Part 7 -Building Your AI-Ready Data Stack: Types of Analytics techniques

10-part series about building your first Data stack from 0 to 1, and be ready for AI implementation.

Hey there,

Can’t believe we are at part 7 of our series. I am sure you got some time to setup your data stack on AWS and already loving it!

Before we dive into last section of our series, which is build predictive analytics model using ML model, I wanted to share importance of understanding why only "Predictive” analytics, what “other” type of analytics techniques out there?

let me share a little story that might sound familiar to some of you.

The Analytics blunder that (nearly) cost me a promotion

Few years ago, I was managing a data team at a fast-growing startup. Our VP comes to me with a big ask: "Gunjan, we need to forecast our user growth for the next year. Can you handle that?"

Now, at this point, I should've taken a step back and thought about the best approach. But no, I was too eager to impress. I dove headfirst into the data, crunching numbers, creating the most beautiful descriptive analytics report you've ever seen. Pie charts, bar graphs, trend lines – you name it, I had it.

I presented my findings with full confidence "Based on our historical data," I proclaimed, "we can expect a 15% increase in users next quarter!"

Then our CTO cleared his throat and asked, "But Alex, what about our upcoming product launch? And the new marketing campaign? How do those factor into your prediction?"

My heart sank. In my eagerness to show off my data skills, I'd completely missed the boat on what we really needed: predictive analytics. I'd looked backward when I should've been looking forward.

I know promotion in headline was just clickbait ;) But you know what? It was still the best learning experience of my career. It taught me a valuable lesson: knowing which analytics technique to use when is just as important as knowing how to use them.

Busting the Big Data Myths

Now, I know what some of you might be thinking. "But , isn't all analytics the same? You just crunch the numbers and get either bar or line or pie chart, right?"

umm, I wish, but it is more complex than that. Let me explain.

Myth #1: All data analysis is the same. Reality check: Analytics is like a toolbox. You wouldn't use a hammer to tighten a screw, would you? Different analytics techniques serve different purposes.

Myth #2: More data always means better insights. Plot twist: Sometimes, more data just means more noise. It's not about how much data you have, but how qualitative and diverse data you have.

Myth #3: You need to be a math genius to do analytics. Spoiler alert: While math skills help, analytics is more about asking the right questions and understanding your domain.

Myth #4: Analytics is only about looking at past data. News flash: Some types of analytics can help you peer into the future. and we are talking ML yet. Mind-blowing, right?

Now that we've cleared the air, let's dive into the four main types of analytics techniques. By the end of this article, you'll know exactly which tool to reach for in your data toolbox.

---

1. Descriptive Analytics: The Foundation of Data-Driven Decisions

Think of descriptive analytics as your campaign's rearview mirror. It tells you what's already happened, giving you a clear picture of your past performance. This technique is all about summarizing historical data to give you insights into things like:

• How many people clicked on your last email campaign?

• What was the conversion rate for your recent social media ads?

• Which products were bestsellers last quarter?

Descriptive analytics is your starting point. It's simple, straightforward, and essential for understanding your current position before you start plotting your next move.

When to Use Descriptive Analytics

You'll want to reach for this technique when you need to:

1. Establish a baseline for your campaign performance

2. Identify trends or patterns in your historical data

3. Generate regular reports for stakeholders

4. Understand what's working (and what's not) in your current strategies

Putting It into Action

Here's how to get started with descriptive analytics:

1. Identify your key metrics. What numbers really matter for your campaign? Is it click-through rates? Conversion rates? Total revenue?

2. Gather your data. Pull from your CRM, Google Analytics, social media platforms – wherever your campaign data lives.

3. Crunch the numbers. Calculate averages, percentages, and totals. Look for patterns or anomalies.

4. Visualize your findings. Don't just stare at spreadsheets. Turn your data into charts or graphs to make trends easier to spot.

5. Share your insights. Present your findings in a clear, concise way that tells a story about your campaign's performance.

For example, let's say you're running an e-commerce store. You might pull data on your top-selling products, average order value, and busiest shopping days. Visualize this in a dashboard, and suddenly you've got a clear picture of what's driving your sales.

2. Diagnostic Analytics: Uncovering the 'Why' Behind Your Data

So, you've used descriptive analytics to see what's happening in your campaigns. But now you're scratching your head, wondering why your email open rates dropped last month or why that social media ad performed better than expected. Enter diagnostic analytics.

Diagnostic analytics is all about drilling down into your data to understand the reasons behind trends and anomalies. It's like being a data detective, looking for clues and connecting the dots to solve the mystery of your campaign performance.

When to Use Diagnostic Analytics

Reach for this technique when:

1. You spot an unexpected trend in your descriptive analytics

2. A campaign underperforms (or over performs) and you need to know why

3. You're trying to identify factors that influence your key metrics

4. You need to explain performance changes to stakeholders

Making It Happen

Here's how to approach diagnostic analytics:

1. Start with a question. What specific trend or anomaly made you go - uh, that’s weird.

2. Look beyond your initial metrics. If you're investigating email open rates, for example, you might also want to look at send times, subject lines, and subscriber demographics.

3. Look for correlations. Are there any patterns that align with your anomaly? Did open rates drop when you changed your subject line style?

4. Create hypotheses and test them. Think of possible explanations and see if the data supports them.

Let's put this into practice. Say you're a content marketer and you notice your blog traffic dropped by 30% last month. Here's how you might approach it:

1. Question: Why did our blog traffic drop so significantly last month?

2. Look at metrics like bounce rate, time on page, traffic sources, and publishing frequency.

3. Look for correlations: You notice the drop coincided with a change in your publishing schedule.

4. Use tools: Run a correlation analysis between publishing frequency and traffic.

5. Test hypothesis: You hypothesize that the reduced publishing frequency led to the traffic drop. Test this by temporarily increasing frequency and monitoring traffic.

By the end of this process, you'll have a much clearer understanding of what's driving your campaign performance. But don't stop there – use these insights to inform your future strategies. That's where predictive analytics comes in.

3. Predictive Analytics: Look into the future

Imagine if you could peek into the future and see (or predict) how your campaigns will perform. That's essentially what predictive analytics offers.

Predictive analytics takes the insights you've gained from descriptive and diagnostic analytics and projects them forward. It's not about certainties, but probabilities – giving you educated guesses about what might happen next.

When Predictive Analytics Shines

You'll want to leverage predictive analytics when:

1. Planning future campaigns and want to estimate their potential success

2. Identifying which leads are most likely to convert

3. Forecasting sales or revenue for the upcoming quarter

4. Predicting customer churn and taking preemptive action

Putting Predictive Analytics to Work

Here's how to get started:

1. Choose your target variable. What exactly are you trying to predict? Is it sales numbers, click-through rates, or customer lifetime value?

2. Gather your historical data. And make sure data quality and diversity is in place, this will improve the scenario testing.

3. Select your predictive model. This could be regression analysis, decision trees, or machine learning algorithms depending on your needs and data.

4. Train and test your model. Use a portion of your historical data to train the model, then test it on the remaining data to check its accuracy.

5. Check the model output, build the feedback loop for continuous improvement

Let's see this in action. Say you're running an e-commerce store and want to predict your sales for the upcoming holiday season. Here's how you might approach it:

1. Target variable: Monthly sales revenue

2. Gather data: Collect past sales data, including variables like marketing spend, website traffic, and seasonal factors.

3. Select model: You might choose a multiple regression model for this scenario.

4. Train and test: Use data from previous years to train your model, then test it against last year's holiday season to check accuracy.

5. Predict and plan: Use your model to forecast this year's holiday sales. Based on the prediction, you might adjust your inventory levels or marketing budget.

Predictive analytics isn't about perfect forecasts. It's about making more informed decisions based on likely outcomes. Use these predictions as a guide, but always be ready to adapt as new data comes in.

And speaking of adapting, that's where our final technique comes into play...

4. Prescriptive Analytics: Data-Driven Strategy

We've looked back with descriptive analytics, dug deep with diagnostic analytics, and peered into the future with predictive analytics. Now it's time to answer the all-important question: "So what should we do about it?" That's where prescriptive analytics comes in.

Prescriptive analytics takes all the insights you've gathered and uses them to recommend actions.

When to Embrace Prescriptive Analytics

Turn to prescriptive analytics when:

1. You need to make complex decisions with multiple variables

2. You want to optimize resource allocation across different campaigns

3. You're looking to automate decision-making processes

4. You need to balance competing objectives (like maximizing reach while minimizing costs)

Making Prescriptive Analytics Work for You

Ready to let data guide your strategy? Here's how to implement prescriptive analytics:

1. Define your objective clearly. What exactly are you trying to optimize?

2. Identify your constraints. What limitations do you need to work within?

3. For this, You'll need both historical data and real-time inputs.

4. Choose your analytics tool. This might be optimization algorithms, simulation models, or machine learning systems.

5. Generate and evaluate scenarios. Your tool should provide multiple options based on different possible conditions.

6. Implement and monitor. Put the recommended actions into play and closely track the results.

Let's see this in action. Imagine you're a digital marketer with a fixed budget, trying to allocate resources across different channels for maximum ROI. Here's how you might use prescriptive analytics:

1. Objective: Maximize overall campaign ROI

2. Constraints: Total budget of $50,000, campaign duration of 3 months

3. Data: Historical performance data from each channel, current market trends, competitor activity

4. Tool: You might use a linear programming model for this scenario

5. Scenarios: Your model generates several allocation strategies based on different market conditions

6. Implement and monitor: You choose the most promising strategy, implement it, and continuously monitor performance, ready to adjust if needed

The beauty of prescriptive analytics is that it doesn't just tell you what might happen – it tells you what you should do about it. It's the ultimate tool for data-driven decision making.

Bringing It All Together: Your Data-Driven Campaign Strategy Toolkit

There you have it, from looking back with descriptive analytics to charting the course ahead with prescriptive analytics. But here's the kicker – these techniques aren't meant to be used in isolation. They're most powerful when used together, forming a comprehensive approach to data-driven campaign strategy.

Think of it as a cycle:

1. Use descriptive analytics to understand your current position

2. Apply diagnostic analytics to understand why you're in that position

3. Leverage predictive analytics to forecast where you're headed

4. Employ prescriptive analytics to determine how to get where you want to go

Then rinse and repeat, constantly refining your strategy based on new data and insights.

A Word of Caution

As powerful as these techniques are, they're not magic bullets. Here are a few things to keep in mind:

1. Data quality matters. The insights you get out are only as good as the data you put in. Always strive for accurate, comprehensive data.

2. Context is king. Numbers don't tell the whole story. Always consider the broader context when interpreting your analytics.

3. Balance data with intuition. Use data to inform your decisions, not to make them for you. Sometimes, your gut instinct is worth listening to.

4. Stay flexible. The market is always changing. Be ready to adapt your strategies based on new data and insights.

In our next article, we'll dive into setting up your predictive analytics model with hands on coding. Trust me, it's going to be a game-changer for your campaign strategies!

Until then, keep questioning, keep exploring, and most importantly, keep learning from your data.