Since the beginning of July the improved device detection for Mapp Intelligence is available. In the past few days, we have received numerous inquiries in this regard which we would like to answer here for all customers.
Why has Mapp rebuilt the device detection?
Some customers have reported variations in the device detection in the past. Specifically: Many devices were not recognized at all, others were assigned to the wrong device classes. The distribution of the traffic between the groups "Smartphone", "Tablet", "PC/Laptop" and "TV" was too inaccurate and thus caused false assumptions in the interpretation of the data.
An example: The share of smartphones was often reported as too high, since all unknown devices with Android were placed in this category (including media hubs and smartwatches).
What changes were made to the device detection?
We have summarized all changes in detail in our Release Notes. The essentials in brief:
- More devices are detected more accurately – the share of traffic, which is attributable to so-called fallback names (e.g. "any Android") has become significantly smaller.
- In addition to "Smartphone", "Tablet", "PC/Laptop" and "TV", other device classes are now available: "Console", "E-Reader", "Mediahub", "Smartwatch" and "Other".
- New dimensions are available for the separate analysis of model family and model name.
Why does the share of my mobile traffic change?
There are two reasons for this:
- Because the device detection was not always accurate in the past, some devices were assigned to the wrong device classes. This applies in particular to mobile traffic respectively to the "Smartphone" device class. The new solution can better distinguish between smartphone, tablet, smartwatch, e-reader, console, TV and media hub, resulting in shifts in traffic distribution.
- The so-called fallback devices, which contain unrecognized devices (e.g. "any Android"), have been moved to their corresponding device class. Example: In the past, the fallback device "any Android" belonged to the "Smartphone" class – now it’s being classified as "Unknown".
Why has my data changed retroactively?
The improved device detection has been running in the background since March 29, 2018, parallel to the old device detection. Since July 2018, the improved device detection has also been used for analyses in Mapp Intelligence.
The effects of the improved accuracy in hardware identification can be observed very well in the analyses in calendar week 13 – from this point on the new device classes are filled with data and you will see the already mentioned shifts in traffic distribution. As a rule, this ensures retroactive deviations in the single-digit percentage range.
A clearer change in the data can be observed in the "Unknown" device class. Before Mapp's internal introduction of the improved device detection in calendar week 13, the share of "Unknown" devices is comparatively high. At the same time, mobile traffic is lower when compared with old analyses that may have been exported or transferred to Mapp-external reports. The reason for this is the shift of the fallback devices to other device classes, e.g. the above mentioned "any Android" device, that was previously part of the "Smartphone" device class, and is now "Unknown". This shift also has a retroactive effect. More examples are available in the Release Notes.
From the time when the improved device detection was in operation, a major decrease in the "Unknown" share can be observed, since it now recognizes significantly more devices exactly.
Can I still use the old device detection?
No, the change of the device detection is permanent, a downgrade to the old system is not possible.
We are aware that the changes in the data, especially retroactive changes, can cause confusion and discussion within companies. Nevertheless, we hope that the advantages, namely significantly improved accuracy, outweigh this circumstance.
Can Mapp announce such changes in advance in the future?
Although we have announced all changes 4 weeks before the release via Newsletter, we must admit a misconduct: We have failed to communicate clearly the consequences of the changes, namely the changes to the existing data. We apologize for any inconvenience this may cause and make a solemn resolve to get this right the next time.
If you have any further questions, please contact firstname.lastname@example.org or your personal contact.