Trying to make sense of the amount of data you collect in Kissmetrics can be daunting, especially if you notice a difference between our data and data from your databases or another analytics platform.
This page discusses the following:
- Reasons we might not capture 100% of the data
- How to determine where the root of the discrepancy may come from
Kissmetrics, like all analytics vendors, makes its best effort to achieve 100% accuracy. However, despite all of our best efforts, there are many things that are not within our control.
- Slow connections
- Browser software that is outdated or has limited features (mobile browsers)
- Privacy-protection software
- Browser settings or modifications
- Spiders and bots that can throw off counts
- When one person uses multiple computers or browsers
- When multiple people use one computer or browser
- DNS issues
- Hardware failures
Much literature is dedicated to these challenges. But in spite of the challenges, we still continue to try to innovate and get closer and closer to the goal of delivering 100% of your data.
There are some events, especially Revenue-related events, where it would be very important to ensure that we capture every instance. For these events, it’s effective to use our server-side libraries or other integrations, if possible. Though this means it can take more time to implement, you get two main benefits:
- You have a definitive source of data (your own database) on which to ground your Kissmetrics data.
- Our server-side libraries include mechanisms for retrying the sending of data, if for whatever reason our tracking servers did not receive everything on the first attempt.
If you suspect that Kissmetrics’ reports disagree with the data you were expecting, these are the types of investigative questions our support team would use to approach the situation.
For almost all of our tools, you can segment all the way down to the individual person level, by segmenting by Customer ID. At this point, our support team looks for some sample people who are affected by this data discrepancy to try to find patterns in what was recorded, what wasn’t recorded.
Additionally, our People Search lets you narrow in on sample people affected.
Of the people you’re reporting on, are there any patterns you can see?
First off, are we looking at the right event or property? If you weren’t the one who originally implemented all of the events, you might discover some similarly named events in your account. It’ll help to sync up with the person who implemented to get on the same page.
- Here’s a table of the total counts of all of your events. This lets you see the most recent instance of the event, to help you differentiate between two very similar events.
Does this event represent the action I’m interested in?
Let’s check if the events are properly implemented. When you step through the expected flow, does your activity trigger the events?
- Perhaps the easiest way to verify your events work is to use our Live tool. If you leave Live open, you can step through.
- For our more advanced users, you can even look for network activity to our API. You’ll see this in the form of HTTP requests formatted according to our API specifications. This is the format that you’ll see in the logs generated by our server-side libraries.
- If you’ve set up the events through the Event Manager, here are some Event Manager tutorials that may help indicate whether you have set up the rule properly or not.
Do we consistently receive an event when someone does this action?
It is often overlooked that your Kissmetrics events won’t be fired when your site redirects users to a third party Authentication or Payment platform. EX: Facebook OAuth, Paypal. This occurs when users are directed to a third party domain.
As a result, for data like how many users signed up, logged in, or paid, you will see a discrepancy between the numbers Kissmetrics reports, and what your other resources report.
There are several ways to handle this situation:
Use the Funnel Report to track the steps involved. For example, you may create a funnel that counts how many people viewed the signup page -> how many people actually signed up -> how many people viewed the confirmation page. In our scenario, you may see 1000 people viewed the signup page -> 500 people completed the signup process -> 800 people viewed the confirmation page. It would be safe to say 300 people completed the signup process via the third party platform in this case.
Use the third party platform’s SDK, if they have one, in conjunction with Kissmetrics API calls to fire the events. Here is the support doc we have created to demonstrate how you could measure Facebook logins and signups. The same principle applies to any other third party tools, as long as they provide ways to return the status needed.
Import the data that corresponds to the events in question from the third party tools to Kissmetrics. You may refer to this support document for the multiple ways you can get the existing data into Kissmetrics.
If you’d like some help with the above steps, please contact our support team at [email protected]! We’re happy to help.
What Other Analytics Professionals Say About Data Discrepancies?
Ensure your analytics strategy is to reduce data inaccuracy as much as possible. Don’t focus on getting 100% accuracy. That does not exist. I believe that. There are no “right numbers”. I am a part of a generation that believed that and we always tried to reach that goal (I grew up in the world of data warehouses and business intelligence and ERP and CRM systems).
But painfully I have learned that you can either focus on that, or you can use the data you got. Analytics data gives you 900% more information than you have through traditional channels. It is only 90% “right”, but the missing 10% is outweighed by the fact that you can now make decisions that are so much better informed.
It will take a while for the current crop of business leaders to “get it” – and sadly many many web analytics practitioners/consultants/vendors to get it. We need to realize there is more money to be made not peddling our services that make things “accurate” but rather peddling our ability to find raw awesome insights (whatever the tool the company has).
example taken from Avinash Kaushik’s (Author of Web Analytics 2.0, Web Analytics: An Hour A Day, and Digital Marketing Evangelist for Google Analytics) blog post on data reconciliation
If there are any known site-wide issues, we will post a notice to kmstatus This message will also appear at the top of the website when you log into your account.
Updated 7 months ago