Posts

Predictive Analytics Answer the Great Thanksgiving Brussel Sprout Controversy

Predictive Analytics Answer the Great Thanksgiving Brussels Sprout Controversy

Given the current pandemic, we can all expect our Thanksgiving to look a bit different this year. There is of course one constant no matter the challenges being faced; turkeys are going to be in high demand. Over 45 million turkeys are consumed each year during the Thanksgiving holiday season. This means, we as a nation, collectively chow down on about 730 million pounds of delicious turkey every Thanksgiving. A truly mind-boggling amount!

Now, we all know everyone will be eating turkey but what about the truly fundamental and critical question posed each Thanksgiving, should we serve brussels sprouts? I am sure most of you can hear your kids screaming, NO WAY, but amazingly Americans have a love-hate relationship with this round, green cruciferous vegetable and consume $4.3 million worth each holiday.

So how do we go about making this critical choice and potentially avoid family strife? We could send out a survey to all our attendees, listen to our kids, OR we could use a predictive analytics software like Squark! Below outlines a step-by-step view of how you would find the answer to the vexing question and how easy it really is!

Lets use a dataset (see appendix 1) provided by fivethirtyeight.com to find out the answer to this vexing question in less time than it takes you to clear your first plate of food. 

  1. First, let’s grab the data using our intuitive Google Sheets wizard:
Predictive Analytics Answer the Great Thanksgiving Brussel Sprout Controversy - Squark's Google Sheets Wizard

2. Once we have the data, Squark will perform some initial analysis. Once completed, let’s tell the system which column contains the historical data that answers our question of “do you serve brussels sprouts?” so our system can learn:

Predictive Analytics Answer the Great Thanksgiving Brussel Sprout Controversy - Upload your training file

3. Select which columns you would like to include (hint: throw them all in at first and make choices later once the machine has had a chance to explain to you what is most important to the prediction you are making):

Predictive Analytics Answer the Great Thanksgiving Brussel Sprout Controversy - Include all relevant data to increase predictive accuracy

4. Now grab a file from your computer with all of your relatives names, along with their associated data to match what we are training on to determine who loves brussels sprouts but would never admit it:

Predictive Analytics Answer the Great Thanksgiving Brussel Sprout Controversy - Easily upload your production file

5. Click “Start Training” and let the machine figure out not only who is going to be sneaking some brussels sprouts in the corner, but what traits or attributes are most likely to identify a true brussel lover.

This is just a snippet of the attributes we examined but we can see those from the West North Central and West South Central regions are highly likely to go straight for this green veggie. And it goes without saying that if you are a fan of canned cranberry sauce, you of course love brussels sprouts. Predictive analytics is this simple.

Now that we know who is likely to be the first in line, lets export our results to Excel and see who in the family is a secret brussels eater.

I always suspected Mom and Dad!

Amazing, another mystery solved in just minutes. All with the information you have right at your fingertips. No code required. Even better, out of the same data you just utilized, you could even find who is likely to steal the last slice of pecan pie!

Now imagine what you could do by applying the same capability to any range of marketing related questions you would like to answer. All within minutes, with the data you look at every day. Reach out at any time and let us show you how predictive analytics can maximize your efforts to increase conversions, revenue, customer retention, and so much more!

HAPPY THANKSGIVING EVERYONE!

3 reason why elections are hard to predict for humans and machines

3 Reasons Why Elections Are Hard to Predict for Humans and Machines

What if a model said no one will win the election? That model would be rejected immediately, but what if it wasn’t wrong?  What if the learning picked up a signal from historic data that knew the election would be contested and no one wins?  That’s not an unprecedented outcome when a “hanging chad” is a known feature. Perhaps the model predicts the US courts will decide. Plausible. Probably unlikely.  Someone always wins eventually so what’s the ground truth – was the model right or wrong? 

Perhaps the wrong question was asked; are election outcomes really binomial?  Inherently with more than two political parties running, the results aren’t binary- they win or not. Elections can be  truly a multinomial classification; however, we rarely treat elections as multinomial, but rather usually binomial.  Right Mr. Nader?  Right Mrs. Stein?  Thousands of Markov chain simulations are cool on a random walk to the polls but there’s always the drunk leaning on the lamppost. 

What causes bad predictions and why is it difficult for not only humans but machines to predict the outcome?  So many reasons beyond the scope of this blogivation, which we will tell you in Squark Symmetry, but here are a few to get started:

  • Bad data.  People lie.  Citizens do it to pollsters who inform politicians who lie to citizens about the results; it’s an unvirtous cycle, for sure.  A strong empirical indicator that the current US President will win again is when you ask Americans who they will vote for they say Biden, but when you ask the same American who their neighbors will vote for, they say Trump.  This of course varies by geography, but it’s the same trend seen in 2016. 

    I’m told people are cautious to announce their support for Trump or Biden, and don’t want to disclose that they align with a particular politician.  The reasons are as myriad as snowflakes.  With the models ONLY being as good as the data used, predicting election outcomes – within unprecedented times with viral externalities and the potential for foreign interference  –  means new features harder to engineer.  Feature named “Pre-Covid” yes or no may become the “new normal” column for machine learning. 

  • Skewed data.  It is well known candidates increase their positive polling trends directly after a political convention. Post-convention bounces regress to the mean (or to the moon) but still hold perhaps some predictive power.  But are these candidate bounces just dead cat bounces? When trying to predict November are post-convention polling increases mostly meaningless earlier in the year?  Seems that way. The polling data can be skewed by bias around the convention based on who was sampled in the data frame. Coverage error, selection bias, and so on.

  • Overfitting. Which means the results of the analysis match too closely to the data the machine trained on, and thus fails to fit new data in a reliable way.  It’s a problem in prediction, and it’s not always easy to work around.  One way to do it is to early stop models when they begin to degrade and get erroneous. Other methods exist too, but these are complex math that aren’t always obvious or even known to data people. It helps when someone has written the code to do this work for you, like Squark.

There are so many other reasons why election probabilities are just that, probabilities, that may or may not happen.  People fail to clean data; they overclean data, they don’t use enough observations.  Sometimes analysts, despite best intentions, just don’t know what they don’t know. Recently on LinkedIn I saw some “expert” with an agency produce an inflammatory Covid related correlation analysis, and the guy didn’t even make the data stationary.  Doh!

The cool thing about Squark is we know this and a lot more too.  When we invented no code predictive analytics, we created a powerful entire automation, with deep tech sub-automations built it, that help reduce and work around the programmatically solvable issues above and many more. 

We started with a vision to democratize prediction for business users by putting an advanced AI capability into their hands, without coding.  We’ve done it now, and we want it to be affordable and powerful, so everyone can use it, even politicians. That is, if we would actually sell it to them.  But like Jack Nicholas said in that Tom Cruise movie, perhaps they “can’t handle the truth!” 

Please don’t forget to vote wherever you are, whenever you can!

Predictive analytics helps to drive results in uncertain market conditions

Predictive analytics helps drive results in uncertain market conditions

Nowadays, marketers are having to deal with extreme uncertainty due to the impact that Covid-19 has had on the majority of industries. Strategies are reevaluated and plans are pivoted over and over again. With the majority of companies’ marketing efforts now being digital, marketers are facing a huge influx of data collection across multiple channels.

In times of uncertainty such as these, scenario planning is critical to combat the uncertainties of the future. The most crucial ingredient for effectively planning scenarios is knowing how to use your data. When used correctly, data can offer a doorway to current trends, it can help avoid costly errors in decision making, and aids in charting a way forward. The obstacle most companies are facing is that data can be challenging to collect and maintain (we’ve all seen the CRM with dirty data). It’s usually with the help of technical experts such as data scientists, that companies are able to extract insights from their data and create accurate predictions. 

If you don’t have a team of data scientists, not to worry, because no code predictive analytics is bridging that gap and helping marketers understand their data to help craft and design winnings strategies during these uncertain times. 

Here are 3 ways no-code predictive analytics is navigating the ‘COVID’ era:

1. Responding to Ever-Changing Customer Preferences

Marketing departments are seeing major changes in consumer behavior. Changes in the way consumers interact with brands, changes in preferences and purchasing habits and so on. Some of these changes in behavior could be temporary due to current circumstances. Others, however, may reveal themselves to be long-lasting. 

As a result of these constant changes in customer behavior, our data is also changing and adapting. Predictive models created just a few months ago, are no longer accurate and have to be updated manually by data science teams. Updating models could take months which could lead to a never-ending cycle of outdated insights. No code predictive analytics is solving this problem by allowing marketers to generate the most advanced models and predictions in a matter of minutes. This gives teams the advantage of predicting data as fast as it changes. Ultimately, meeting our consumer’s preferences and building loyalty in the nick of time.

2. Allowing for Personalization at each Touchpoint

The customer’s journey has changed due to this current pandemic. It’s become important to meet customers’ new expectations and ways of interactions. However, touch-points throughout the journey still remain to be an opportunity to engage with the customer and align resources more quickly. In order to reach potential customers at various touch-points, marketers will need to use the influx of new data to better personalize messages and offers. By adding predictive analytics, marketers can transform data into insights to essentially personalize services, offers, and products to their audience, outmatching the competition. This helps guide customer interactions throughout the buyer’s journey and lead to more conversions.  

3. Efficient Budgeting

No code predictive analytics is not only useful when trying to predict a future outcome such as, “when is my customer going to buy from me next?” or “which service is my client most likely to buy?” No-code predictive analytics software such as Squark will tell you why these outcomes will occur. By incorporating explainable artificial intelligence (XAI), Squark is going to tell you what the main drivers are for the predictions. This power of explainability can help guide some data-backed attribution modeling. AI decisioning becomes transparent for business users.

For example, you could ask, are more people interacting with our boosted social media posts or should we focus on webinars to help drive sales? If more people in the sales process have been associated with certain programs along their buying journey than using predictive analytics can uncover which would produce the most sales. Knowing how to attribute marketing efforts, budget expenditure can be allocated more wisely to channels that can generate a better return. As you prepare for the next quarter, forecasting necessary budgets has never been easier than doing so with predictive analytics.

These are only a few ways in which no code predictive analytics is helping companies navigate and drive results, even during these uncertain times. Marketers and business leaders can take advantage of the new digital space we are entering by being early adopters. Data doesn’t need to intimidate you and you don’t need an army of data scientists. Early adopters of no code predictive analytics will pave the way and have the advantage over competitors by constantly driving results. Read more about how Squark is helping marketers succeed.

Predictive Analytics

Statistical techniques gathered from predictive modeling, machine learning, and data mining that analyze current and historical facts to make predictions about future or otherwise unknown events.