Image credit: Sarah Alex CAFOD

Why learning from failure is key to anticipating humanitarian crises

Start Network's Sarah Klasson looks inside the 'black box' of the Crisis Anticipation Window, to see how we can learn from the failure to act early before crises hit this summer.


Time to read: 7 minutes

Area of work:

For the nervous flyers out there, it’s comforting to know that flying has become the safest form of travel. It’s safe because we’ve recorded and learned from past failures- all captured in black boxes which are a compulsory part of all commercial aircrafts. Matthew Syed writes about this in his book ‘Black Box Thinking’. Syed reminds us that breakthroughs always begin with multiple failures and that true success lies in capturing, understanding and overcoming these failures. I think he’s onto something.

The Start Network launched the Crisis Anticipation Window in November 2016. For the first time, the 42 humanitarian charities that make up this network can apply for funding- ahead of a potential crisis. Start members can use forecasting information to raise an alert to the network about an escalating risk of a crisis. This alert can trigger Start Funds which support projects to mitigate potential damage. This approach is pretty intuitive and so far we’ve seen a lot of promising proactive aid interventions. We are no longer just waiting and reacting to chaos after it unfolds -humanitarians are on the offensive.

But right now, I don’t want to focus on our successes. I want to look inside the black boxes that emerged over the summer, because there’s so much in them that’s key to breakthrough for our network moving forward. And there’s too much at stake to not learn from our past.


On 13 July, Start members raised concerns about potential flooding in Niger and submitted an anticipation alert to the network. When decision makers were considering the alert, they knew that there was a real risk of flooding in the country and they knew that communities were vulnerable. But they were also (rightly) concerned that the alert only focused on one area in the country that already had a relatively high NGO and UN presence and funding. Although they could have expanded the geographic scope of the alert to focus on the most vulnerable areas, they felt that not enough information was provided about the needs and risks in other areas and so the alert wasn’t activated.

On the 26 of August, torrential rains arrived in Niger- as expected. Within a week, the rain had destroyed hundreds of homes in the capital, Niamey. Niamey lies along the Niger River, so it’s particularly vulnerable to flooding. Thousands of people were forced from their homes. According to Niger’s government, flooding killed at least 44 people in Niamey and other parts of the country. Frustratingly, one of the main causes of flooding was simply the obstruction of drainage canals. Clearing drainage canals is one of the projects that could have been funded if the alert was activated.

What’s in the black box: There’s a lot to learn from this experience. Uncertainty was the ultimate stumbling block here. As a network, if we’re going to be proactive, we need a greater appetite for risk. Anticipating crises always involves a greater level of uncertainty than responding to something in front of you. If there is a clear risk, decision-makers should be empowered to broaden the scope of alerts to the most vulnerable areas - even when faced with uncertainty. The ECHO funded Crisis Anticipation window is a welcomed opportunity for humanitarians to experiment with proactive responses and to learn from these experiences along the way.  

Sierra Leone

On 14 August, a mudslide linked to flooding killed more than 1000 people in Sierra Leone in a town called Regent on the outskirts of Freetown. The government initially pegged the death toll at 450, while rescuers warned that many of the more than 600 people still missing would likely not survive. It was a devastating and traumatic experience for the town’s residents. As funding often takes a while to arrive, the Start Network began responding rapidly to fill the critical gap immediately following the mudslide.

Storms and torrential downpours are common in Sierra Leone in August and September. Back in July, Start Members began talking about the escalating risk of flooding and landslides in the country. We knew that the risks were increasing. Sierra Leone had seen 104cm of rain since 1 July, which was three times more than expected during the rainy season according to the US National Weather Service’s Climate Prediction Center. We then discussed the potential flooding and landslides in Sierra Leone for two weeks. Many people in Regent live in informal settlements on steep hillsides. The country’s officials had warned against widespread unregulated construction on the hillsides. The uprooting of trees for construction on the hillside is also known to have made the soil particularly unstable and more vulnerable to collapse. The Sierra Leone metrological department didn’t issue a warning to hasten evacuations from the danger zones. As we were discussing and drafting the alert note, hundreds of people were killed.

What’s in the black box: It’s important to take a step back and reflect on this experience: How can we learn from this moving forward? When we know a risk is escalating, humanitarians need to act- and we need to act quickly. Natural disasters don’t work according to our timelines. When multiple agencies are concerned about an escalating risk, we need to make sure that this quickly translates into action.  If we’re uncertain about a risk- we can analyse it. The Start Network provides a small grant for inter-agency risk analysis. It’s called the ‘Analysis for Action Grant’ because when people come together to collectively analyse risk, they’re more likely to do something about it.

Hurricane Irma

You’ll have seen in the headlines that Irma was the most intense hurricane in the Atlantic since 2007. The category 5 storm caused widespread and catastrophic damage, particularly in parts of the northeastern Caribbean and the Florida Keys. Irma developed on 30 August near the Cape Verde Islands. Shortly after, Start Network members who were actively monitoring the hurricane raised it to the network. Still facing uncertainly about the possible impacts and final path of the hurricane, members raised an anticipation alert to respond to the approaching hurricane on 6 September. It made landfall on the coast of Barbuda later that day, and continued along its path of destruction until the 11 September.

Discussions around the alert on 6 September were complex and difficult, but ultimately, the alert wasn’t activated. Decision makers felt that it was raised too late to implement activities that could mitigate harm and loss related to the hurricane and that the scale of the crisis was seen to be beyond the scope of the Start Fund. It was also rightly pointed out that other sources of funding would be made available. Agencies involved in the alert likely walked away from the experience discouraged. Anticipating hurricanes is a very tricky business.

What’s in the black box: Although the alert wasn’t activated, this experience was a huge step forward for the network. The Start Fund seeks to enable quick and proactive aid interventions. It’s important to remember that in 2015 and 2016, for a variety of different reasons, our members in the region have only been able to raise alerts for hurricanes no sooner than 9 days after the storm has hit landfall. This time they raised the alert in advance. We’re moving in the right direction. Earlier alert activations- even very close to storm landfall, will allow our members to begin their project proposals sooner. This means we reach impacted people sooner.

I share these black box reflections because I believe they capture the journey that we’re on as a network. Working to shift the humanitarian sector from a reactive to a proactive way of working happens in small, sometimes painful, steps. We need to recognise this journey and we’ve got to keep embracing and learning from our ‘failures’ in order to move forward.

Acting in anticipation of humanitarian crises is new for many agencies. It can be difficult to know when to trigger an alert and what projects can be the most effective. We have created guidance notes to help our members anticipate various hazards. This month we are releasing two new guidance notes: anticipating disease outbreaks and tropical storms

Read more about the Start Fund Crisis Anticipation Window.