Skip to content

5 lies you talk about your analysis (and how to fix it)

The consultation data are good.

But being a slave to the data is not.

There is such a thing as being too obsessed with data. The confirmation bias appears. And you miss the good things, though intangible, that accompany your efforts.

The solution is to discover prejudices and misunderstandings that mislead you.

This is not easy. Or even intuitive. But that's the only way to avoid these five blinkers of analysis.

Here's how it goes when you least expect it.

Here's why you fall for it.

And here's how to avoid it by bringing in other types of comments and analyzes.

Lie # 1. Your "conversions" are perfect

You have three AdWords campaigns.

  • The first reports zero advance on 78 dollars spent.
  • The second brings in one for $ 135.31
  • The third earns two wins at $ 143.28 in advance.

Nine times out of ten, the campaign with more "conversions" is declared a winner.

But what do you really, really know about this scenario?

Which campaign is the most successful? Which puts the most money in your pocket?

There is simply no way to say it at this point.

First and foremost, these "conversions" are leads – not closed clients.

Second, they could be for different products or services. So, different average control values ​​or LTVs come into play.

Third, this is nowhere close to statistical significance. For example, the third campaign has the most leads because you have spent the most money on it.

Not because it's "better".

And if you just spent the same amount on the first two? What would happen if you left them both around the same value, $ 150 / mark?

Do you see what I mean?

Too much "what ifs" for my taste.

Yet, that's exactly what's happening in any marketing department. The same final result appears after each client or superior meeting.

Everyone points to the third campaign. He gets adulation. He gets the increased budget. He receives staff and additional resources.

Thus, it becomes a self-fulfilling prophecy.

One solution to understand all this is the closed-loop analysis.

Ideally, your goal is to match customer information (name, email, phone, credit card) to the main data you see in Google Analytics.

Haha – I'm kidding.

This would mean that you were collecting personally identifiable information, which is a big no-no in Google Analytics.

Do it and they will delete your account right away.

The simplest alternative is to simply use a tool that gives you that power without compromising your data. Index, index.

Lying # 2. Your "best" sources of traffic

What are your main sources of traffic?

A quick look into Google Analytics usually tells you (1) a natural and (2) direct search. Maybe a little (3) referrals for good measure if you have the press last month.

Here is the problem.

Two of these three are legitimate. The other is not it.

The problem is that your direct traffic is not, in reality, all that is "direct".

Technically, this should be the number of people typing the URL of your website in the address bar and typing "Enter".

Instead this is a healthy mix of email, social media, and good old organic search.

The bigger the site, the bigger the problem.

For example, The Atlantic could not explain or explain how 25% of its visitors came to their site.

One of the world's largest publishers. One of the most respected. Who is paid based on the number of visitors and page views. I do not know how a quarter of their traffic comes to their site.

See also  4 Basic Tips to write a great headline for Twitter

This is not good.

But how can you really tell where people come from, if most of the analysis programs can not tell you precisely?

For example, let's say that your new fantastic email campaign is about to come out.

He received the green light. "Legal" gives you the A-OK.

But wait! You have not correctly tagged promotional links.

Now you've spent all this time on a campaign that will have nothing to show, because the traffic you get will now end up in the dump pile officially called "Direct Traffic".

This is not just an email. It also affects every social message, press statement, and post-blog reference, too.

This can even affect your organic search traffic.

Groupon found it the hard way. Literally. By deindexing completely for a few hours.

What did these crazy cutters find? That nearly 60% of their direct traffic was actually resulting from organic research.

Seventy percent.

But do not panic right now. There are solutions here.

First, you can use Google's UTM generator to make sure you're tagging your links correctly. It means everything and everything you have control.

Mark them manually before going out, or copy and paste them into a lightweight application like Terminus.

If you have a long and cumbersome URL, you can be sure that traffic to this page is not from Direct traffic.

People are not going to remember it. Which means that they are not going to type it spontaneously.

Instead, these buddies probably come from another place, such as an organic search or an email.

However, in the same breath, you can probably consider home page traffic as legitimate.

Create a segment based on these URLs and traffic sources to identify "Dark Traffic" in its tracks. And prevent it from ruining your data in the future.

Lying # 3. Top of the funnel performance = results

Yes, we want traffic.

Yes, we want page views.

They make us all feel warm and fuzzy and proud. As our hard work does not go unnoticed.

But they should not be the end-all, be-all.

Use them to see how you are going in the last month. But do not misunderstand the numbers to be the Holy Grail either.

Like this, for example:

Looking only at this, you leave feeling like a boss for all the numbers you have accumulated. Seriously, I can not even count so high.

But what about when you consider the bounce rate and output for each of these pages? Do people stay? No? Color embarrassed you.

Are you still so excited by your thousands of page views if most of them are gone immediately?

Bounce rates are real. And you must take them into account when you consult your statistics.

They mean that people have not had the chance to interact with your micro-conversions. They did not have the opportunity to activate it.

So take a look at the big picture.

Do your blog posts and your site attract people, but do not make them stay?

This is not a horrible problem because it is a problem that you can identify.

The traffic is there. They do not really like what they see once they arrive on your site. That you can repair.

First, set up events to get a better idea of ​​what is happening on your pages. Then make sure you have achievable goals that will allow you to track movements.

See also  How to acquire customers with new Facebook ads

Or use the Kissmetrics Customer Engagement Automation tool to analyze what people are actually doing on your site and with your products. Then you can interact with behavior-based messaging to keep them longer. Or bring them back to find out more.

This way you can increase conversions, engagement and retention without guessing.

 kissmetrics populations "class =" full-size alignnone wp-image-34167 "/> </p>
<p> Always remember that numbers do not tell the whole story. Use them with a grain of salt and a little context. </p>
<h2> Lie # 4. A / B Deceiver "wins" </h2>
<p> I'm just going to be honest with you. These A / B tests "wins" that you just received? Do not always have the best record. </p>
<p> I am sorry to be so hard from the start. Sometimes the truth hurts. </p>
<p> What's more disturbing? Often, the tests <em> look like </em> they succeeded. But this is not always the case (or at least not the whole picture). </p>
<p> Start by using the Google Analytics content tests. </p>
<p> You can use it to compare your different pages to see if there are any important adjustments that are causing positive changes. </p>
<p> Instead, it allows you to compare different page variations to see which "bigger" changes result in improvements. Maybe it works a little better because it adds an extra letter – it's a test A / B / N. </p>
<p><center> <img decoding=

And we're talking, conversions-conversions here. Like, bottom-of-the-funnel, paying for customer conversions.

Context is the key when you analyze the data.

Do not test landing pages or simple field edits by only evaluating the top of the funnel. Be sure to dig to see how the changes affect the rest of the customer experience and the trip.

To do this, use the Funnel Report so you can see exactly how the changes at the top of the funnel are impacting sales down the funnel.

Lying # 5. Attribution of the source of your chain

A few years ago, a Forrester Research study found that 33% of all transactions of all transactions took place after new customers had gone through more than one point of contact.

This number rises to 48% when considering regular customers.

The same report showed that paid search is the biggest source of conversions.

Is this, however?

Or is it just the last point most commonly used before a sale?

Just because it's the last one, does not mean that's the only one.

What other marketing tactics work to increase growth? Forrester went on to say that while email works for repeat conversions, social media reports less than 1% of sales.

See also  Report: Microsoft seeks to help retailers compete with Amazon Go

Ok. So how do you explain SpearmintLOVE?

You know, the baby blog that boosted their business by 991% using Facebook and Instagram.

The only reason I know about them? Because my wife bought clothes from them. After discovering them on Facebook and Instagram .

A simple, Google graphic puts this myth to bed. Fast.

If you look at the left side, or "help interactions", you will see that social channels will bring new products to people.

When you move to the middle, customers get more product information and options for the help of research. In the end, they are en route to the website.

Note all possible interaction options here. This is not just the last touch that brought the customer to the website. They can take a lot of steps to get there.

Google Analytics incorporates several different attribution options to help you change the way conversions are awarded.

Source of image

These include:

  • Last non-direct click : This allows you to ignore direct clicks and access the channel used just before.
  • First interaction : She uses the social network or advertising that brought them to the website.
  • Linear : Here, each channel used by a customer before the purchase will get an equal allocation.
  • Time Decay : This will consider the channel that was used immediately before the conversion, rather than the channels used in the last days / week / month.
  • Position : This pattern gives priority to the first and last channel used before conversion. Everything in the middle receives less attribution.

The depressing part, though?

There is no good answer here. The attribution model you choose depends largely on your sales cycle, your customers and even the specific goal you are trying to determine.

For example, if you spend a lot of money on advertisements, you might want to see what the first interaction looks like. Especially when you use social ads that often bring people into your ecosystem for the very first time.

In other cases? It would be a terrible choice.

The trick is to find out first what you are solving. Then back.


The data is important. It's huge.


But, be careful.

Google Analytics is a wonderful, profitable and revolutionary tool.

However, we know that he lies a little from time to time. (Yes, we still talk about Google Analytics here.)

Remember that conversion results are not always positive. Direct traffic data may not be correct. Vanity metrics are not everything. A / B results can trigger false positives. And the last touch is not everything.

Discovering the bias is never fun.

But that's the key to creating campaigns that actually achieve results.

Without just blowing a lot of hot air.

About the author: Brad Smith is the founder of Codeless, a B2B content creation company. Frequent contributor to Kissmetrics, Unbounce, WordStream, AdEspresso, Search Engine Journal, Autopilot, and more.