A Business-Oriented Framework For Web Analytics

by

Data, data and data, anyone talks about it but only a few actually show a robust methodology for Web Analytics.

If you have ever wondered how to use Web Data to drive business changes, this is the place for you.

Almost anyone can apply this framework but the best ROI is for those in charge of larger websites.

Why Do We Need This Stuff?

Many websites have tons of data they don’t use which hides a lot of profitable insights.

Analytics helps you uncover these secrets and to simplify processes as a whole.

The issue is that people focus on technologies and NOT on how to use this data to make money.

It all boils down to the following:

  • Deciding faster
  • Improving the quality of decisions
  • Understanding which decisions you should take and when

Most people are obsessed with growth, forecasts and new theories when in reality the execution and management of many projects is weak.

Yes, we are talking about building systems and not “campaigns”.

The Framework

Analytics is by tradition split into multiple stages depending on your goals and maturity.

The idea is to ask the right business questions and understand what to focus on at each stage.

The picture below showcases the framework taken from one of my favorite books (source in the pic):

I don’t often ask predictive questions because I rarely work with Machine Learning & co.

Some say you can skip some steps and avoid following a linear path.

I think you can only skip Predictive and maybe Prescriptive but the first 2 stay there.

These frameworks are good BEFORE you start analyzing data and if you feel lost when working.

Business Level

This step involves understanding business needs and asking the right questions to your client.

Marketing is functional for a business and can’t work as a separate component; the same goes for data.

Analytics should be put into context to work and you need to define clear goals and relevant KPIs.

Using conversion rates as a measure of success for informational posts will be a failure, I can already tell.

This is why you should pay particular attention to your business needs.

This step embodies:

  • How much money are you making with the website?
  • What’s your most profitable channel?
  • Which KPIs work the best?

Many stakeholders will be skeptic because of the many problems of Web Analytics, like accuracy and attribution.

My school of thought is that a strong marketing knowledge trumps both of them.

Optimize your work to satisfy stakeholders and be strongly action-oriented, as we will see.

A big concept behind the business level is considering data as a product:

Not sorcery, not witchcraft, just a means to achieve your business goals, given the conditions listed in the picture.

Many marketers are familiar with product management and it’s the same concept.

Being business-wise also implies evaluating solutions by channel, like I did below:

You don’t even need deep technical knowledge to understand that AI stuff is inflated and you are fine without it.

The picture above classifies pages by Complexity and Business Value, another perspective compared to what we are often taught.

Descriptive Level

The majority of my content sits here.

Descriptive Analytics tells you what happened in the past and is represented by aggregations and the evergreen pivot tables.

It won’t tell you the why but it will show you the what.

Based on your experience, you can then use the insights to elicit action.

Analyzing Google Search Console/Google Analytics 4 data can unlock several opportunities, such as:

  • Finding new topics
  • Spotting dangerous traffic drops
  • Uncovering CTR Optimization

All of this is part of Descriptive Analytics.

  • What happened?
  • Can we quantify gains/losses?

Descriptive Analytics is often underestimated because it’s not glamorous and cool like AI and some can’t accept that simple solutions work.

The picture below shows an example I covered in my Content Auditing guide:

This is purely descriptive, it’s a snapshot of what happened in the past and that’s it.

You can understand a lot from this chart, for example drilling down on the last months that got more unique queries.

Still, this doesn’t prove anything or show what the future could look like.

The same goes when evaluating the traffic of a domain.

Sure, maybe a domain has a good distribution of clicks per page, you don’t have imbalance…

but it can be risky anyway because life is unpredictable.

Descriptive Analytics is what companies often need but NOT the only one they need.

Predictive Level

Predictive Analytics involves predicting future events based on available past data.

This means using Machine Learning which is based on training data models to make predictions.

In SEO, making reliable predictions is hard and Machine Learning is cost-expensive to bring decent results.

In PPC and even Social Media, I am a big fan of predictions and even estimates.

  • What is more likely to happen in the future?
  • Which factors may contribute the most to X?

Since updates are frequent and not rare events, your predictions will be largely off.

I find value in estimating costs for SEO or even potential revenue but it’s close to impossible to give accurate estimations for rankings.

To be fair, conversion rates and specific events are the most fun and valuable to forecast.

Machine Learning isn’t only used for predictions but you can check out ML For SEO if you want to know more about it.

There is also this one article that has some great ideas for forecasting traffic.

Traditional ML methods tend to be more reliable with PPC though.

If you need to forecast traffic, I recommend using GA4 and picking a traffic metric like Users (without filtering by source/medium).

Prescriptive Level

Prescriptive Analytics refers to preventing damage and optimizing processes. If traffic drops, which actions can we take to reduce the damage or prevent it?

You can optimize your processes to spot decaying articles and fix them as soon as possible.

Prescriptive Analytics is generally associated with AI and advanced systems but you can also use your brain.

Preventing problems doesn’t necessarily involve data, models or machines.

The table above shows you some of the counters to the most common problems for websites.

Once you figure out the main problems, you can efficiently work on solutions to contain the damage.

  • What can we optimize?
  • Which is the best course of action?
  • How can we improve our existing processes?

This synergizes with Project Management, Airtable and managing your data sources.

Speaking of which, it’s also necessary to mention the DIKW framework.

It’s a mental shortcut to think about how to make your recommendations actionable.

This model doesn’t fully overlap with our business framework because it measures different things.

BUT… use it to guide your actions as you walk through the 4 steps.

A Sample Application (Again)

Say you have to analyze a content website with a lot of pages (20,000).

You first figure out what they want and what’s the goal (Business Level) and then move on to gathering all the necessary data.

You run your analysis and explore data (Descriptive Analytics) and uncover some insights on their pages.

This can be used for reporting because it’s past data and you can show the current situation.

Armed with this knowledge, in rare cases you may want to predict if a page is going to be successful or future traffic (Predictive Level).

Anyway, once you figure out the most common problems, you hurry up to work on feasible solutions that prevent the problems (Prescriptive Analytics).

Processes: The Missing Link

It’s all cool but asking questions alone won’t save you…

after all, you need action but single tasks alone are useless.

That’s why you need to think in terms of systems and processes, seek the bigger picture.

Companies are built on processes and you need to understand what’s going on or what you could create.

This is an example of SEO, a sketch of what a process could look like:

I am not a fan of agile, scrum, stories, counter-stories and stuff but I preach common sense.

Measure things without overcomplicating the obvious too much.

If we talk about Web Analytics (not SEO, not PPC, etc.) you just need to provide frameworks and good recommendations, you won’t execute the job.

Politics Is Life

You read online that only big companies have office politics and startups are cool and trendy.

BS.

Life is politics. You have to understand what people really want and how to persuade them or your project will be ditched.

Business models help here and one of them is the Stakeholder Map.

A quick recap for those who don’t know it:

Manage closely.

These people make or break the project. The CEO calls the shots and can stop everything.

If you succeed, they will promote your project!

Keep satisfied.

Low interest but high in power/influence. The recommendation here is to show them what they get in exchange.

If you manage to catch their attention, the rest will follow like above.

Keep informed.

This group of stakeholders doesn’t really have decisional power but they support you. And as your “fanbase”, it should be included in what you do.

Don’t make the mistake of only going for the big shots or it will quickly backfire.

Monitor.

The people in this bracket aren’t influential enough or even interested. Keep them under observation and focus on the 3 other groups.

But things can change quickly, so monitor them!

If you can figure out who is who and your potential supporters, the rest is up to your social skills.

A great Analyst is also a good communicator.

Which Data Do I Need?

It’s kind of tough to say without knowing your objectives but you can be sure GA4 will have some saying.

For SEO, you can consult my post on the ideal SEO Analytics stack.

For other marketing channels, there is too much to say:

  • Google Ads
  • Meta data (Facebook, Instagram)
  • External tools (e.g. Salesforce, SAP)
  • Crawl data
  • Log files
  • Clarity/Hotjar
  • CRM data
  • CMS data

If you work with big companies or do actual data projects, you won’t be in your comfortable silo with GA4.

It’s usually about juggling with different sources and gathering requirements here and there.

Sometimes they can have solutions like Salesforce Data Cloud that add even more complexities.

As an Analyst, you are not required to be an engineer, mind you.

But you still need to understand what’s going on and where to get your data!

Requirements may change according to the complexity and the goals of the project.

How Is Analytics Misused In Marketing?

This can be the topic for another article too, as there are too many misconceptions flying around.

So far, the most dangerous I have seen are:

  • Analytics treated as knowing some basic Pandas (not even Python)
  • Confusing correlation with causation
  • Wrong methodology

This falls outside of Marketing but it’s important to remember that Marketing isn’t solo play. You can rely on other professionals and ask for their feedback.

If you want to learn more about Analytics for SEO, I’ve prepared a course just for it:

Unlock Powerful Insights Now!

The same applies to UX, Web Development and pretty much anything else.

My recommendation is to create immediate value with the Web (and non) data you have and build some basic processes.

This is the best way to get started instead of wasting time with FOMO.

SEO is one of the industries that are affected the most by it but also PPC, to be honest.

Small VS Big Website (Again)

One of the things I blame the most is equating a large website to a local business or a simple website.

The reason why so much advice from the big shots feels “detached” is that you can’t apply it to small/medium websites.

The gameplan is completely different and the goals too but that doesn’t mean you can’t profit from data with a small website!

Small Website

Your first priority is to grow your website and don’t waste time on checking Google Analytics/Search Console every hour.

That said, I recommend connecting BigQuery to the Search Console/GA4 as soon as possible to start getting data from the get-go.

Some people claim it’s expensive but I don’t know what they are on about because it costs pennies.

Most of the advice for this phase involves having a solid setup and a clear idea of how you want to use data.

Business and Descriptive clearly win here. Predictive and complex Presciptive Solutions are no use.

Google Sheets can solve many of your needs and provide basic automation.

There is no particular need to hire someone or invest some budget into data, think about growing and getting it first!

Large Website

The game for big websites is completely different, although the basics always work.

They have much more data and mature SEO/Marketing channels and most importantly active processes, so you can’t simply barge in.

The very first thing to do is investigate what’s already there and how they use existing data.

If you are lucky enough, they will have BigQuery and maybe some automation.

At this level of complexity, you need to define a strategy, define goals and think in terms of maintenance too.

If a solution isn’t future-proof, it means it will cause problems in the next few years (and we don’t want that).

Big companies can also feature a Data team that handles most of the crucial data-related tasks and even Marketing data.

It’s possible to work on all the 4 steps of the framework!

In most cases, they will have weak Marketing knowledge and use the wrong data or report on metrics that we don’t care much about.

This is your chance to shine and prove your worth.

Useful Resources

These are the usual cool resources I recommend in my newsletter because they are so good:

If you are a data person, these are essentials. They have changed my way of looking at data.

Marketers may prefer more technical reads to begin with as (I suppose) they are already good at communication and abstraction.

At least, they should be!

Leave a Comment

Terms and Conditions