Blended data feedback


#1

Summary
I have seen that the way to blend data takes a lot of clicks, and also is too technical.

Too technical meaning, not many people might be able to write SQL queries to blend data. I can, but you are limiting this functionality for the highly technical users. In this case, there are several challenges. First, the technical background of the people. Maybe include a visual builder, something like creating charts and drag and drop, similar to how you create a single chart, but try to do the same technical stuff in the background, but with a simple way.

Also, there are problems on which the data is not on the same format. For example, the screenshots below are sessions from Google Analytics and clicks from Google Ads. I want to have a correlation between both, and it seems I cannot do it, because the format of data is not the same. I can very well do this on excel, but then, I have to create an excel and just create a simple chart. I understand the technical aspect on which you need a primary key to blend the data, but in this case, this is not a relational database, this is data visualization. Maybe date can be a single metric on the X axis, and what I want to measure on the Y axis, for example.

Lots of clicks, since I have to create a chart, then another chart, then the SQL, then try, etc. I suggest include all in the same step.

This is a great addition, but the technical stuff might be tougher for some users and also makes me just wanna use excel and do a simple graph. If doing blended data in here is easier to export to excel, join information there and then do a simple chart, then using this functionality works. Right now it seems the solution takes more steps than what I am doing right now.

Screenshot

Google Analytics

Google AdWords


#2

Thanks for the detailed feedback, @epiclemon. Really appreciate it!

I agree SQL is not for everyone. Even for people with basic SQL skills, doing a proper join is not trivial. We do have the plan to add a visual builder to make this functionality available to a broader audience. Maybe these two things will coexist eventually (and blended to some extent), serving people from different background.

For the data format mismatching, right now the tables simply inherit data types of fields from the original data source. For Google Analytics, “Day Index” returns “Day of the month” as plain text. If you want to join with Adwords, it’s easier to pick “Date” field from Google Analytics side instead so that they’re both date type. Alternatively, you can also keep the current configuration, but use SQL functions to extract “Day of the month” from Adwords side to join with GA, which is more complicated but it should work. Our SQL is mostly compatible with MySQL so most MySQL built-in functions can be used. That’s the power of SQL isn’t it :grinning:?

For “lots of clicks”, @Frank_Marent gave similar comments in the other topic and I totally agree the current design is quite verbose and doesn’t have enough shortcuts. We’ll think hard on this.


#3

Great, thanks! Yes, I know that many stuff could be done with SQL, but again, I don’t see this as a relational database and might be too much to write a complete SQL function, rather to just export, do manual work and then use a regular chart. I am aware that this means no automatic refresh, but again, depends on which scenario we are talking about. In my case, even if I can create SQL and can dust off my knowledge on this, it is a greater effort to use this right now, than to continue as I am doing manually right now. I get the feeling that the solution is more complicated than the workaround I have at the moment :wink:


#4

Yes, for simple join creating a Blend is a bit overkill. A possible more light-weighted solution is to directly correlate two pieces of data in a single chart - something also in our radar :sunglasses: