A Family Affair: The Reality of Household Device Sharing

By Marisa Lucchesi October 18, 2020

A Family Affair: The Reality of Household Device Sharing

Measurement of advertising outcomes relies upon data that tells us who saw an ad. But how can we know for sure who saw an ad if they’re using the same devices as their spouse, parent, or child? Think of the mom passing her iPad to her daughter at the dinner table; the teen doing homework on the family computer. As an industry, we have never been quite clear on who has actually been exposed to an ad. Yet despite understanding that the prevalence of household device sharing leads to flawed data, we have yet to invalidate measurement based upon this data.
At first glance this approach makes sense — if your device was exposed to an ad, you should be part of the deterministic exposure group used for measurement. But dig deeper and you’ll find two things that beckon for change: One, this model is under fire right now due to the changing privacy landscape — Chrome’s erosion of 3rd-party cookie support; Apple’s changes to IDFA. And two, it’s becoming clear that perhaps we never had the deterministic exposure data we thought in the first place – or anything even close to it.
A new study by Upwave reveals much larger rates of device sharing within households than previously suspected. The actual rates of household device sharing cast new doubts about conventional measurement, suggesting the need for more sophisticated methodologies that don’t make naive assumptions about device ownership and usage. Here, we’re unveiling our findings into just how widespread device-sharing is, what the implications are to the ad measurement industry, and the parameters you need for more accurate data.

Methodology and Findings

Upwave conducted a census-representative research study with 1,227 individuals on our Digital Network in an effort to better understand the prevalence of device sharing in the United States.
Our research also examined the differences in sharing across device types, different household and family structures, and across different demographic and regional compositions. The following summarizes our key findings from this research:

  • Device-sharing is common practice among those who live in multi-person households, which comprised 78% of the people surveyed.
  • Generally, the more people in a household, the higher the rate of device-sharing.
  • Device-sharing tends to be higher among multi-person family households, in particular those with children, parents, and among minorities.
    • Sharers are overrepresented in parents and their children under 18.
    • African American and Latinx households are more likely to share smartphones than White households. Additionally, multi-member family households share devices more than other household types (e.g. non-family).

Sharing by device type

Our study revealed that if households share one device, it’s likely they share other devices as well. Smartphone and tablet sharers tend to have a higher likelihood for other types of device-sharing like computers or CTV.

Rates of sharing

Sharing rates vary significantly depending on the device type itself, but are highest among CTV and computers.

  • Connected TVs (e.g., smart TV, Roku, Chromecast) are shared among a majority of multi-person households (47% of all multi-person households, on average). 68% of connected TV sharers share their device at least once a day.
  • Computer-sharing is also high: Almost 35% of all multi-person households share computers or laptops.
  • While smartphones are the least likely to be shared, still 10% of multi-person households share these devices. And, 58% of smartphone sharers share their device at least once a day.

What Does This All Mean for Measurement?

It’s clear device-sharing is high in multi-person households, and that the rate of sharing itself varies by household size and device type. This ubiquitous device-sharing makes accurately measuring brand lift or purchase attribution on the individual level nearly impossible, and introduces an array of biases into measurement:
In general, when exposed persons were not exposed, and non-exposed persons were exposed, the contamination of the data results in lower lift estimates. Can we just assume this bias affects everyone equally, and look at relative performance? No, we can’t, because this contamination varies systematically. In particular:
Media partners and campaigns disproportionately targeting computers and CTV are treated unfairly due to the depressed lift estimates that result from contaminated control groups.
Media partners and campaigns targeting families are also unfairly penalized by their datasets, with lift results that underestimate outcomes due to contaminated control groups.

What can be done?

The demands that household device-sharing place upon measurement data, it turns out, also solve for privacy. The data needed for measurement to account for device sharing must clearly meet the following parameters:
The probability of exposure is accounted for in exposure data.
This probability of exposure accounts for household size and environment/device type, as exposure probability varies.
Once measurement data is framed in terms of exposure probability, exposures can be tracked at the household or mini-cohort level, rather than the individual or device level. This unlocks innovative solutions to privacy-safe measurement and leads to actual improvements upon deterministic data that never was truly deterministic before.
Fortunately for brands and their media partners, the future of ad measurement is moving in this direction. Digital advertising leaders like Upwave and major measurement and social media platforms are innovating in the area of household and micro-cohort level tracking to better account for the prevalence of household device sharing.