Skip to content

● Evidence · Kenya long rains 2026

A Kenyan lender warned its clients, and more of them prepared

With Atram, a microfinance lender in Kenya sent its clients early warnings by text message before the rain. After the rain event of 22 April to 3 May 2026, 74.2% of the surveyed clients who got a warning had prepared, against 53.5% of a comparison group.

This is one pilot season with one lender. The method and every table are public.

● Clients who prepared before the rain of 22 April to 3 May 2026

74.2%of the surveyed clients who received a warning

53.5%of the comparison group

The gap is 20.7 points, and the test finds it unlikely to be chance.

Atram, phone survey of 772 clients of one lender, of whom 376 in the comparison group, Kenya, 22 April to 3 May 2026. Read the method. See every table of the season.

Explore a pilot

A pilot runs for one season, with your own clients.

Small businesses that prepared reported smaller losses

Among the small businesses surveyed after the rain event of 22 April to 3 May 2026, those that prepared reported a loss of USD 80, against USD 115 for those that did not (p=0.04). Read the method.

How large the test was

● One lender.

A microfinance lender in Kenya enrolled about 10,000 of its clients for the warnings of the 2026 long rains. Atram names a lender only with its written consent.

● One season.

The 2026 long rains had ten rain events between February and June. Warnings went out for six of them, and phone surveys followed three.

● One test.

After the rain event of 22 April to 3 May 2026, 772 clients completed the survey, of whom 376 were in the comparison group.

A second season or a second lender can change the number.

How the survey worked

After the rain event, automated voice calls asked a sample of clients what they did and what they lost. The comparison group has two parts: clients that the lender's files set aside in advance, and warned clients who did not recall the warning. A statistical test compares the two groups, and the method note describes it.

The method is a first version, shared for feedback.

What this season does not show

  • It does not show that warned clients repaid better. The effect on arrears and defaults is not measured.
  • It does not show that warnings cut losses. The loss line sets prepared clients against unprepared ones.
  • The comparison group was not picked at random, so the gap may lean in either direction.
  • Losses are what clients reported, in bands of Kenyan shillings.

The season also has an estimate of the losses that the warnings avoided. We do not publish it, because it is a model built on self-reported answers. It also counts the clients' losses, not the lender's. The method note explains it.

What your committee can check

Run one season with your own clients

We run the same early warnings and the same surveys with your clients.

Explore a pilot

Not ready? Read how we measured the season.