Thank you for registering for our talk, How To Build a Content Calendar For An Entire Year With Predictive Analytics. Your materials are below.

Video

Can’t see anything? Watch it on YouTube here.

Audio

Listen to the audio here:

Download the MP3 audio here.

Slides

Build an Entire Content Calendar with Predictive Analytics PDF (6.2 MB)

Machine-Generated Transcript

What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for watching the video.

Welcome to predictive analytics for social media marketers a recipe for success.

My name is Christopher Penn, co founder and chief data scientist here at Trust Insights, a data detective firm, helping you find answers in your data.

So let’s get started.

We’ll bypass the normal stuff.

Today, we’re talking about social media analytics, predictive analytics to help you build more impactful and more effective social media.

And the most important thing that I want you to take away from this session if you if you had no other time for anything else is this analytics without action is a recipe your cook, but you never eat.

That is the most important thing to learn and to know, all the analysis, all the data, all the insights don’t matter one bit, unless you do something about it unless you do something with the data and bring it to life without it.

It’s just a decoration, right? It’s just another binder on the shelf.

Another thing that you have and maybe it feels good To have it but it doesn’t bring impact to the business.

So it’s so important that you need to have this analysis that leads to action.

So let’s start off with this.

What is this predictive analytics stuff? You heard the term a bunch of times usually mentioned with terms like data science and machine learning and artificial intelligence.

What is what is this? predictive analytics is exactly what it sounds like.

It’s using analytics to predict things to predict stuff.

And the reason that this works is because we humans are predictable, right? An individual may not necessarily be completely predictable, but as groups of people, we are very, very predictable.

We do very predictable things.

And predictive analytics is looking at our behaviors.

Looking at our what we do over time, the seasonality, cyclicality trends and then 14 Casting ahead, what’s likely to happen? Now, predictive analytics, we’ve been doing this for a really long time as marketers, right? We know when Christmas is, for example, it’s going to be December 25 of every year.

I know there’s some holidays, for example, that like Diwali, or Hanukkah, where the calendar date may move, but you know that it’s going to happen.

So you can predict it.

Predicting stuff, in general is very, very easy, right? We know that there are times of the year when people are going to be less busy and more busy.

What we don’t know.

And we’re predictive analytics has its value is predicting specifically what day what week is something likely to happen so that we can plan ahead is that specificity that makes all the difference between Yep, something’s going to be happening.

Or we know Fourth of July is here or Independence Day is here.

Or the Buddha’s birthday is here.

Yeah, you know those things, but why? What else can you predict? And how specific can you be so that you can time things for exactly the right time for your marketing have most impact? Fundamentally, predictive analytics is about using the past to predict the future.

And we do this with math, math, the mathematics of statistics and probability to take past data, look at the variations and trends and then see where that trend is going.

The most common algorithm for this is what’s called auto regressive integrated moving averages with seasonality and some of you’re like, well stop the train we got enough.

It’s okay.

You don’t need to dig super hard into the math, right.

You don’t need to have this formula memorized to be able to understand the mathematics about it in order to do predictive analytics, but you do need to know that it is fundamentally about probability, the probability that something is going to happen and the reason that simple is because there are some things you can’t predict.

Right? There are certain events called, you know, Black Swans or anomalies or whatever, which there’s no way you can forecast them.

Predictive Analytics works best on things that are predictable things that you know are likely to happen.

The the textbook example that I often give when I’m doing this as a keynote is nobody knew.

No one could have predicted on September 10 2001 was going to happen the next day, it was an anomaly, right? There’s no way to predict that no one could have accurately predicted the results of the last presidential election or when the UK was going to vote on on the Brexit referendum or any of these things that are one offs, you cannot predict what is unpredictable prediction is all probability.

So understanding that has a basis in math takes the magic and the mystery out of it, and converts it into something we can understand it is math.

So why why do we care? about this one.

Yeah, it’s cool to know that how to how to plan ahead of things.

Why is this important? A couple of reasons.

In the most recent Pricewaterhouse CEO survey, CEOs were asked a question, thinking about the data that you personally used to make decisions about the long term success and durability of your business, how important the following.

And the four most important responses here are all about data and our ability to act on it data about your customers and clients this preference in these financial forecasts and predictions, data about your brand and reputation and data about the risks to which your businesses expose CEOs want to know what’s likely to happen so that they can plan so that they can mitigate risk so that they can increase results in the most recent cmo survey chief marketing officers.

It’s interesting to look at how wildly variable the use of data and analytics is right.

People like I don’t know if I should do this is not a big reason for that.

Because going back to what we said the beginning analytics without action is a meal you cook but never eat.

Right? When we hand just data to a CMO, they don’t know what to do with it.

They’re like, this is great.

So nice chart, but they don’t take action on it.

Right? predictive analytics is all about taking the guesswork out of as much as possible so that we can deliver better results so we can focus our efforts.

So we can achieve operational efficiencies, which is code for do more with less.

Back to the CEO server when they were asked, which of the following activities if any, are you planning in the next 12 months in order to drive revenue growth, the number one was operational efficiencies, which means budget cuts, right? do more with less.

Predictive Analytics helps us do more with less by helping us understand what’s likely to happen so that we don’t plan to spend money when It’s going to be least effective.

We don’t plan campaigns that are going to not resonate because our audience isn’t paying attention.

Remember, analysis without action is a recipe cook, but every, that’s what we want.

That’s why predictive analytics is so important, because if we do it well, we will get people to take action.

So how do we get started with this? How do we get start with predictive analytics? predictive analytics, and analytics in general, is just like cooking.

Right? Cooking is a very straightforward process.

Everybody understands it.

First question, who we cooking for? And why? Right? We need to know those questions to answer those questions.

So that we’re putting together the right types of predictions.

Imagine you have this lovely org chart, right? Do the people at the top of this org chart need the same type of prediction the same type of forecast that the people at the bottom do of course not be silly, and yet, when marketers use ad analytics predictive or otherwise, they have a tendency to try and do like a one size fits all.

That’s never the case.

When you look at this an organization, you divided up into levels and divisions, every one of these boxes in your org chart probably needs its own type of predictive analytics forecast.

And the way you get out that the way you you figure out who needs what is by doing what are called user stories.

This is a concept that my CEO Katie robear, did for about 10 years in software development.

And she says user story is simply a who, what, why statement that helps outline the needs of somebody when you’re building software.

Well, predictive analytics and machine learning and data science, essentially, as a lot of a lot of ways is like software building.

So if we fill out user stories will have a much better sense of what types of predictive analytics we’re going to be doing.

For example, we if you ask somebody, why do you want to predict the forecast, and the CMO says as cmo who I can see big picture trends, but what so that I can plan my overall marketing strategy that tells us a lot about what needs to be in this person’s predictive analytics forecast.

Right? Take it down a step as marketing manager, I can anticipate periods of high demands that I can plan staffing and resourcing appropriately.

Right? This could be a very different kind of forecast.

Much more granular on timing, but maybe not necessarily on on volume.

And if you are a social media marketer, so I’m marketing specialist, as a marketing specialist, I can know what topics are important weekly, so I can create timely content.

And you see what the CMO needs and what the marketing specialist needs are very, very different.

So when it comes to predictive analytics, we want to make sure that we are collecting these user stories in advance so that we are building the right kinds of forecast.

So that’s step one, know who you’re cooking for.

Next, what’s the recipe how are we going to cook this thing? We look at a recipe recipes are actually really straightforward, right? There’s sort of a what is the outcome you’re after that lovely picture at the top there.

What goes in the recipe? And how do you do the thing? and probably some equipment, right? If we refer to our predictive user stories, we can really start getting into understanding what needs to go in the recipe has Social Media Manager, I can know when people are in the office so that I can schedule the most important content the right time.

So let’s do this as a case study, right? Let’s walk through the process of building a predictive forecast.

We’ll start with this user story, the Social Media Manager.

The social media manager wants to know when are people in the office so that they can schedule content, probably the most impactful like a recipe we’ve we know, we know roughly what we’re doing.

Let’s find the ingredients.

What kinds of stuff do you have in your pantry? When we’re building predictive forecasts? We’re talking about data, right? We need that data to forecast from in order to make our predictions.

What do you use to predict this is the broad data You need numbers, and dates.

To create predictable data and you need a fair amount of it.

The general rule of thumb is, for every one period, you want to predict, you need four periods of data to look back on.

more is better.

So if you want to predict a year in advance, you need four years of back data to reliably account for trends and seasonality and stuff.

If you want to predict seven days advance, you need 28 days of pass data to predict for it.

So numbers and dates are what you need to do prediction, but you need to have that four to one ratio.

The good news is as social media marketers as digital marketers as marketers, we are have no shortage of numbers and dates, right? We have search data, what people type into Google, by day, we have social media data, tons of social media data.

We have email data, we have web analytics data, we have ad performance day we have YouTube data, podcasting data, you name it.

We’ve got tons of it.

So we’re not running short on numbers that have dates attached to them, we have to make sure that they’re, they’re in good condition.

So that’s our ingredients numbers with dates attached to them.

Next, we need to have something to do the prediction with.

And this is where you’re not gonna like this section.

The reason why is that you’re going to need some very sophisticated tools to do the actual math.

Remember I said earlier, you don’t need to know the mathematical formulas, you need to have the right tools that can do the math for you.

So let’s look at a couple of these.

If you want something that is a comprehensive tool that is enterprise grade, and enterprise pricing sometimes and has a drag and drop sort of environment for assembling the statistical needs, the statistical algorithms you need to do predictions.

IBM Watson Studio is a great choice for disclosure, my company Trust Insights and IBM business partner.

Anything you buy through IBM through us creates a potential indirect financial benefit to me.

So FTC disclosure, by the way, Watson Studio is a good environment for doing pretty good stock time series forecasts, predictive analytics.

However, the best stuff, you need to learn how to code or you need to have access to developers or coders that can write in one of these two languages, and a data scientist to help you with the algorithm in order to get the best quality predictions.

Now you’re saying, Oh, isn’t there something I can just get off the shelf? Isn’t there a SAS based software company out there that just does this for me? Not really.

Not really the I’ll caveat but not really, with any of that I have interacted with that are doing a more credible job than the sets of tools that I just mentioned, the sets of tools I just mentioned, particularly these open source libraries, languages are in Python.

They don’t cost money, they cost knowledge and time.

A lot of the SAS software companies out there cost you money.

And their quality is iffy.

I was talking to a vendor at a marketing technology show.

And they had all this great stuff about their predictive analytics.

And I pulled their engineer aside.

I said, so what are you using are using this S-ARIMA method, or using LSTMs, which is a type of machine learning and and I’ve named a few of the technologies and the person said, very candidly, yeah, we’re using recurrent neural networks.

And the engineer said, I got to tell you don’t tell anybody.

Don’t tell the sales guys at the booth.

But we’re having a real problem getting much beyond eight days of the algorithm so just not keeping up.

And that made me really stop like we’re talking about here.

You got a booth and got sales guys promising you know, the world world to people.

And then you’re telling me behind the scenes, you can’t resolve more than eight days where the data without your predictions falling apart.

That really made me hesitate.

like, Okay, you know what there is, in terms of recommendations like why is there anything off the shelf? Well, there it is.

I don’t know that I would trust it.

I would rather I’d rather and I would rather you invest the time and effort into either using Watson Studio or using our using Python as languages.

So that’s the tools you need.

And again, there’s not a lot that I think is super awesome right off the shelf.

Right.

But there are a lot of super awesome tools that you can use to cook with.

But you it’s, it’s kind of like you can’t buy the frying pan but you can go to this lovely Ironforge and make your own frying pan.

Takes a little more expertise takes a little more time.

But in the end it will be yours and will have your secret sauce embedded in it which is awesome.

So we got the recipe we know who we’re cooking for.

We’ve got our ingredients.

We’ve got our utensils, time.

Do some preparation, we’ve got to prepare our data.

When we’re doing predictive analytics, remember, we’re going to start remember that user story.

I’m a social media manager, I want to know when people are going to be in the office so that I know when to time my campaigns.

That’s our use case we’re working against.

A big part of analytics, a big part of data science, is understanding humans, I’m the same people.

If we can forecast specific words or phrases that could indicate whether somebody was in or out of the office based on their behavior, we get some very robust forecasts.

So what’s something that people would talk about what something people would ask? I love search data.

I love search data, because even though we’re talking about social media, people will type things into Google.

They will not say to other people certainly not say out loud and they definitely won’t post on public social media.

There’s a great book a few years ago called everybody lies all the questions that people ask Google that, again, they wouldn’t say in person.

So what could we, what could we ask? What can we find out what other people ask? Well, here’s a really good example, out of office, when people search for Outlook, out of office or out of office message or Gmail out of office, what are they asking? They’re asking, How do I type turn this feature on so I can do what, go on vacation, leave the office.

So we could understand all the different ways people could could ask Google that question, and then take the data from that, from these hundreds of variations.

That would be a robust data set to understand when are people going to ask these things when people looking for these things.

And if you look, here, this five year forecasts a five year back data from Google Trends.

You can see there’s very clear seasonality about when people are not going to be in the office.

So this is our preparation.

We need to get this data out of search engines, maybe some of our social media monitoring tools, maybe Google Trends, wherever it is, we get numbers with dates attached to them about the behavior that we’re interested in.

That’s the preparation stage.

Now we got to cook.

And again, this is where you’re going to want to have those expert tools, maybe some expert assistance to use those tools.

This is an example from Watson Studio.

of taking that data we were just talking about the search data combined with Google’s trends data, which is the the purple data asset box and drag and drop, there’s no coding here is drag and drop these boxes together, saying I want to forecast the search volume.

And you can see the orange result box there says hey, the forecast is made.

And I want to be able to take that data and and work with it.

This is an example from the our environment.

Again, same idea.

Take the data in forecasts that search volume when will people be searching for out of office message or outlook out of office or Gmail out of office? When will they be searching for that the most because if we know that, if we know, when people are typing that into Google the most we know, they’re about to leave the office.

And we don’t want to be running majors paid social media campaigns to them.

We don’t want to be sending them emails, we don’t want to be trying to engage with them on Facebook, especially if they’re leaving the office, and they’re going on vacation.

The last thing they want is a marketer saying, hey, while you’re on vacation, think about work.

And think about buying some stuff.

Right? That’s again, knowing humans.

So these two examples, this is the cooking part.

And this is where that that expertise does come in handy to be able to run the software or build the model in a in a Watson Studio environment to get your forecast.

Now, what comes out of these tools isn’t the most friendly, right so part of What we have to do is, is take the the raw, first run at cooking and refine it, refine it, clean it up, make it more palatable in cooking parlance.

This is the test kitchen.

Right? We’ve, we know who we’re cooking for.

We’ve got the menu, got the recipe, got the ingredients, utensils, we’ve done our prep, we’ve cooked it now it’s Test Kitchen time.

What do we get? Well, that’s what comes out of Watson Studio, right? That’s not super helpful.

Right? It is technically correct.

But remember, analytics without action? This is why you wouldn’t take action.

You’re like, what the heck does this say? So we need to take it into a tool that can refine the data a little bit something like in this case, I’m using I’m using Tableau software.

But there are a bunch of other tools that can take this raw output and refine it.

Here looking at those general terms Gmail terms, Outlook travels, you can start to see the the week so in the next year, that graveyard line is the day that I ran this forecast, you can see the ups and downs of when people are going to be out of the office, when those lines are the highest.

That means people are going to be out on the blinds a little lowest these people are going to be in.

right because they’re searching less and less for those terms.

The graph is helpful.

The table is a little more helpful where we can see okay, what are the weeks when this term is going to be highest here in deset in q4, 2019, the week of December 15.

And the week of December 22.

Those weeks people are gone, gone, gone, gone.

They’re not around there, nobody’s home, don’t run major campaigns, looking out a couple of quarters week of June 21 and June 28 of 2020 people are gone on vacation out.

So we know not to market during those times.

flip side, Jet the week of January 12 and q1 Hey, people are in the office right there that search volume drops really low.

That’s the week that we want to to email them to run paid social campaigns to run retargeting campaigns to do influencer outreach, whatever we can do whatever levers we can pull to get our audience’s attention.

That’s the week that we want to do it.

So the final product in our test kitchen probably is something that looks like this.

Right and screenshot this because you you you can have this year, it’s our gift to you.

These are the weeks for your big social campaigns the week of January 12, in q1, April 26, and q2 September 30.

And q3, October 18 and q4 do not plan big campaigns for these weeks, March 8, and q1 June 21, and q2, July 5 in q3, and December 20 in q4.

Now, you may be saying, Well, duh, Chris, that’s pretty obvious.

July week, July 4 week of Christmas, don’t send campaigns there, at least if you’re in the United States.

Right, you knew that.

But have you done anything with that information? Have you taken action on it? said, as you’re building out your 2020, social calendar, this these little weeks where we’re just not going to do stuff, we’re going to do a limited version we’re going to do analytics without action is a meal you cook, but you never eat.

So this is sort of what should come out of the test kitchen, to go from this really, really confusing mess of data to more and more refined products until you just have this to hand to your marketing manager, your social media manager, your ad manager, your your cmo, to say this is what’s likely to happen, here’s what we need to do about it.

Now, this is the walkthrough for an example of how you would use this data to plan your campaigns.

What is a more refined, industry specific version look like? What What would it look like if we were to roll this out for something that was more specific to a certain line of business? One of the one of the examples I love to use when I’m talking about predictive analytics is cheese shops.

Jesus, right? Because people know what Jesus.

Some people like cheese, some people don’t.

But suppose you worked at a cheese shop, think of all the different cheeses that you could have.

And for those who who are Monty Python fans, that’s where we’re going with this Gouda iida case, Nespresso flip side chatter.

You name it.

They’re all these different cheeses have parties and goat cheese and fed us.

Wouldn’t it be great for you as a as a cheese shop marketer to know what cheese people are searching for each week, and be able to time your campaigns, your specials, your social media posts, your videos, your walkthroughs to the cheese that people care about that week.

So here’s an example of taking those cheeses and the questions people have about them and putting it in that exact same predictive analytics framework.

We’re going to repeat that whole process we went through Get the recipe, get your equipment, get your ingredients, prepare it, cook it, Test Kitchen it to see for see a social media manager, you look at this and go, okay, the week of January 1 2020, I need to have some social content about how many carbs and how many calories are in cream cheese.

Right? Of course, again, this is part of knowing the humans, we know the humans.

What’s happening in the beginning of the year, people are winding down from the holidays, people are suddenly really obsessed about the calories and carbs and their food.

That’s humanity, human nature.

So we want to build content for that week.

And then if you go out and just illustrate here, go out to like, end the March.

And now we’re setting into like how many calories and mozzarella what’s changing? Well, of course, is especially in certain parts of the world.

It’s getting warmer, it’s getting nicer People are grilling more.

Right? And so the type of cheese changes and the content you would create changes, right? Going out a little bit further down here.

How many calories in a bagel with cream cheese right, that goes up right on the third week of January and then comes up again in April, what’s happening in April.

And so some parts of the world it’s people thinking about vacationing and beach season and, and looking a certain way.

They’re going to care more about those things.

So the predictive analytics tells us the timing, remember the specificity The timing is so important if we know that’s going to be a thing beginning of the fifth week of the first week of April.

Boy, we better have our content ready? We better have our YouTube video scheduled if we’re a cheese shop.

Hey, like we got low calorie Jesus here.

We got cheeses with lots of that are nutrient dense.

So yeah, you’ll get some calories but it’s going to be healthier choices for you.

Here’s how to make a cheese lower calorie Hey, you can fry cheese and pour off the extra oil and have like cheese crackers that are mostly the proteins.

There’s all these different things that you could do as a social media manager with the data.

And what you see here is that we’re looking out if you use the tools, well, we’re looking out months in advance, predict predictive analytics.

You can refine this even further.

And maybe just take the top two or three questions each week.

And that’s your content.

This is something you could hand to your agency, your junior staffers, anybody that would really help them out like okay, this is what you need to focus on this week, the week of December 15.

How many ounces of cream cheese in a cup of what are people doing? Hey, they’re making stuff for the holidays.

help people with content that that addresses their needs, meet them where they are.

This is so important.

This is the value of predictive analytics.

You can create content that is months in advance.

How many of you had this problem.

You’re in your marketing meeting with your team or your agency or whatever.

It’s Monday morning.

Everyone’s got their cup of coffee.

They’re staring at the bottom of that one fast.

And you’re like, so, we’re going to blog about this week.

We’re gonna put on YouTube this week, what’s going up on Facebook this week? And I was like, ah, I don’t know.

Um, what what are people talking about? Open up the social media monitoring system.

What are what are people asking about? That’s not bad.

But wouldn’t it be so much nicer, so much less stress for you, your team, your agency, just hand somebody this calendar.

For the next 26 weeks here the questions people are most likely to be asking using predictive analytics.

Now we can plan ahead.

Now we can build our content and we have to rush to get rid of that stress.

What are we going to do can be a lot more efficient.

I remember when I was working agency side, I would watch teens do this sort of thing, this kind of planning, and they would burn hours and hours each week, just saying well, what are you going to do? This client, what are gonna do for this client? What are we going to do? And they never, they lost that time, hours and hours each week losing that time instead of having this predictive calendar to say, Hey, this is what we’re doing this week.

Other advantage, it gives you the ability to have higher quality content if you know, by the time December 15 comes around with it, we’re recording this the week of November 7, if we know where we’re out like here, and what is ricotta cheese tastes like ricotta is a neutral cheese, right? That doesn’t taste like much of anything can’t taste like tofu.

Which means it’s great for recipes.

We don’t want an overwhelming cheese flavor.

I have enough lead time now to build a video to you know, to get the ingredients together to shoot a video to record some audio with it.

And to prepare a nice piece of content instead of that Monday morning going I don’t know.

Let’s talk a quick blog post about this.

Maybe tweet about it and hope it resonates.

By planning this far ahead.

You know what’s going to happen and then you can take All that planning and all the content that you’ve done and loaded into a system like Agorapulse to have your content ready long in advance, have it scheduled out weeks and months in advance, using this data, taking away that scramble and and timing it with what the audience really cares about.

So that’s the power of predictive analytics.

That’s what I want you to take away.

So we’ve talked about what predictive analytics is using analytics to predict what’s likely to happen, why it’s important to end that scramble to deliver higher impact and how to do it.

From the data.

You need.

numbers with dates attached to them.

The tools you need these statistical and machine learning tools, and then the process of doing the cooking the test kitchen, refining it to a usable product, and then giving it to people and taking action on it.

That last part again, can’t say this enough.

Analytics without action is a recipe you cook but neveri we’ve talked a lot about food.

We’ve talked a lot about cooking This, this recording today, I can’t emphasize enough, you’ve got to take action on your data and your analytics without it.

Nothing else matters, right? It’s just a waste of time.

So please take action on this stuff, please do the thing.

Use predictive analytics to take away the scramble to deliver more impact and to make your life better and to feel good about the work you do.

Because you know, with great planning, you’re going to create high quality work that you can be proud of.

You can reach me at a CS pen at TrustInsights.ai dot AI, visit my company Trust insights.ai.

If you like to pick up a copy of my book, have one back here probably should go to AI for markers book.com it talks a little bit about predictive analytics, and a whole lot more about machine learning and AI.

So thank you for watching and I hope to see you soon at an upcoming event.

Feel free to close this out and move on to the next thing you want to binge watch Thanks for watching, want help solving your company’s data analytics and digital marketing problems? This is Trust insights.ai today and listen to how we can help you