What does Discrepancy mean?
Discrepancy usually means an unexpected (or unexplainable) difference between two sets of information. Unfortunately, in digital marketing, a discrepancy is always to be expected between any two sets of analytics platforms.
This is because, although they often use the same terminology to describe a stat (eg “Page View”) different analytics programs will often measure these items slightly differently (eg when a page starts to load vs when a page finishes loading). This will create an unnoticeable difference for a single user, but across a large number of users, the difference in stats will become large.
Even when both analytics programs work in the same way, one always has to load after the other, and a myriad of issues could cause one to not work occasionally.
What it looks like
If you’re still wondering why a discrepancy is occurring between your analytics platforms, there are simply too many reasons to list them all. It is certainly not the case that there is always someone to blame. Many times a discrepancy occurs simply due to a unique set of circumstances.
In general, you would expect discrepancies between different analytics platforms to remain somewhat stable. This doesn’t mean that month-on-month you should see exactly the same discrepancies occurring, but there should be a stable range (5-10% for example). However, if you start seeing this range dramatically change, then there is likely a problem that needs addressing.
Here is a list of some (but not all) reasons why discrepancies might occur:
- A user leaves before a page fully loads
- One or more set(s) of tracking is incorrectly implemented
- Tracking doesn’t work on some devices/browsers
- A changing element of the site (eg advertising) affects tracking, meaning it is disrupted sometimes but not others
- Geo-targeting means some element of tracking is occasionally disrupted
- An update to some part of the system has broken some tracking
- A large increase in traffic to a site has overwhelmed servers
- Two sets of tracking interact oddly
- A tracking service is being blocked by some ISPs
- A large image (or another element) means a page loads slowly and times-out
Advice For Website Owners
Choose a measurement and stick with it when reporting. If you report Google Analytics for one month, make sure you use it again next month. Otherwise, you are adding in a level of randomness for no reason.
When selling advertising, again use consistent stats for reporting. In most cases the advertiser or ad network will have a preferred method (often Comscore) for reporting, so make sure you consistently use that.
Your stats will also be inconsistent with the advertiser’s stats, so always ensure that you get reports sent over from them throughout campaigns (and within one week of starting). On top of this make sure you agree beforehand which (and whose) stats you are using to record information. This will help you make sure you do not think you are hitting all your targets, only to find out the advertiser is refusing to pay for anything as their analytics don’t work.
In the USA it is standard to use advertisers reporting for payments, but in the UK it is still mostly standard to use publisher or ad network stats. There are exceptions to the rule in both cases.
Advertisers will, of course, want to use their stats, and agreeing to do so can help you close deals. However, if you do agree to this, you should plan in at least a 20% buffer zone (as in selling 20% less inventory than you think you have). Keep checking their stats against yours throughout any campaign and adjust delivery accordingly. This is the only way to make sure you don’t under-deliver.
It should be noted that conversion data is especially prone to discrepancies, as the advertiser will be de-duping data, whereas you will not. This means the advertiser will almost always be recording fewer conversions than you are. Sharing reporting is therefore especially important in CPA campaigns.
Advice For Ad Buyers
Ask for regularly scheduled reports from anywhere you advertise to minimise discrepancies.
It is, of course, advantageous for ad buyers to insist that any campaigns you run are paid out based on your own third party reporting. This can be a large concession for website owners/ad networks so treat it as such – you are essentially asking for up to a 20% bump in inventory. This is because while your reporting may more accurately record how many times your ad is shown or clicked on, it will likely have lower numbers than the publisher records. This is because your tracking pixels will load after the publishers in most cases.
It should be noted that demanding to use your own statistics for billing will also hamper publishers optimising efforts. If they are optimising towards clicks, for example, they won’t be able to optimise towards the clicks you record, only the ones that they do. To try and keep campaigns running as well as possible, you should definitely provide as much reporting as you are comfortable with. This includes broad stats from other people you are advertising with (if you are happy with that) as this helps create competition between the publishers you are working with.
It may seem like getting this extra inventory is a nice bonus, but if you take advantage of the system too much it will backfire. Publishers and ad networks will be measuring their own eCPMs to check if your campaigns are worth it. Trying to pull a fast one on the people you work with (especially if they provide good performance) is not a good business strategy.
Please note: conversion data is especially prone to discrepancies, as you will be de-duping data, whereas the publisher will not. This makes sharing reporting especially important, as you want them to be working towards as many conversions as possible.