3 ms·

In fact, from my reading the rate of signups was increasing before hand and continued increasing afterwards, implying the design made no difference. That's what

In fact, from my reading the rate of signups was increasing before hand and continued increasing afterwards, implying the design made no difference. That's what I can see from the trendline anyway, and the rest of the graph is too uneven to draw any conclusions.

Fogest 12y ago
Yes, exactly. The graph is full of spikes and I do not see any indication that there was any real change present from the CSS change.
eCa 12y ago
They are wrong to draw the trendline the way they did. Since what they are claiming is not steady growth (then the trendline would make sense) but a shift in behaviour they should instead have drawn one line for average before the change, and one line for average after the change. It is quite obvious that something has changed. Before the change there are five days above 7.5%, and after there are more than 10 days above 7.5%. The valleys are much higher too.
- cbsmith 12y ago
  > It is quite obvious that something has changed. Yup, but that's not an 8% increase in AR's. That's maybe an 8% increase in the rate of increase in AR's. That could be attributable to a LOT of things though (there isn't even enough context to control for network & seasonal effects, let alone whatever increased attention there might have been to sign ups after the rollout). Ironically, probably the best way to isolate the impact to the launch is to narrow the time window and resolution to the minutes before and after the redesign launched.
- richardwhiuk 12y ago
  More days above 7.5% would be true even if the underlying trendline was going up anyway. They really should plot the moving average and include a longer timeseries - plotting the trendline before and after would be trying to prove what they are trying to show, rather than trying to work out whether what they are trying to show (our fancy new design wasn't a complete waste of time) is actually true.

Fogest 12y ago

Yes, exactly. The graph is full of spikes and I do not see any indication that there was any real change present from the CSS change.

eCa 12y ago

They are wrong to draw the trendline the way they did. Since what they are claiming is not steady growth (then the trendline would make sense) but a shift in behaviour they should instead have drawn one line for average before the change, and one line for average after the change. It is quite obvious that something has changed. Before the change there are five days above 7.5%, and after there are more than 10 days above 7.5%. The valleys are much higher too.

cbsmith 12y ago
> It is quite obvious that something has changed. Yup, but that's not an 8% increase in AR's. That's maybe an 8% increase in the rate of increase in AR's. That could be attributable to a LOT of things though (there isn't even enough context to control for network & seasonal effects, let alone whatever increased attention there might have been to sign ups after the rollout). Ironically, probably the best way to isolate the impact to the launch is to narrow the time window and resolution to the minutes before and after the redesign launched.
richardwhiuk 12y ago
More days above 7.5% would be true even if the underlying trendline was going up anyway. They really should plot the moving average and include a longer timeseries - plotting the trendline before and after would be trying to prove what they are trying to show, rather than trying to work out whether what they are trying to show (our fancy new design wasn't a complete waste of time) is actually true.