When Analytics Go Wrong: Projecting Missing Data
It happens... the best laid plans of mice and men. You made an update to your analytics suite, and suddenly, your data is not coming through at all, or it's not coming through accurately.
It's fixed, but now you have a data gap. So what are the best ways of projecting the missing data?
If it's a short time period for consistent data over a time period, projecting can be done by linear estimation - or simply drawing a line between the two points, adjusting so it fits the period of time that's missing (hours or weeks or even months).
If the data collection is inaccurate but consistently inaccurate, normalize the data to match the pre (and/or) post inaccuracy data. Multiply it up or down by the factor it was off.
If you have some numbers that were consistent and just a few that were off, see if there are consistent conversion rates or ratios between the metrics being tracked. If the conversion rates are consistent, then use those to project back to what the numbers likely were.
Unfortunately, there is not much that can be done to fix historical data when something goes wrong with analytics. But using the techniques above, you can approximate the missing history.