How did everyone get it so wrong? Again.

donald trump

Just like Brexit, last year’s election, and Honey G. Everybody was wrong. Again.

Unless you’ve been living under a rock, you will have know that practically every political insider, “trusted source”, pollster, and analytics nerd in America has been left with an egg in direct alignment with their face after they predicted that Hillary Clinton would trump Trump and waltz into the White House.

However, the UK arose from their collective slumber this morning to see that those projections had been left in tatters by the man with the ‘fun-size’ hands — just like the ones in June that all-but ruled out the possibility of Britain leaving the EU.

The future POTUS’ surprisingly strong performance in the election has left analysts and experts struggling to explain how everyone could have been so far off the mark. The result also seems to at least partly prove Mr Trump’s claims that many polls “just put out phoney numbers.”

Although surveys may not exactly pull numbers out of thin air, the inherent limitations of polls* have led experts to proclaim that “they just don’t work anymore.”

Analytics are good, but not perfect.

The same can be said of your Google Analytics data; it is all-too-easy to be fooled by complicated data and just take things at face value. In our experience, most organisations we come across either under-utilise, mangle, or ignore their analytics data altogether.

Is Your Analytics Telling The Truth?

Below, we have listed the top 5 mistakes companies can make with their data, and how you can avoid these in your own analytics.

1. Relying on nebulous “User” data

You are presented with a myriad of metrics and dimensions in Analytics that are intended to provide you with an accurate view of how visitors to your site are behaving. User/visitor-based KPIs have never reflected actual users.

One such example is the New vs. Returning visitors report in Analytics. We regularly see this metric in SEO reports, but this paints a false picture as it falsely assumes that visitors only ever use one device and one browser on that device to browse the web.

Now, as we all now this isn’t true. This metric categorises a visitor as a particular combination of device and browser someone visited your site from. Therefore, making reporting on them a wasted KPI.

The only way to safely use user-based KPIs is if you meet the following requirements:

  • Users can log into your site
  • Users ACTUALLY log in to your site
  • You use Universal Analytics
  • You have modified your GA tracking code to capture a user ID
  • You have a suitable view set up that uses user ID

If you do meet the above, then make the most of them by utilising the plethora of custom dimensions that will add real value to your data. If you don’t meet the above requirements, then we’d recommend moving to session-based KPIs for your reporting, of which there are plenty.

2. Not getting your referral data in order

One of the biggest issues clients have within Google Analytics is with referral data, either in the form of self-referrals or referrer spam.

With regards to self-referrals, these have been a significant problem for a long time. In a nutshell, sometimes your Analytics can show your own site as being one of the top referrers to your site, which obviously doesn’t make sense.

To make matters worse, if this happens, Google will count that referral as a new session, which again will skew your data. The same is also true if you send someone to a third-party site (i.e. to PayPal to process a payment or Eventbrite to sign up for an event) – this will again trigger a new session when they return and land on your confirmation page.

To get round this, you can use the Referral Exclusion list, which contains domains that you do not want visitors who arrive at your site from to trigger a new session. Therefore, if you do not want traffic that comes from a particular domain to trigger a new session or show up as a referral, you need to include that domain in your Referral Exclusion List.

We also see a number of clients who are plagued by referral spam. We have already covered how to remove referrer spam in a previous post, but not doing so can falsely inflate the number of visits to your site, thereby giving you a false impression of how well you are performing.

3. Not taking your data with a pinch of salt.

When Google says jump, it can sometimes be hard not to jump. Often we take Google’s word as gospel, but just because Google tells you something, it doesn’t necessarily mean it’s true.

You should always approach your data with a critical eye; otherwise, you risk making poorly-informed decisions (like all those people who would have put their house on Clinton winning last night).

To use a real-life example, we had a bricks and mortar client in Newcastle who suggested to us that because a large number of sessions were coming from London, customers would be willing to travel all the way up from London to purchase their products.

It wasn’t until we pointed out that this anomaly was more than likely due to some servers being based in London, that they realised that trying to expand into London might not necessarily be the best idea!

So the key takeaway for this is ALWAYS QUESTION YOUR DATA! One of the most commonly-used phrases in our office is “How do you know?”

If any insight seems surprising to us, we question it, and you should do the same.

4. Having data is more important than using it to drive clear actions.

I’m not sure which is worse; using unreliable data to make poorly-informed decisions, or just collecting data and not using it for anything at all.

Either way, collecting data for the sake of having data is completely pointless. It chews up time that could be better spent running campaigns, refining strategy or creating awesome content.

Moral of the story: Don’t let your data go to waste. When you are looking at your Analytics data, make sure you have a clear plan on what you are going to do as a result of what it tells you. If you don’t know how to use the data you’re collecting, then there’s probably not much point in collecting it at all.

The most important part of any reporting in Analytics is being able to extract actionable insights that will lead to clear actions for your organisation, for both current or future campaigns.

5. Reporting on every single metric.

This point logically follows on from the previous one as, in theory, if you had the inclination then you could spend 100% of your time reporting on every single metric available in Analytics. This might sound far-fetched, but we have seen some monthly SEO reports that made War and Peace seem like a Haiku.

There is an endless rabbit hole of data that you can dive into to conduct analyses. However, not all of it is relevant to your business. So before you perform your quarterly/monthly/weekly/daily/hourly analysis, note down the questions that you to answer with your analysis:

  • Are there any new markets that we should be targeting?
  • Are there any issues with our mobile experience?
  • What is the impact of assisted conversions?
  • Which marketing channels are our unsung heroes?
  • Don’t get carried about by reporting on everything; try to stay focused.


In summary, marketers and political pollsters alike often rely on data as a means of reaching the truth.

However, even though data is still one the best tools we have at our disposal, it is critical that we understand and respect just how difficult it is to get a valid result and not to take our results as an edict.

*Seeing as we have no credibility in the field of political science, we haven’t dissected the failings of the US poll forecasts, but you can read about them here, here and here.


The Digital Allies newsletter brings you free resources and new thinking about all aspects of digital marketing. By signing up for it you will get notice of our events before anyone else and we will only be in touch if we have something that we think you will be interested in.  You can unsubscribe at any time. See our privacy policy for more information on how we take care of your information.