This is a tough time for agencies with client work drying up. In order to make the most of the economic recovery (when it arrives) why not use some of the downtime to get some training for your team?

Below are some of the areas where I can deliver training courses. All courses will be delivered virtually through Zoom or Hangouts.

R for Web Analysts

This is my most popular introductory course.

It takes complete beginners from zero to being able to query the Google Analytics API from R.

After the course participants will be able to start automating reporting and be in a strong position to learn more advanced analytic methods using R.

The material on this course takes three to four hours to get through. I can do this as a single half day session but it works better to split into hour long chunks so there is time for people do digest the learning and come up with questions between sessions.

You can see a sample of the course materials here.

Effective use of Google Ads Scripts is a very important part of modern paid search management.

But many practicioners don’t yet have the skills to write their own scripts. Editing and making small customisations to someone elses scripts is very useful but it is limited compared to what you can do by writing your own scripts.

This training course comes in two parts:

  1. An introduction to scripting which takes two to three hours
  2. A follow up session two or three weeks later when we will work through some of the problems and challenges that students have had trying to apply the lessons from part 1

Model Based Forecasting

The “model” in “model based forecasting” means that you have a mental model about how the world works which you use to make a forecast. Weather forecasts are an everday example of this; meterologists have a model of how tomorrow’s weather will evolve from today’s weather and how the weather in different places will influence the weather at the forecast location.

This course goes through common forecastings models, how to tell which one to use in a given situation and how to customise the models further to include extra business specific knowledge.

For a group that know a little R or python already this course takes about three hours in total. Splitting it into three one hour sessions will make the material easier to digest.

For those who are complete beginners when it comes to programming I can run an accelerated version of my introduction to R course which will cover only the bits needed for forecastings. This will add an extra two hours to the total time.

The last set of students I had through this course were able to develop a future traffic forecast that was 17% more accurate than the predictive forecasting models available in Microsoft’s Power BI.

For more about model based forecasting Rober Hyndman has a good explanation for why it is important on his Hyndsight blog.

Introduction to Stan

Many things about the world can be modelled as stochastic processes; something with a bit of randomness.

For example, in the chart below y is equal to 2x with a bit of random variation added.

If we have the data (the y and x values) and our stochastic model for how the data were generated we might ask questions like:

Stan is a clever tool for figuring out estimates of model parameters from data that was generated by the stochastic model. It allows you to specify highly customised models without (for the most part) worrying about how to fit them.

If neural nets and deep learning are the revolutionary tools of the 2020’s for learning from big data then the techniques enabled by Stan are just as useful for “small” (or everyday sized) data.

This course is an excellent companion to Model Based Forecasting above