. $\endgroup$ – Johannes Mar 6 '12 at 12:35

Target's approach is everyday, ordinary statistics. They have a list of products that are purchased by expectant mothers and know the significance of each "factor" (ie product purchase) towards indicating pregnancy. These are used to calculate a percent chance that the buyer is pregnant. Once that percentage exceeds a human-chosen threshold, they start sending them coupons and advertisements for baby stuff.

The products which expectant mothers tend to buy are unscented lotions, soaps, sanitizers, washcloths, cotton balls and containers (like extra large purses). If someone starts buying these things for the first time or in increasing amounts, each instance is counted as a factor.


Target Inc's data science team developed a customer loyalty system using a machine learning technique called Bayesian classification network that represent conditional dependencies against sets of random variables. In this case Target developed a strategy to understand customer behaviour and shopping habits and encourage customers to try products similar to their usual brand purchases. Target's CRM data warehouse contain customer engagement key figures with a brand dimension are fed to Target's Customer Loyalty system to answer for the best method to communicate with customers for brand experiences. Target's CL system was able to mine the teenager's shopping baskets and segmented her as a potential pregnant mother, therefore sending her baby coupons which is her preferred communication method. This ML technique is also used in Google Mail to filter out spam mail.

  • $\begingroup$ Target's Customer Loyalty Machine Learning is not a Statistical Tool like SAS or Excel spreadsheet. SAS or Excel require manual explicit programming to process and visualize reports from which the user derives patterns, trends and outcomes. Machine Learning computer programs are said to learn from experience with respect to some class of tasks and performance measure, if its performance at tasks as measured by performance measure, improves with the experience. Statistics tools do not learn from running statistical tasks, therefore, their performance do not improve over repeated usage. $\endgroup$ – James K Chau Apr 29 '17 at 20:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.