Accurate audience measurement in Africa through mobile tracking

Peter Searll

The media consumption landscape is transforming at an exponential rate, yet there is little innovation in media research addressing the impact of technological advances.

Harmonising market research across Africa is the key objective of the Pan African Market Research Organisation (PAMRO) and in line with the theme of this year’s conference held in Zimbabwe in August, ‘Africa Media Research in a Globally Connected World’, Dashboard Marketing Intelligence presented a solution for accurate audience measurement in Africa through mobile tracking and the results of a Zambian pilot study.

Mobile is often described as the second screen, however, in Africa it is often the first or only screen. We addressed two questions: how to measure mobile media consumption accurately; and, how this can be used to augment traditional media data. The objective of our nine month case study was to measure and track mobile behaviour with a specific media focus.

Digital research challenges in Africa

There is a global shift from traditional media consumption to online and there will inevitably be a critical mass tipping point where online will become dominant, it is just a matter of time. According to Price Waterhouse Cooper’s (PWC) ‘Entertainment and media outlook 2015-2019’ the compound annual growth rate (CAGR) on spending on internet will increase by 21.7% in the five years between 2014 and 2019 in South Africa and by 31.6% in Nigeria. In South Africa the projected CAGR in this same period of radio spend is at 5.9% and TV is at 6.2%. Television now gets five times the advertising spend as internet, this will be down to 3.5 times by 2019. Media buyers and advertisers are slowly catching up to market realities of media consumption.

Online media consumption in Africa is on the rise, but has not been properly measured. Most audience measurement methodologies still use diaries, which are notoriously problematic. The media industry must keep pace with consumers and so must research methodologies.

The current research is not necessarily sufficient to give advertisers a clear, holistic perspective and digital measurements are often not comparable to traditional media currency. Digital space in Africa is not understood properly and researchers need to make sure that they accurately reflect the changing media consumption landscape for advertisers, media agencies and media owners.

Keeping ahead of the curve with online measurement

We set ourselves the challenge at Dashboard to create a research methodology tailor-made for the African market that observes real, rather than reported mobile behaviour. Our aim was to observe and gather data on actual mobile behaviour and then analyse this ‘big data’ using familiar audience measures such as reach and frequency, so that the data could be integrated with traditional media research.

We wanted customer centric rather than website centric research which tracks an individual’s mobile usage through their device, rather than traffic through a specific website.  The aim was to gather an accurate, complete permission based record of actual behaviour - not diary, recall, or interview based research. We wanted truly longitudinal data where no surveys are required, just continuous passive data collection.

Our data covers all websites and apps (very close to real-time), not just the large ones. It was important that there were no tags from site owners and that we did not rely on cookies. In order to contribute to the bigger framework of media audience research in Africa it was imperative that the data was scalable and could be seamlessly supplemented with survey data or merged with existing media data (like TV and radio data sets), if required.

A Zambian pilot

For our Zambian pilot study, Dashboard built an app that our sample downloaded onto their mobile devices. Once activated, it tracked all their mobile behaviour continuously for nine months. We opted for an android mobile app, as iOS is not widely used in Africa.

Demographics were established from the sample of 60 respondents, recruited from our existing panel in Zambia. Respondents were incentivised with airtime to download the app and give permission for their device to be tracked. Once activated the app collected data all the time and data was uploaded to our server three times a day in efficient packages. Individuals were not identified to protect privacy and POPI principles were adhered to. The data was then collated, cleaned and analysed.

The key metrics collected about device usage included:

• Websites visited - including time visited and number of times.
• Apps - including when used and time in foreground.
• Wi-Fi versus GSM data usage - uploads and downloads.
• SMS - sent and received.
• Calls -made and received.
• Other phone functions like settings, calendar etc.

It was critical to clean and code this very complex data. Analysis requires sophisticated protocols to extract the mass of complex data. During the study there were in total over 242 000 website visits and 402 000 app usage occasions. The results are not market representative due to the sample size, but this is simply a matter of scale. The outputs demonstrate what kind of data is available.

Top-line insights

1. Weekly phone usage patterns showed that distribution during the week is the same for both genders, but spikes up on Fridays and is lowest on the weekends. Women speak for longer than men, the average female call duration is 129 seconds compared to 84 seconds for men.  Men, however, make more calls than women, on average three more per day.
2. Wi-Fi accessibility was measured and the percentages of Wi-Fi and GSM uploads versus downloads were compared. GSM accounted for 54% of data usage and women used Wi-Fi much more for downloading than men.
3. Website reach by category was tracked, 100% of the sample was on social media and used email on their mobile devices. 80% of women and 85% of men used their browser and search engine, 70% of women and 75% of men accessed sport news. Women interacted with activism, tech/device news and adult websites more than men who were on career and sports/sports betting websites more.
4. The reach and frequency of specific news websites was measured in line with traditional media currency. While The Mirror (UK) had the highest reach, the Zambian Watchdog had higher frequency and Lusaka Times also had high frequency. This intrigued us and on further inspection we noticed that the pages visited most on the Mirror were football related, and a trusted source of EPL news.
5. Social media usage was tabulated and 100% of both genders used Facebook, next was LinkedIn used by 50% of women and 56% of men. Twitter was used by 38% of women and 29% of men. Rate n Date and Bb Dating were more popular among women, while Slut finder and Date hot dolls were exclusively male. Waplog had the highest combined reach among dating sites. This type of information is unlikely to be revealed in an interview!
6. Versatile data and analysis enabled detailed profiling by time of day/day of week. Fascinating Facebook insights showed very similar gender usage. Sunday usage exhibits peaks before and after lunch and around dinner time. ‘Dead zones’ in terms of usage were between 2-4am and 4-5pm. 10am was good for weekday ad or content placements, but not on Sunday, when 3pm or 8pm were better.
7. App reach by genre was measured and communication apps were the most widely used (100% reach), this was closely followed by productivity apps (96%), social media apps (93%), internet browsers (87%), media and video (87%), music and audio (84%), photography (84%), while shopping apps had 20% reach. WhatsApp was the platform of choice, followed by Gmail and over 25’s used Messenger and Chrome more, while the youth preferred Opera Mini and Internet for Samsung Galaxy.
8. An App drilldown measured usage frequency and patterns by time-of-day, this compared app and website daily usage patterns and different demographics. Usage and share over time was tracked to reveal trends. In addition, a day in the life was viewed for individual and group daily usage patterns to provide an aggregated mobile perspective - a powerful tool for digital segmentation.

Reach and frequency at a granular level

Our approach provides data versatility and multiple time scales: by hour, day, week, month (or minute if really needed). It puts the full usage of mobile devices in perspective, websites and apps can be grouped by type or genre for full competitive profiling. A detailed analysis of reach and frequency at a very granular level is readily available for any demographic.

This app research can stand alone or be merged with traditional media research data to get the full picture of media consumption. Dashboard is not trying to steal anyone’s lunch, but provide the gravy. This home-grown African solution is the only research of its type in Africa that provides this level of complexity at the price point that we are able to offer.