Month: February 2020

How to set up your first alert in YetPulse

Intelligent alert uses an AI powered algorithm that builds a forecast for your metric and checks it on an hourly basis. Once the value of your metric goes beyond that forecast, you will receive an alert.

This is extremely helpful for overseeing advertising traffic or revenue. Because it helps you get alerted just in an hour when you get a spike or drop in traffic, revenue or any other metric. Spikes and drops indicate possible problems. By getting an alert in real time, you can react faster.

To set up your first alert you need to be logged-in.

Create a new project

Click on a “add new site” on the home screen

Choose a name for your alert, enter it and click “apply” in the top right corner.

Connect your Google Analytics

Once the account is created, you will see your project home screen. Click on “add analytics”:

Then click “grant access”:

Then you will need to choose the Google account that you use to access Google Analytics for the website you want to add.

Then, grant permission to YetPulse to view your Google Analytics data:

Click “allow”

Click “allow” again:

Select your Google Analytics profile (or view):

You should see this:

Next, click the “apply” button in the top right hand corner.

Create an Alert

On a project home screen click the “add new intelligent alert” button:

In a popup that appears , choose the dimension that you want to oversee:

Then choose your preferred metric and click “apply”:

Once you create an alert it will appear on your project home screen:

This alert is set for all sessions from Facebook. YetPulse will check it on an hourly basis and send you an alert if there is a spike or drop in sessions.

Having problems setting up your first alert? Contact us by live chat for help.

How To Set Data-Driven Alerts in Google Analytics

3 years ago, disaster struck.

I was working in an advertising agency. A week before a monthly results meeting with a client, we realized that client’s landing page hadn’t worked for 3 days. Nobody knew, not us and not the client.

We were running Display & Video 360 and a Facebook campaign on that landing page and we thought everything was fine. You can imagine how tough this meeting was. It was one of the hardest meetings of my career. The client was furious with us for not noticing, though we knew the fault came from the client’s team. But that didn’t matter. As an advertising agency, part of our job was to foresee problems and proactively mitigate them.

Of course, we weren’t about to start blaming anyone for the failure. We went to the meeting full of ideas of what to do and how to react faster in the future. Our key idea was to leverage the power of custom alerts in Google Analytics. It seems obvious, but you might be surprised that usually, less than 10% of marketers use custom alerts because of their limitations.

We set up alerts for that client to tell us if conversions dropped by more than 20% compared to the same day the previous week. Problem solved? Not yet. Because we then received an alert right the next day, which turned out to be false. And then we received the same false alert after another couple of days.

We realized that if the false alerts kept coming, we would end up ignoring them. And then we’d inevitably end up ignoring an alert that wasn’t false. This continues, we will stop reacting on alerts at all. Even if one of these won’t be a false.

We turned to Google to find a solution. Back then, there were no anomaly detection services like YetPulse, so we had to figure out what to do using the alerts in Google Analytics. Our analysts came up with a brilliant approach based on our data. It wasn’t complicated, but it did require some Excel skills and a bit of time, both to build the alerts and to keep settings updated as the data changed.

I’m going to share this solution with you. It’s so simple, but powerful. Using this approach you will be able to set alerts that work. These alerts won’t disturb you with false notifications and they will capture key incidents with your data. You’ll just need to update all your settings once every week or two. When my analysts shares this principle with me, I thought “why I didn’t come up with this first?”

You can set intelligent alerts for your Google Analytics using our FREE service. Also, you will get same day notifications, since it’s checking your data on a hourly basis.

I’m going to assume you know the basics on how to set alerts.

To start, you’ll need to choose the metric and dimension you want to track. In the example below, I’ll set alerts for session drops and increases for the CPC channel. But you can set alerts for other channels and other metrics, like conversion rate or revenue. This is up to you.

To start, you’ll need to choose the metric and dimension you want to track. In the example below, I’ll set alerts for session drops and increases for the CPC channel. But you can set alerts for other channels and other metrics, like conversion rate or revenue. This is up to you.

Export your daily data in Excel format

Then, open an Excel document and find a tab called “Dataset2”. This contains daily data for a metric chosen on a chart in your Google Analytics report.

In this tab you will see something like this:

Next, we need to calculate week over week change in %. Usually I recommend using a percentage if you have enough data. Also, I prefer to set week over week change alerts to mitigate seasonality in data. In my case, I have very strong weekly seasonality: during weekends we got significantly fewer sessions than on working days.

A very simple formula is used for that: ( current_value — last_week_value ) / last_week_value. For small amounts of data (like less than 500 sessions/conversion a day) I would suggest to subtract, like this: current_value — last_week_value.

If you have zeros in your data, so if on some days you got zero conversions or sessions, it might be wise to switch to weekly reports instead of daily.

And then, just copy this formula down to the bottom of column.

At the end you will get something like on the screenshot below. (I’ve formatted this one to make it look better).

Now we’re very close to our solution.

You just need to draw a chart for the last column. This chart going will illustrate how much your data usually changes.

I’ve got this:

It represents a % of change, week over week, for each day. If I pick a value for the 10th of October, it will show my by what percentage the number of sessions changed that day, compared to the same day in the previous week.

All these steps are needed to us to understand normal behaviour for your metric (in my case it’s sessions). You can do the same math for revenue, or conversion rate, or any other metric in Google Analytics.

Now, we’re almost ready to set thresholds for our data-driven alerts, but first I would recommend to cleanse your data.

What do I mean by cleanse?

Imagine that we ran a big new campaign a month ago. Of course this will be reflected in our chart. Due to the nature of change, this is not normal behaviour. We need to exclude that day from our chart.

Again, our goal is to determine the normal rate of change for our session. If some of the days are not “normal’’, they should be excluded.I personally just replace this cell with a zero.

To investigate whether you’re seeing normal behaviour or not, open the Google Analytics report with two dates to compare: the date that you are interested in and same day the previous week.

After cleansing the chart, we can define our threshold for custom alerts.

Here is an example of a best fit threshold, which represents: 37% increase, 25% decrease.

I used maximum values to determine thresholds.

I’m sharing because I want to help businesses react faster to incidents. If you have stuck with this approach you can use our free tool to set intelligent alerts: using this link.

6 Things You Need to Know About Custom Alerts in Google Analytics

Custom alerts are difficult to use and have poor functionality. Most marketers don’t use them because they don’t do anything to make their lives easier.

Here are 6 reasons why.

1. You have to manually specify every alert

I’ve seen a list of 55 valuable custom alerts for Google Analytics, and each one had to be manually defined. That takes a lot of time that I, and most other marketers, would rather be using to do something else. It’s crazy to have to set every alert up from scratch at the start of every new project.

2. It’s hard to set thresholds

You have to dig deep into your data to find the right threshold, using Excel and custom formulas. Again, this takes time… and you can’t help thinking that it could and should be so much easier. It’s difficult to set it up so that you don’t get false alerts constantly, which take you time to deal with. It doesn’t save you time.

You need to constantly adjust thresholds to your data. If your data changes you have to manually update all your alerts settings, which costs time rather than saving it.

3. It’s too easy to ignore notifications

If it’s not set up correctly, Google Analytics provides multiple notifications, many of which are irrelevant. This means that they’re ignored…because users feel like they’re being spammed with constant, pointless alerts. Any alerts that are relevant get ignored along with the rest. One of the reasons for the excessive number of notifications is that custom alerts aren’t adaptive. The threshold doesn’t update automatically and not everyone has the skills to set it correctly.

4. Notifications aren’t real-time

Even if you manage to set up alerts properly in Google Analytics, they’re slow, not coming through until the next day. This means if you’re running PPC ads, you could be losing a lot of money. If there’s a problem with your ad, you won’t get a notification about it until it’s too late. I have 15 years’ experience of working in PPC agencies, I know just how much money can be lost in a day.

5. It has limited functionality

At times, Google Analytics doesn’t seem to offer even the most basic information, such as notifications of a campaign that’s getting zero conversions. You cannot set up useful alerts until you get some data, as you don’t know what thresholds to set.

6. It’s not scalable

Ads are designed to grow your business. But custom alerts simply can’t be scaled. So, if your ads are working, Google Analytics won’t be. This is particularly true if you have multiple people managing ads for you.

Can custom alerts be made to work better?

I’ve seen suggestions that Google Ad Scripts can make custom alerts work better, but that’s not my experience. Not least because many of the points I’ve made here also apply to Ad Scripts.

And what if you’re running ads outside of Google (e.g. Facebook, Bing, Criteo, Instagram or Display & Video 360)? Ad Scripts can’t help you on those platforms.

The bigger your budget, the bigger your problem. Once you reach an ad spend of millions, you have a lot to lose and really need to be able to detect and fix issues as soon as possible. Mistakes become very expensive.

I have a client who once lost 100,000 Euros over a weekend by accidentally enabling an old campaign…that had become a 404 page.

That experience is what prompted the client to sign a contract with us at YetPulse. We now monitor their ads 24/7 and if there ever is a problem, we can deal with it immediately.

Why Anomaly Detection is Vital for Digital Marketers

When you’re working on multiple campaigns, you end up with tons of data. Say you’re running 1000 ad campaigns. Naturally, you want to understand if everything is working with each one.

That’s a lot of data to work through.

You might end up with that data on multiple spreadsheets. But reading and understanding it all could take hours. You’re almost certain to lose focus and miss something important. Maybe something that could change your approach completely and make you and your clients much more money…if only you knew about it.

You need a system for analyzing data

That system is anomaly detection. This is a problem-solving technique used by statisticians.

An anomaly is a statistical outlier. It’s easy to see outliers when you put the stats into a graph.

For example:

A perfect example of an outlier
This is also an outlier

These are obvious outliers. But it’s not always so simple.

Consider this graph:

Example of a non-obvious outlier

Is the last point on the graph an outlier? It could be, but whether it is depends on what you’re hoping to achieve.

Outliers are not the only anomalies. There are also inliers.

Here is an example:

Example of inlier anomaly

This graph shows website sessions. The number of sessions increases on Mondays and the number of weekday visitors is around double that at weekends. Except in the last week, which shows Monday and Tuesday sessions stuck at weekend levels.

This is a good example of an anomaly that cannot be tracked by rule-based alerts.

Anomaly detection would highlight it automatically. It’s exactly the kind of unexpected anomaly that you’re likely to miss if you’re using spreadsheets.