PC Usage Monitoring

For tech products and services ranging from jet engines to Facebook, data on user interaction with the product is the baseline “ground truth” for understanding user behavior. It enables A/B testing, behavioral segmentation, and many other types of user research. However at Intel, given our role as an “ingredient” to compute devices and lack of a direct relationship with end users, we were cut off for years from this kind of real-world input on how people were using our products. This project was Intel’s first attempt at solving this problem.

01 Background and Goals

The introduction of the smartphone in 2007 and then the first viable consumer tablets in 2010 caused a lot of introspection about how those two devices would diminish the role of the PC. A narrative in the industry sprung up saying that the smartphone would be for quick hits of information and communications, while the PC was for usage occaisions requiring immersion like productivity, “serious” gaming, and creativity. Testing this hypothesis spawned a ton of diary studies in both my group as well as Intel Labs, but we all found traditional diaries lacking in capturing the “rhythm” of device usage. I had run across an internal IT project using tracking software to measure employee usage, so I team up with the Intel Labs researchers to re-purpose this platform to understand patterns of PC usage.

02 Methodology

  • Developed software to continously monitor PC usage

  • Recruited a representative panel of >150 users in 3 cities

    • 30 minute survey

    • 2 months of PC usage monitoring

  • In-home interviews with each respondent

    • Review of usage data visualizations

    • Discussed routines as revealed by the data

03 Insights

This was the first data of its type that we had ever produced at Intel, and it changed a lot of perceptions about how our products were used.

  • We totally blew out of the water the notion our internal product teams had that PCs were mostly for long interactions. In fact, we saw the average time of interaction (AMB keyboard and mouse activity) was measured in a few minutes.

  • Likewise in our contextual interviews respondents showed us the physical context of their interactions, we saw the PC being used througout the house as a companion device throughout the day for buying movie tickets, checking scores, and crafting email.

  • We produced a lot of quant analyses about the share of time respondents spent on different activities, and in later studies used this to segment them. But we learned from the interviews that time spent does not always equate to the meaningfulness that users place on different activities. So even if they only spent a couple hours each month editing pictures and videos, media editing and storage may have weighed quite heavily in their experience of the machine (not to mention the purchase decision.)

04 Actionability

  • The response to this research was profound and immediate amongst our senior engineering management. Based on our new point of view around the intermittent quality of user interaction with PCs, they approved a project which had been on hold that eventually cut in half the boot up and wake times of all Intel-based PCs.

  • The findings on “meaning” vs actual usage also informed subsequent work in Marketing on purchase triggers and user segmentation.

  • After this project, Intel made a significant investment in PC usage telemetry software development, which had its biggest impact in engineering. My research team benefited greatly from this investment, and we used improved versions of the collector software for anything from an iPad placement study to understand the effect on household PC usage to building predictive big data models of PC replacement based on device usage data.

05 My Learnings

  • Power of Cross-Group Collaboration - This project would not have happened had my research team not reached out to our collegues in Intel Labs, IT, and our Quality team to build this new capability.

  • Risk Taking - This was an incredibly expensive capability to get off the ground, both in programming resources and cost of recruiting the panel. However the prospect of solving our usage data access problem made the investment worthwhile.

  • Governance - This was the first time we’d had access to such detailed personal data, and required the team to not just work closely with Legal but also build processes on the backend around data retention, data minimization etc.