• Analytics Experts Always Ask These 5 Questions

    Sometimes you might have all the data in the world, but still, if you fail to make adequate inquiries, you are going nowhere. Excelling both as a marketer, business developer or analyst requires asking the right questions at the right time. Digital analytics is demanding and not only does it require expertise but also a […]

    Sometimes you might have all the data in the world, but still, if you fail to make adequate inquiries, you are going nowhere. Excelling both as a marketer, business developer or analyst requires asking the right questions at the right time.

    Digital analytics is demanding and not only does it require expertise but also a flair for what might work. Here are the top 5 questions digital analytics experts always ask.

    What do the 5 analytics questions have in common?

    Regardless of the nature of the gathered data, the analysis and conclusion almost never turn up too surprising for the initial hypothesis. And if sometimes you get an astonishing result you weren’t expecting, at first you feel suspicious toward the validity of the results. When we forget such outliers, in the majority of cases, the numbers are widely accepted without much pondering.

    Analytics people
    Analytics people, Image courtesy of Digitalhrtech.com

    What makes a good analyst and how do they ask the right questions?

    Simply knowing the advanced interface of an analysis tool doesn’t make someone a great analyst. Simply conducting a few checks to brag about in front of a team or a manager doesn’t do the trick either. Becoming a  good analyst requires critical thinking and a flair for asking real questions. Of course, it also takes quite a degree of experience in the field and knowing the business you are analyzing like the back of your hand.

    Each business faces unique challenges and can rarely be compared with other business unless they operate in the same field as competitors. Regardless, there are a few questions that can be asked for any business and they work wonders at the initial stages of analysis.

    Let’s see what they are.

    #1 Is something missing from the data?

    Missing data
    Missing data, image courtesy of https://www.displayr.com

    Whether you trust the validity of the data gathered or not, it might be missing something. The most common causes for missing data links are found in the initial setup of an organization’s google analytics. To thoroughly examine the setup in GA, you should ask yourself some questions, such as.

    1. Is enhanced link attribution working?
    2. Are the default URLs up to date?
    3. Is Google Analytics properly linked to GSC (Google Search Console)?
    4. Are your google ads properly configured?
    5. Is the data from PPC campaigns appearing in GA?
    6. Are all sessions, impressions, and clicks recorded appropriately?
    7. Have your referral exclusion settings been saved properly and are working?
    8. Do you use demographics reports and interest reports?

    Of course, there are many other questions to be asked, but these are the main few that could attribute to finding what is missing in your Google Analytics account.

    Imaginary example of disconnected analytics

    Imagine a service has two separate apps – a web app and a mobile one. And the data between them doesn’t properly link. The AI algorithms would know you on the web app but won’t suggest the same pieces of content on the mobile one, completely losing you as a user. Eventually, if your habits are the same across both platforms, and if they use the same algorithm, they should catch up. But that allows the user to feel frustrated at first.

    Sometimes, using different tracking IDs for our users across multiple platforms can pose problems. Typically, in Google Analytics such cases are tracked by implying parameters in a URL, for example, the used UTM tags by GA are:

    • Content
    • Term
    • Source
    • Medium
    • Campaign

    Only two of these are 100 per cent required – the source and the medium. Of course, you can adjust that and add another tag as well, for example, the campaign one. This would allow you to track your campaigns and whether the amount of traffic is according to plan.

    A “Tag” means the identifier of the traffic of paid search leading to actual visits or clicks. Most advanced platforms attach these tags on URLs themselves and require nothing of you. But some don’t. It is important to research what type of service you are using.

    Moreover, user data might be missing due to technical difficulties. The most common ones found on websites are:

    • 404 Pages and 500 Pages (server ones) could miss a tracking code.
    • Erroneous or missing e-Commerce categories.
    • Behavioural reports experiencing fractured URLs.
    • Violation of Google’s terms and conditions. (For example, by collecting personal information without the need to do so and without the agreement of the user.)
    • Non-encoded ANSI Characters in a category or in a product name.
    • Non-recorded visits due to missing tracking codes on a certain page.

    #2 Is your analytics data accurate?

    Predictive Analytics, Image courtesy of Gartner

    Analytics audits should be performed routinely and with good intention.  Assessing the quality of your data should be an essential part of your decision-making process. Talking about Google Analytics Account-level data,  a lot of things can be questioned.

    In some cases, your proper set up could allow you to gain insights into the future, even. Of course, such cases are rare, and all future predictions are based on empiric evidence from the data gathered in previous models. Thus, they should be as close to 100 per cent accurate, as possible.

    For instance, if your GA data isn’t sampled you could bump into a number of issues arising from that. For some businesses the data precision level isn’t required to be as high and analytics are used mainly for directional purposes. Nevertheless, maintaining these 8 Google Analytics principles throughout the gathering and analysis of data is essential:

    1. Make sure you use standard reports.
    2. Adjust the data period range accordingly.
    3. Ensure that you decrease the traffic per property.
    4. Don’t be afraid to create new views and also make the most use of filters.
    5. Modify GA tracking codes.
    6. Big Query can be useful.
    7. Adobe Analytics could replace Google Analytics unless you’ve upgraded to premium.
    8. Get familiar with GA API best practices.

    Try to go into your GA Account and survey report by report whether everything is all right and if something is missing, try to figure out what is the reason. Moreover, if there are any coefficients that don’t look quite right, like for example a 0 per cent bounce rate, then you should examine what’s happening.

    A full health check of your GA account should be done at least once per year, and it could help you improve a lot.

    Check whether your GA data aligns with other software

    If your data doesn’t align, something might not be working properly. And figuring out which data set is erroneous should be of utmost importance, in order to quickly fix the issue. Google Analytics data sometimes doesn’t quite align with the data of the eCommerce tracking software you use. But how much of alienation should be considered reasonable? If your data sets differ one from another by more than 5 per cent, then you have a problem.

    For example, if your GA data says you’ve had 2000 conversions over the past month and your eCommerce system states 2100, that would be within reason since it’s only a 5 per cent deviation. Anything more than that should be considered excessive and the issue causing it should be found.

    #3 Do you have the right data?

    Each organization should create a list of questions they would like to answer with the help of data gathering and analysis. Simply collecting data for the sake of data collection isn’t always the best idea. Sometimes, you would end up simply wasting precious resources without actually getting any answers.

    Asking what each data point means for the particular business you are analyzing might put you on the right track.

    When performing A/B testing, your main goal is to keep all parts of the environment the same for both tests, whereas you change only one thing in the design or system, so that you can be completely sure that the change in the results is a clear reflection of that change. Otherwise, there can be a wide number of reasons, responsible for the results.

    What does this have in common with having the right data? Well, sometimes we overestimate the data we’ve gathered. For example, if you test two different campaigns but one is during the weekend, you might end up concluding one of them performed better than the other, without actually taking into account the different time during they’ve both run.

    This would often lead to reaching a hasty conclusion.

    ETL Process, Image courtesy of elderresearch.com

    More often than not, we have the right data, but somewhere along the chain of transforming, loading and validating it, a malfunction might occur. Thus, having a precise plan of what needs to occur in each of these steps is a must.

    #4 Have you done an Audit to your GA?

    Performing QA for almost any aspect of your website, app or business is essential toward optimisation. Be it campaign management, A/B testing, design, or simply analytics and data gathering, you need to test your process and make it better one step at a time.

    Of course, the case might be that you don’t find anything missing or malfunctioning in 90 per cent of the cases, but when you find it, it totally pays off for each and every unsuccessful QA attempt.

    Erroneous data could lead to:

    • Bad business decisions
    • Loss of revenue
    • Conversion rate decrease
    • Loss of brand image

    Moreover, if you are an analyst, and you vouch for a certain data set, that eventually proves to be imprecise, the marketing team would have to take responsibility for the data on your behalf. Data sets and analytics are used for major decision making in big organizations, thus second checking everything shouldn’t be frowned upon.

    Numbers that look high off the charts should be closely inspected. If the usual conversion rate is 12 per cent and you see a whopping 35 this month, well either something is off, or something has been changed by another team and you haven’t been informed. The same goes for much lower numbers than usual, as well. Don’t take any value for granted.

    Know the usual mean of your data points and look for outliers and proofread all data analysis reports.

    How to proofread reports

    • Survey each query separately by pulling it up in the exact tool you used for its creation. Check what’s included and what’s not.
    • Re-check the data in the tool of creation by comparing it with the respective data points wherever you are translocating the data.
    • Proofread whether you haven’t swapped the data string somewhere along with the report.
    • If there are any correlations between data points and coefficients, formulas in the form of scripts can be devised to inform you whether something is off the charts.

    #5 Can We get more out of the gathered data?

    Levels of Data Governance, Image Courtesy of Gartner

    If you take a close look at the image above, there are 5 levels of data governance in terms of maturity, according to Gartner. The more data-aware and advanced a certain organization gets, the more insight they have, and the more control over the future course of actions is obtained.

    At the start, you would experience uncertainties and question your data. The more defined your process gets, the more credibility your data would yield, eventually becoming the basis of every major business decision of your organization.

    There are questions you can ask anywhere along the curve of data maturity. Some of them are:

    • Are your conversion points properly defined?
    • Are your KPIs set? Do you have all the dashboards you need?
    • Are all meaningful events on your website correctly mapped? Is there something you would like to track in terms of customer behaviour but you don’t?
    • Is your user segmentation optimised?
    • Can you view the entire customer journey? Does your data allow it? If not, perhaps consider merging data sources.
    • Is your model taking advantage of behavioural targeting or predictive analytics?
    • Are you using behavioural personas that are data-driven in targeting?

    While these are just a few questions you might ask anywhere along with the development of your data gathering, storing and analytics process, they can help you set the basis.

    Data is only useful if there is the intention behind gathering and analysing it. There is no point in tracking absolutely everything that goes within a business unless there is a model that requires it. It might be confusing, and time-consuming, and still, fail to yield fruitful results. Thus, at first, be advised to limit yourself to the coefficients and data points that are most important to your business and that you are most familiar with.

    Final Words

    Asking the right question is the essence of understanding data analytics and making most of it. Avoid data puking just for the sake of reporting. Question the accuracy of every single metric, especially one of outliers. With proper auditing and a modelled process, you will optimise the data analytics of your business in no time.  After all, having data is all about giving you the confidence and information needed to make the right decision.