Analytical history - how long?


(Bas van Leeuwen) #1

When you look into your analytics history (google analytics, community analytics, whatever), what is the absolute maximum that you ever go back?


(rhogroupee) #2

I have done some very occasional year-to-year comparisons, but my norm is more like the past 6 months. Good question, I’ll be curious to hear what others say.


(Darren Gough) #3

We had 10 years worth of data at MSE. Was useful to understand how certain campaigns worked and how they changed over time. I think in practice we probably used about 3 years worth of historical data for most things which felt a good way to do year on year and allow a little for how things adapt over time (culture etc).


(Bas van Leeuwen) #4

Would anyone be surprised/offended if analytical software only showed the past two years?


(Sarah Hawk) #5

No, that would serve my purposes. I think you’d need to make sure it was clearly communicated when people signed up though.


(Bas van Leeuwen) #6

Wouldn’t want to create false hopes, true!


(Darren Gough) #7

2 years sounds reasonable. Only time it might be a problem is if it was a community around something that happens every X years (general election/government is the best example I can think of) and they’d need to compare like for like data.

Seems pretty niche though - i think for most people 2 years is very workable


(Richard Millington) #8

For some background, what are you trying to do with the data? Context would help us here I think.


(Bas van Leeuwen) #9

Fair enough, didn’t share the background because I don’t want to be overly promotional :smile:

I’m working on a platform to analyse communities; for the interface we need to make some decisions about what to keep, what to show etc. So i thought I’d ask our target audience what they’d like

I think we’re going with a 2-3 year history, that makes presentation and selection of the data a lot easier. Probably give the possibility to export the data from whenever you want so that the outliers can get their data as well.


(Gear Buzz) #10

Can you take a csv file of past data and dump it into Google Analytics?


(Bas van Leeuwen) #11

We do provide .csv exports; haven’t tested it against the GA import
functionality, but that should work as far as I can see :slight_smile:


(jennita) #12

Letting people know up-front is great. Plus if you make it easy to download the data, and perhaps yearly email reminders would be nice as well.


(Bas van Leeuwen) #13

Ah yes of course! That is obviously a good choice… and one I didn’t think of :slight_smile:
(And let’s just send end of year CSV files attached while we’re at it)


(Bas van Leeuwen) #14

Actually, I just found a counter-example where 2 years isn’t quite good enough. It’s a seasonal forum, and the trend is much clearer if you look at four years than if you look at two. You can’t truly see how well you are doing until you compare it to a couple of other peaks.


(Adrian Speyer) #15

I know Google used to have a 25 Month limit of data. They could keep data beyond that point, but only their premium service was guaranteed though. Maybe offer more as an extra feature :slight_smile:


(Steve Bridger) #16

The think the main concern is as time passes, it gets harder to remain consistent.

You think you’re comparing apples with apples, then you discover an error in your GA set up in Year 2, and you were in fact, comparing apples with pears.

So, @Bas_van_Leeuwen, I think if you need to set expectations around the integrity of the data and how it might become slightly less than 100% accurate after, e.g. two years. You can sweeten the pill by delivering better and better features.


(Bas van Leeuwen) #17

That is an excellent point Steve! That said, I think the end user is probably much better equipped to deal with these weird things that we are, but it’s probably nice to remind them of potential issues.

Reminds me of another point: When we analyse new communities, when we weird spikes happening and ask what happened on date X it is invariably a migration. Those tend to be very bad for data-integrity. Be aware of this :smile: