Last week was our first ever Marketing Tech Friday post. We had a lot of fun researching and writing about tag management. If you’re interested, you can find that post here.
Also, don’t forget to let us know what you want to learn about next week! You can tweet us here or find us on either LinkedIn or Facebook.
Like we promised last week, we are covering a topic you are interested in. You tweeted at us, and we listened.
[hr style=”3″ margin=”40px 0px 40px 0px”]
You see predictive analytics a lot. A lot of sites use it, like Target, Amazon, and even Netflix. You probably see the results of predictive analytics on a daily basis, without even realizing it.
Most predictive analytics programs, which are much cheaper than hiring pricey data scientists, start by linking with a company’s internal CRM and marketing automation data. From here they add in data from thousands of public sources, such as company revenue and income, number of employees, and number/location of offices, to name just a few.
Once these programs have this data, they use data science to identify common characteristics of the accounts that were won by your sales team, and then predict the likelihood of closing each prospect. For example, a good signal for an equipment dealership to contact a prospect may be when they bid for a new account or project.
[hr style=”3″ margin=”40px 0px 40px 0px”]
In short, uplift modeling is the output of the predictive analytics process. It is used to predict the change in purchase probability, attrition probability, and spend level or risk that results from a marketing action or changing some aspect of the service the customer receives.
In other words: uplift modeling predicts the change in behavior that results when someone is subject to an influence. Mathematically, it looks like this:
In terms of retention, the state-of-the-art approach to customer retention is to predict which customers are at risk of attrition. You can then target those at high risk who are also of high value with some kind of retention campaign. Uplift models are often really successful at identifying the people who can be saved by retention activity.
In terms of cross-sell and up-sell lead generation, uplift modeling has proven to be particularly effective in high-value financial products. These purchase rates are often low, and the overall incremental impact of campaigns is often small.
[hr style=”3″ margin=”40px 0px 40px 0px”]
It depends on the company.
Predictive analytics requires less time and resources to implement than does uplift modelling. For years, the only companies with access to the data and capable of employing data scientists were the largest B2B companies, i.e. IBM.
There are many companies that are now providing cloud-based B2B predictive analytics services for a fee. They employ the data scientists, they add in the external and internal data sources, and they determine the analytics output that is useful to you.
If you’re interested, here are a couple of predictive analytics companies that we came across in our research.
Uplift modeling requires a lot of time, is narrower in scope, and can easily be misinterpreted. One of the biggest issues is that valid control groups have to be maintained. They also need to be of a reasonable enough size to accommodate a full range of potential actions. And this can be difficult even for larger organizations.
There is also evidence that uplift models need to be refreshed more frequently than other, more conventional models. There are clear clases where either data volumes are not adequate to support uplift modeling or where the results of uplift models are not different from conventional models.
So that’s the rundown. Now that you have seen our research and analysis, you have to make your own decision about whether or not these tools are right your business.
We want to know which marketing technology you want to learn about next! Just tweet at us or comment on LinkedIn or Facebook. You never know, your suggestion may just be the subject of our next Marketing Tech Friday article!
