One thing online businesses need to use as a cornerstone of their business decision making process is their digital analytics data (analytics data from a variety of sources: i.e. web analytics, search console, paid search, paid social, social media, etc.). Yet, according to a MIT Sloan Management Review only 15% of more than 2,400 business people surveyed trust their data. While there is no analytics method available that will guarantee 100% accuracy of your digital analytics data, by auditing your data you can ensure the data is as accurate as possible. This will provide you with the confidence to not only trust your data but to leverage the information provided in making objective business decisions, instead of subjective decisions. It is that lack of trust that explains why a mere 43% (according to the same survey) say "they frequently can leverage the data they need to make decisions." This low confidence in one's data equals failure.
As marketing and analytics professionals, we need to work together to not only increase the accuracy of our data, but to educate people about the data and how to leverage it. The first step in this process is auditing your analytics configurations and thereby identifying issues, and correcting them to ensure the integrity of the data.
The analytics audit process
Step 1: Acknowledge analytics data isn't perfect
When you start your analytics process gather together all those who have a stake in the outcome and find out why they don't trust the data. Most likely they have good reasons. Don't make claims that your goals is to make it 100% accurate because that is impossible. Use the opportunity to explain at a high level that analytics data captures a sampling of user activity and for various technical reasons, no system will be perfect and that's why they are most likely seeing data difference between things like their Adwords account and their web analytics data. Use an example of taking a poll. Pollsters take a sample ranging in size of 1,000-2,000 people of a total population of over 350,000 in the USA and then state their data is accurate within a few percentage points 4 out of 5 times. In other words, they are way off 20% of the time. However, businesses, politicians and the general public respond and trust this data. When it comes to web analytic data even at the low end of accuracy, your data is likely still capturing an 80% sample which is far more accurate then the data presented by pollsters, yet it is less trusted. Let the stakeholders know, as a result of the audit and implementing fixes, you could be improving data capture accuracy to 90% or even 95%, and that is data you can trust 100%.
Step 2: Identify what needs to be measured
One of the biggest issues when it comes to analytics data is, the analytics software isn't configured to collect only the correct data. The software becomes a general catch-all. While on the surface it sounds perfect to just capture everything (all that you can), when you cast a huge net you also capture a lot of garbage. The best way to ensure the right data is being captured and reported on is to review the current marketing and measurement plans. Sadly, too few organizations don't have these, so during your meeting, make sure to ask what the stakeholders' primary items they want measured are.
Identify and gather all the "Key Performance Indicators" (KPI) currently being reported on. You'll need this before you start the audit. Verify all KPI are still valuable to your organization and not just legacy bits of data that have been reported for years. Sadly in many organizations, they are still reporting on KPIs that actually hold little to no value to anyone within the organization.
Step 3: Review the current analytics configuration
Now is the time to roll-up those sleeves and get dirty. You'll need admin level access (where possible) to everything or at a minimum full view rights. Next you'll need a spreadsheet which lists the specific items that you need to review and ensure are configured correctly and if not, a place to note what is wrong and a column to set a priority to get them fix.
The spreadsheet I've developed over the years has over 100 standard individual items to review, grouped into specific aspects of a digital analytics implantation plus depending on the specific client additional items may be added. The following eight are some of most critical items that need to be address.
- Overall Analytics Integrity: is your analytics code (or tag manager code) correctly installed on all pages of your website? Check and make sure your analytics code is correctly deployed. Far too often the javascript (or the code snippet) is located in the wrong place on the web page or perhaps it is missing some custom configuration. Simply placing the code in the wrong place can cause some data to not be captured.
Verify the code is on all pages/screens. Too often either section of a site are missed or the code doesn't work the same on all pages resulting in lost data or potentially double counting.
If you run both a website and an app, are their analytics data properly synced for data integration, or is it best to run them independently?
- Security: Review who has access to the analytics configuration and determine when the last time an individual's access and rights were reviewed. You'd be surprised how many times, it has been discovered that former employees still have admin access. This should be something that is reviewed regularly, plus a system has to be in place to notify the analytics manager when an employee departs an organization to terminate their access. While you may think since you're using the former employee's email address all is fine because HR will cancel that email address, they may still have access. Many analytics systems do not operate within your corporate environment and are cloud-based. As long as that former employee remembers their email address and the password to that specific analytics account they'll have access.
- Analytics Data Views: This is an especially critical feature when it comes to web analytics (i.e. Google Analytics, Adobe Analytics, etc.). Is your analytics system configuring to segregate your data into at least 3 different views? At a minimum, you need "All Data" (no filtering), "Test" (including only analytics test data or website testing) and "Production" (only customer-generated data). In many organizations, they also segment their data further into "Internal Traffic" (staff using the website) and "External Traffic" (primarily external users).
If these don't exist, then it is likely you are collecting and reporting on test traffic and internal users. How you're employees use a website is completely different than customers and should at a minimum be excluded or segmented into their own data set.
- Review Filters: Filters are a common tool used in analytics to exclude or include specific types of activity. Most filters don't need to be reviewed too often, but some do need to be reviewed more frequently. The ones that need to be reviewed most often are ones that include or exclude data based on a user IP address. IP addresses do have a nasty habit of changing over time. For example, a branch location switched ISPs and received a new IP address. When it comes to IP based filters it is recommended they be reviewed every 6 months, but if not possible at least once per year. As a tip, after they've been reviewed and verified, rename the filter by adding the date they were last reviewed.
Don't forget to ensure that exclude filters are in place to exclude search engine bots and any 3rd party utilities used to monitor a website. This machine-generated traffic has a nasty habit, of getting picked up and reported on which skews all the data.
- Personally Identifiable Information (PII): Too many developers have a nasty habit of passing PII via the data layer especially on ecommerce sites. They are unaware that it is possible that this data can end in the company's analytics database. By storing PII on a 3rd party system, you either need to reference this in your privacy policy and even then you might be breach of various privacy laws. One of the most common types of these errors is passing a users email address as a URI parameter. The easiest way to check for this is to run a pages report for any URI containing a "@" symbol. Over the years, I've seen customer's names capture and much more.
If this happens, ideally your developers should fix what is causing this, but at a minimum you'll need a filter to strip this type of information from the URI before it is stored in your analytics database.
- E-commerce Data: This is the most common issue we hear from organizations: "The sales figures reported in the analytics doesn't match our e-commerce system!" As stated above, analytics isn't perfect nor should it be treated as a replacement for an e-commerce/accounting backend. However, if you are capturing 85-95% (and possibly higher) of the transactional data then you can effectively leverage this data to evaluate marketing efforts, sales programs, AB tests, etc.
From an e-commerce perspective, the easiest way to audit this is to compare the reported data in a given time period to what the backend system reports. If it is near 90%, then don't worry about it. If it is below 80%, you have an issue. If it is somewhere in between, then it is a minor issue that should be looked into but is not a high priority.
- Is everything that needs to be tracked being tracked: What does your organization deem important? If your goal is to make the phones ring, then you need to be tracking clicks on embedded phone numbers. If your goal is forms driven submissions, are you tracking form submissions correctly? If you're trying to direct people to local locations, then are you capturing clicks on location listings, on embed maps, etc.?
What about all those social media icons scattered on your website to drive people to your corporate Twitter, LinkedIn, Facebook accounts? Are you tracking clicks on those?
- Campaigns: Is there a formal process in place to ensure links on digital campaigns are created in a consistent manner? As part of this are your marketing channels correctly configured within your analytics system?
You now have an outline for where to start your analytics audit. Think of your organization's analytics data and reporting systems like a car. It always seems to be working fine until it stops working. You need to take your car in from time to time for a tune-up. This is what an analytics audit is. The audit will identify things that need to be fixed immediately (some small and some big) plus other items that can be fixed over time. If you don't fix the items discovered during the audit your analytics system won't operate optimally and people won't want to use it. How frequently should an analytics audit be conducted after everything has been fixed? Unlike a car, there is no recommended set amount of time between audits. However, every time your digital properties undergo major updates or if there have been a series of minor updates that can easily be viewed together as a major update, it is time to repeat the audit process.
Opinions expressed in this article are those of the guest author and not necessarily MarTech Today. Staff authors are listed here.
Comments
Post a Comment