What you should know about Historical Data in Web Analytics

For many of us, historical data in web analytics is very important because it help us in optimizing our marketing campaigns and efforts.


We often do time travel to get answer to these burning questions.

But the problem with historical data is that it is historical. It is dated.

The insight that you get from analyzing historical data is out of date and it does not always match with the present marketing conditions. The older the data, the more it stinks.


What are my top selling products?

Let us answer this question. Your top selling products are the products which sold the most last week or at most in the last one month.  They are not the products which sold the most in the last 3 months and certainly not the one sold in the last one year.

This is because we live and operate in a constant change of marketing conditions, trends, buying behavior, pricing, competition and multi channel funnels.

For example, let us assume you concluded:

our top selling products are ‘A’, ‘B’ and ‘C’ according to our last 3 months analytics data and we should continue to invest in them”.

But what if the product ‘A’, ‘B’ and ‘C’ are now out of stock or are not sold any more?

What if the sales of these products are declining now but you continue to invest in them because they sold the most in the last 3 months?

What if the potential sales of these products are bleak because a strong competitor say ‘amazon’ has made entry into your market?

What if you no longer run that email campaign and that Facebook campaign which assisting in lot of sales through organic search during those 3 months?

There can be lot of ‘ifs ‘and ‘buts’.

The important thing to remember here is that assisting marketing channels (channels which assist in conversions) change all the time and we often overlook them and rely only on the marketing channels which directly completed the conversions.

Note: If what I am saying is not making much sense to you then I strongly suggest you to read this post first: Attribution Modelling in Google Analytics and then The Geek Guide to implementing Attribution Modelling.Both of these articles explain attribution modelling in great detail.


Possible reasons of change in assisting marketing channels/mediums over time

1. You got a guest post opportunity on a big authoritative website and as a result you got lot of exposure, signups and sales.

2. You changed your website design which dramatically improved the conversion rate of your website.

3. You sold a product which was high in demand but short in supply.

4. Your link bait went viral which resulted in lot of exposure and conversions.

5. You bid on new keywords which turned out to be very profitable because of low competition and CPC.

6. You fixed serious website architecture issues which dramatically improved the crawling and indexing of your website and eventually resulted in more conversions and sales.

7. You captured new top positions both in organic and paid search for number of profitable keywords.

8. You started leveraging two new marketing channels to promote your contents: Slidershare and Linkedin


Now the problem is

History does not repeat itself in online marketing.

It is highly unlikely that you can replicate your success rate by carrying out the aforesaid task once again in the ever changing online landscape.

There is always a possibility that this month:

  • Your post may not get published on an authoritative blog.
  • There may be no changes to the website design.
  • Your top selling product may not be in short supply any more.
  • Your link bait doesn’t go viral.
  • Your profitable keywords are no longer as profitable as they used to be.
  • You may no longer have any website architecture issue to fix which can dramatically improve crawling and indexing of your website.
  • You may get new competitors who capture some of your profitable ranking positions both in organic and paid search.

There is also a possibility that this month a completely new set of marketing channels assist in your conversions.

All these hidden channels/mediums impact conversions and that is why your organic and paid search campaigns perform differently each month even if you don’t directly make any considerable changes to them.

So if you optimize your marketing campaigns and overlooked the role, assisting marketing channels played in your historical data then you will attribute conversions to wrong marketing channels and loose money.


So what is the solution? Should you discard historical data completely?

No. Do not discard historical data but look at the most recent one. As a rule of thumb,


Rule #1: The more you segment your data, the smaller should be your time frame for historical analysis.

Rule #2: The more you look at the data in an aggregate form, the bigger should be your time frame for historical analysis. 

For example,

If you want to know your top selling products then look at most, one month old data. In this way you can minimize the impact of ever changing online landscape on your marketing and business decisions and get optimum results from your marketing campaigns.

If you want to know your top selling product categories then look at most, 3 months old data.

If you want to measure the  quarterly or annual growth of the entire business in terms of sales, leads, total visits etc then choose bigger time frames like 4 months, 6 months or 1 year. 

One year is a very long time period in online world and any period longer than that can be considered pre-historic.


Comparing one year web analytics data to the last year can be like comparing apples to oranges because so much would have changed during that time from website size, traffic, products, competitors to your target market.

If your client/boss insist on using historical data with longer time intervals then educate him about the ever changing online landscape, competition and changes made to your website and campaigns.

Other Posts you may find interesting:

Subscribe to my blog
Join my free newsletter and learn to avoid the analytics mistakes everyone seems to be making over and over again.


About the Author:

My business thrives on referrals, so I really appreciate recommendations to people who would benefit from my help.Please feel free to endorse/forward my LinkedIn Profile to your clients, colleagues, friends and others you feel would benefit from SEO, PPC or Web Analytics.



  • http://www.dabrianmarketing.com/web-analytics.html Brandon Wensing

    Interesting blog. I especially enjoyed the list of examples of how and why yesterday’s tactics might not work tomorrow.

    That being said, I still believe historical data beyond a month prior can still provide some great insights, specifically when it comes to assisted conversions that might be affected by seasonal trends.

  • AnalyticsNerd

    Awesome post Himanshu. We should stop worrying about backward trends in this digital age esp. when the online landscape changes every second. However we often need to dig out historical data with large data ranges (6 months, 1 year, 2 year) to satisfy the fat ego of our senior managemet. It is difficult to changes their mindset mainly because they have been doing such type of historical data analysis for decades even before the advent of ecommerce. Great work. Looking forward to see more analytics posts from you.

  • Jason Spencer

    Great post as usual. I love the way you answered the top selling products question like an analytics pro. Very insightful and valid. The web is changing so fast that we need to switch to almost real time web analytics. Majority of marketeres underestimate how fast things are changing around them. They tend to rely on year old trend which in majority of cases is completely outdated, decayed and provide little to no value in predictive analytics.

  • http://counterforsite.com/ Counter for site

    well yeah in this competitive world historical data may useful but not enough.Specially data older than a year can’t use for building new strategies because trends,competitions changes day by day.Very informative post.Thanks for sharing.