Marketing Forecast – Calculating Error Rates
Posted on 06. Oct, 2009 by Franz in Web Metrics
Calculating Error Rates for Marketing Data
If you’ve been a marketer hardcore enough, you’ll probably realize there is a huge need for information – Particularly in your own demographics and location because most of your clients would probably have a market base here. Data can be collected from different tools and they often have a lot of error rates. Why? Because we’re using fremercialized products in the market.
Cookies & 1×1 images – That’s why.
I’ve said it many times in many tweets that configuration for Google Analytics is important. Also, your data will most likely not tally because of the behaviors each tracker sends and receives cookies. For example, if GA sends a cookie to http://www.elioe.com/ when visitor A visits the page, it’ll also send another cookie to visitor A if he visits http://www.elioe.com/franz/ or http://franz.elioe.com/ or http://www.twitter.com/franznarcis/. Why? Because data sharing is not enabled. By default, GA disables it for many reasons.
You can enable data sharing for cross-domain, cross-directories and cross sub-directories in your website or network of websites.
In any case you’re feeling worried and do not know how to show charts and figures from 5 different trackers, do this first: Imagine you have 5 different trackers, which will be named 1, 2, 3, 4 and 5.
Lay all 5 trackers on the table and you should see similar terms: Visits, Visitors, Bounce Rates, Time/Visit. Use these four (4) and compare it with your marketing goals.
Calculate its error rates – Take an average of those figures from different trackers and multiply it by 100%. Get the percentage. Now, manually multiply each tracker with a -5% (you’ll know why later) on your total error rate percentage. Meaning to say:
If average error rate = 26%, then (26-5)% = 21%. If visits = 1,000 then [ 1000 - (21/100) ]. Thats for Tracker [1]. If you want a global commons error rate for each tracker, you’d need to take an average of at least 2 months data.
That answer is your prudent answer to your boss. Showing lower visits, higher bounce rates (don’t use bounce rates, use the amount of times a page is a landing page), lower time on site and x% of new visits often gives you both a pressure to push further. And to ‘PRUDENTLY’ see what has been achieved so far.
P/S: Usually error rates will not reach 26%. It’ll only be between <0.5 – 2%. But if you use tools like Alexa Traffic/Ranking count vs. Traffic Estimators, you’ll get an error rate of 21% – 38%.
Forecasting Through Error Rates
Forecasting is never easy – It takes a lot of understanding, time, data and factoring in order to predict what wave will hit you next. And if you’re in the technological wave, it’s even harder to expect what’s next. But if you’re forecasting for your marketing campaign’s goals, it could be much simpler than expected:
[ Profile #1 | Profile #2 | Profile #3 | Profile #4 | Profile #5 ] – Obtain error rates from 5 different trackers.
Each profile can represent a marketing campaign, a marketing action or a marketing practice performed for a particular brand. What you can do is take the error rates to calculate tolerance (also called ‘allowance’ in some cases) rates – Tolerance levels are ‘in-between’ figures, where they fluctuate closer or further using error rates as a base.
So if GA yields 10,000 visitors but Woopra delivers only 9,650, then GA. vs WP. is at a -1.036% (Profile #1). Then GA delivers 15,000 visitors but Woopra 16500 visitors, at 0.909% and so forth all the way to Profile #5. Then you’ll get the tolerance levels. It could be -1.036% — 1.987%.
So once you have historical data of at least 4-6 months and you have calculated error rates vs. *x Metric (Visitors, Visits, etc.), you can balance tolerance rates for the forecasting months:
Prucedency - Playing safer or taking larger risks in presenting data
Your choice! =)
