Using data analytics can benefit the social services sector

HALF the money I spend on advertising is wasted; the trouble is, I don't know which half, goes the famous marketing adage, coined more than a century ago by John Wanamaker, an American department store tycoon. Today, businesses are using big data analytics to get better bang-for-buck from their marketing budgets. In a future where our tax dollars will be increasingly stretched, it is time to use the same data tools in the social sector.

Have you ever searched for a product, and then seen ads for that same product appear on your Facebook feed shortly thereafter? This happens because Facebook is using data analytics to help advertisers target consumers most likely to be receptive to their product. It is a win-win for both advertisers and consumers. Advertisers reduce waste from marketing to people who would never buy their product.

At the same time, consumers see fewer irrelevant ads. To enable this targeting, Facebook creates a comprehensive data profile of each of its members, and uses this profile to predict how likely each person is to be interested in different products. Facebook then shows each person only the products they are likely to be interested in.

A similar targeting process can improve efficiency in the social sector. Take the issue of child protection. When a child is severely abused by his or her family, he or she is taken away and placed in foster care. Case workers often receive tips about suspected child abuse. If they intervene early enough, they can help prevent the abuse from worsening and avoid the need for foster care.

However, case workers simply do not have the resources to investigate all tip-offs to the same extent, so they need a way to prioritise their time. Data analytics can help. In the US, New Jersey built a comprehensive data profile of each at-risk child and used the profile to predict the risk of each child needing foster care. The predictions were highly accurate: from past data, children deemed to be at the highest risk were over 100 times more likely than those at lowest risk to require foster care within a year. Targeting allowed case workers to spend more time with children at the highest risk, thereby maximising their impact.

A second tool gaining popularity in business is A/B testing. Before a tech company rolls out a new marketing message today, it is likely to conduct a mini experiment that resembles a clinical trial. First, the company randomly splits its customers into two groups, group A and group B. The company then sends group A the new message, and group B the existing message. By comparing the responses between the two groups, the tech company can tell whether the new marketing message works better than the existing one. Most of the time, the result is that what seems like a promising idea in fact does not work. Google and Microsoft have run about 13,000 A/B tests on promising new messages and products, and only 10 to 20 per cent show any difference versus the status quo.

In the social sector, a nascent A/B testing movement is helping to pinpoint which programmes really work. The results are sobering. Of the skills-training programmes evaluated via A/B testing by the US Department of Labor, only about 25 per cent have shown strong effects. In education, the numbers are worse: only 10 per cent of interventions evaluated by the US Department of Education have shown strong effects. Put another way: for every US$1 spent, about 75 cents to 90 cents is probably not achieving the results we want. By helping to rigorously determine what is working and what is not, A/B testing can help maximise the impact of taxpayer and philanthropic dollars.

Applying data analytics in the social sector must be done with care. Getting it right often concerns lives, not just purchasing decisions. There are several key questions to consider. First, privacy. What information should be allowed to be included in someone's "risk profile"?

Second, ethics. Can rigorous evaluations be conducted without depriving people of services?

Third, cost. What is the right balance between spending on data analysis versus spending on direct service delivery?

In the private sector, similar questions are slowly being resolved by a give-and-take approach between companies and consumers. In the social sector, it will be up to beneficiaries, philanthropists and taxpayers to work out the appropriate trade-offs.

In the coming era of fiscal tightening, we can do more for those in need, so long as we do it smart.