Did the last meeting about the data generated by your business activities not reflect what you’re actually experiencing? If you suspect that numbers don’t add up reflect, your instincts may be right. By enhancing the quality of your data, you can start to see a much clearer picture of how your business is performing. To help you, I’ve summed up a few useful recommendations. If you haven't read Part 1 yet, you can find the blog post here.In my previous blog post, I listed a few things you might pay attention to:
In this second blog, I expand on these key points:
You have one report but how many data sources do you use to compile it and how do you collect them? There are various sources of data; you can upload your data manually or automatically.
Metrics are the units that will determine how your data will be displayed. How do you usually measure your data? What is the timeframe? What is the currency?
If you are active in a local market, it might be easier to manage your data metrics than in an international context. However, in order to have the big picture, we'll review an international company so you can have a good overview of the actual challenge I'm trying to demonstrate with data metrics.
Let's assume you're in charge of creating reports for three markets: CH - UK - JP. You will need to adjust the currency level (most of the time international companies tend to have an internal fix-rate, so there's no market fluctuation during a fiscal year). Another example, which might be a bit more tricky, is to compare conversion rates. I’ll need to illustrate this with some mathematics:
The CH market does an internal weekly conversion rate report, the UK does it bi-weekly, and the JP does it monthly. How are you going to calculate to compare the data? Are you simply going to add them up and divide by the combined timeframe ? This method would be wrong, and the result would look like this:
|Week 1 - 2.5%||Week 1-2 - 3.3%||Week 1-4 - 1.9%|
|Week 2 - 2.6%||Week 3-4 - 2%|
|Week 3 - 3%|
|Week 4 - 1.8%|
|Total conversion rate 2.475%||Total conversion rate 2.65%||Total conversion rate 1.9%|
You want to "flatten" your data first and dig deeper into the metrics used to calculate conversions, meaning that you should also consider the units and data driving that conversion rate. Let's assume it was lead/unique page view.
|Week 1 - 2.5% (lead 25/page views 1000)||Week 1-2 - 3.3% (lead 33/page views 1000)||Week 1-4 - 1.9%|
|Week 2 - 2.6% (lead 52/page views 2000)||Week 3-4 - 2% (lead 100/page views 5000)|
|Week 3 - 3% (lead 120/page views 4000)|
|Week 4 - 1.8% (lead 9/page views 500)|
|Total conversion rate = lead 206/7500 page views = 2.746%||Total conversion rate = lead 133/6000 page views = 2.216%||Total conversion rate 1.9%|
As you can observe CH market has actually a better conversion rate once the data has been flattened. The conclusion is that you should always "flatten" your data with common metrics (convert them into one common metric).
Let's focus now on data visualization and how are you going to demonstrate your learnings in the best possible way.
Tables are probably the most intuitive way to display data as you can really add any type of data with different dimensions. This is a good way to have an overview of your business’performance. My recommendation would be to keep tables clean and display only the most important information. It's not much use if you have a report with a table with 2000 lines or 2000 columns.
This is a great way to show different "products" on one single chart with a timeline. Bars are the best way to showcase if you have a total to calculate for multiple dimensions grouped in a single chart. Let's illustrate this with the evolution of annual sales for three different products from your portfolio and you want to see how they perform.
A pie chart will only display a full 100% set of data. This is a great way to have a comparison in a bagel/cheese representation (depending on your culture and preference) about one unique dimension! We use it to keep track of relative values and compare them to each other. If you have a dimension with 25 value sources, the pie chart might not be the most suitable one as you'll lose track of the smallest values. I wouldn't recommend it if you have above eight different values––unless you want to group them together if some can be combined. I would also avoid it if I know that 90% of my pie chart will be absorbed by one value (it's like Obélix eating a pie, for those who know "Astérix & Obélix").
Illustration from "Astérix et Cléopâtre", by René Goscinny & Albert Uderzo
It is similar to histogram and bar charts but gives a better representation of one period to another in terms of evolution (increase/decrease). You can have line charts with one dimension or multiple dimensions as well. You must remember to be careful when you combine all these dimensions, however, so that they are the same along both the X and Y axis.
Figure 1: example of Geomap with intensity dots to measure two dimensions at the same time
Figure 2: example of gauge to keep your KPI under control
Thanks for reading! This second part was a bit more technical than the first article, but I hope it gave you a better understanding on how to handle data and avoid mistakes on your report. If you're interested to work on a project with us, just click here:
Michel Pham | 17 Dec 2020
Have you ever been in a meeting where all the reports show you outstanding numbers and only green lights? You wonder how that could be possible when your experience on ...
reading time: 9min
Florian Schneider | 8 Jul 2021
Are we the team for you?
reading time: 8min
Winky Fung | 23 Sep 2021
There are 4 billion daily email users, and somewhere in that crowd are your customers. So, you do what all good inbound marketers do: you send lead nurture emails. ...
reading time: 8min
Transform to perform
For mid-market and corporates in the DACH Region that want to transform their business in order to grow locally and globally, BEE is the highest-ranked digital growth solution partner that combines strategy, execution, data intelligence and people in a transformance model.
What BEE stands for