We have improved the values within traction_metrics and added a new field for each traction metric named traction_metrics.external_<traction metric name>
The problem: Historically, when we didn’t observe a new value for a traction metric (say, twitter follower count), we would simply use the most recently observed value. In practice, this would look like periods of flat-lining in a graph, as a value got pulled forward into future weeks.
Our solution: We are fixing this problem on a forward- and backward-looking basis. We have made material improvements to our data pipelines that fetch traction metrics which means going forward there should be far fewer instances of empty observations. When we are missing any weekly value, we will be algorithmically interpolating the time series value based on the values to the left and right of the missing value.
How this affects you
The traction metric data you’re already using will improve, without any action needed from you. In cases where you previously observed flat-lined time series, you will now see sane trend-lines.
A new field will be available for you to access*.* If you want to see the raw observed values for your own modeling or observation, you can access the new fields we’ve added within
traction_metricscalledexternal_{traction_metric_type}(for exampleexternal_twitter_followers).
Example:
Current State of instagram_followers
Date | Instagram Followers |
Jan 1 | 10 (real value) |
Feb 1 | 10 (we know to be bad/stale) |
Feb 15 | 10 (we know to be bad/stale) |
March 1 | 10 (we know to be bad/stale) |
March 15 | 20 (real value) |
Outcome 1: Store all the correct real data in external_instagram_followers .
Final state of external_instagram_followers
Date | Instagram Followers |
Jan 1 | 10 (real value) |
March 15 | 20 (real value) |
Outcome 2: Fill in the gaps of using interpolation. If there is over 30 days gaps between two data points. We’ll add one around 15 days mark.
Final State of instagram_followers
Date | Instagram Followers |
Jan 1 | 10 (real value) |
Jan 15 | 12 (derived value) |
Feb 1 | 14 (derived value) |
Feb 15 | 16 (derived value) |
March 1 | 18 (derived value) |
March 15 | 20 (real value) |