What's in a prediction?

21 Dec 2015 | Vlad Mehakovic

What's in a prediction?

As marketers we are inherently making predictions all the time. When we launch a campaign, brand, website or social ad we are predicting that it will have an impact on some business objective.

Our issue, as marketers, is these predictions are often implicit. How much will the website raise engagement, when by? How sure are we that we're correct?

In his groundbreaking book, Superforecasting, Phil Tetlock uncovers what the world's best forecasters have in common. He's done this after running forecasting competitions for the last 30 years. Using his forecasting competition results, Tetlock was able to beat government intelligence agencies (CIA and others) in Intelligence Advanced Research Projects Activity's (IARPA) forecasting competition - not bad.

Some key insights from the book that I'm taking to change my marketing practices:

Make specific, time-bound predictions e.g. "I am 80% confident that this change will increase daily conversion rates by 20% within a month"
Unpack insurmountable forecasts into component parts ala Fermi
Start with the outside (macro) view e.g. to estimate if a couple will stay together don't look at their wedding photos, look at the divorce statistics
Then adopt the inside view (micro), collect as much opposing information as you can, clearly state your assumptions
Get to an initial, informed forecast as early as possible based on existing information
Update the forecast by small increments when relevant new information presents itself
Review your forecast after the fact, even if you're right assess if it was a close call and understand why.

There's plenty more I haven't covered, needless to say the methodologies in the book are a powerful tool to add to your arsenal.

1. Twitter aims to show advertising to much wider audience

Channel

Twitter plans to start making money from the vast number of people who are not signed in to the service but may click on a tweet they find through a web search or that is sent to them via email or text — an audience Twitter estimates at more than 500 million people worldwide.

2. From quacks to quaaludes: three centuries of drug advertising

History

A brilliant look back at the history of one of the most profitable businesses of all time and its reliance on advertising to sell.

3. The state of technology at the end of 2015

Trends

This post explores the next epoch of computing, argues that it involves communications connected to data and stipulates that the best company to capitalise on it is Slack.

4. Awesome ad tech: what you get when Jeff Bezos buys a newspaper

Channels

The Washington Post continues to debut innovative ad products after focused, consolidated investment into ad-tech R&D. Their latest release, "PostPulse," is a new ad unit that leans what kind of stories readers seem to enjoy, based on their history. Sounds like Amazon product recommendations right?

5. What the KGB can teach you about data-driven marketing

Data-informed

It turns out that the KGB were pretty good at data-mining well before it was a cliche. The article covers some lessons we can all take from their efforts.

6. What’s your true customer lifetime value (LTV)

Data informed

LTV can become infinite, which clearly doesn’t reflect reality. This post offers a new way to calculate LTV based on discounted cash flow analysis that takes into account the risks associated with revenue that is far off in the future, and the time value of money.

Emotion arising
Predictive analytics personalising
Metrics to action realigning
Value-as-a-service emerging
Employee CX & empathy training

10. CrowdFlower

Product of the week

Data enrichment as a service. Your data can be text, images, URLs, business addresses, or whatever else you need enriched and CrowdFlower adds context.

What's in a prediction?