“Make memory unnecessary: put the required information in the world.”
Don Norman. The Design of Everyday Things
In 2012, on the invitation of Stuart Rutherford, a wise mentor I was fortunate to meet early in my career, I visited the rural SafeSave replication in Hrishipara, Bangladesh, and conducted a structured study of 20 illiterate savers and borrowers. As a consultant I had observed that many microfinance clients did not expect to be able to decode their passbooks – though these were often their only record of their personal financial transactions. I wanted to test this observation further, and – if it held up – try to glean insights into possible causes.
The scale of the problem surprised me. Only four of the twenty respondents in the Hrishipara sample could even find a specified entry in their own passbooks. Only two could find the value of their own withdrawals on the slip designed for the purpose (which contained several other numbers as well). They appeared to face three navigation problems in their records – decisive when taken together.
- Column labels are in text they can’t read. While important, this is not decisive by itself.
- Tabular syntax involves a left-right/bottom-top aggregation protocol for sorting the tabular cells on any page, which could number up to a hundred.
- Place value can be disabling by itself. Most can read numerals ‘phone number’ style (for example, if they see ‘4,706’ they may say “four, seven, zero, six”). But only four in the Hrishipara sample could read these numbers ‘cash’ style (by saying “four thousand, seven hundred and six”). This inability blocks learning by trial and error: decoding the number and matching it to the expected transaction amount.
In 2014-15, with colleagues Hawa Mnyasenga at the University of Dodoma, Sou Socheata at Oxfam Cambodia, and Margaret Hazlewood at My Oral Village, I continued this research with structured tests of 115 illiterate individuals in Tanzania, Cambodia and Timor-Leste. These tests offered many insights into oral culture, but there are three to highlight here.
- A small majority could write a single-digit number. This skill was least developed in Timor-Leste and most developed in Cambodia. For our purposes the most important consideration here is that a person who can write a single-digit number can also read and recognize it, wherever they see it.
- A larger majority (76% or three-quarters) could count the cash presented to them, and accurately announce the count when completed. Although the cash sums presented were not large (worth approximately $25 US in Cambodia and $60 US in Tanzania) they are coded in 6-digit numeral strings in the currencies of these nations.
- The most difficult task was reading the multi-digit numeral string, and only 23% (or less than a quarter) could successfully do this. Respondents made many order-of-magnitude errors, in which they selected either ’17,500’ or ‘1,750,000’. These errors were especially common among partly-schooled respondents; unschooled ones often refused to answer the question. (We did not collect error data in Timor-Leste, so this observation applies to Cambodia and Tanzania only.) On the other hand, individuals from Timor-Leste were more likely to respond correctly in spite of weaker numeracy skills. This appears to be related to the use of the US dollar, which involved a 3-digit number due to its high value.
A Cash Keyboard (Input Method Editor)
A literate person may well ask why it would be more difficult to read a 4- or 5-digit number than it is to correctly count a similar sum in cash? But illiterate people live in oral cultures, where pre-literate but quite sophisticated methods of storing and retrieving information without text are commonplace, and where cash – an innovation enabled by the printing press – is still far less trusted as a store of value than as a medium of exchange.
In an oral culture, it is essential to learn how to accurately count cash, but it is not essential to learn how to accurately read a long numeral string, because oral adults rarely encounter long numeral strings, and when they do encounter them, can often find work-arounds, like asking someone for help. The opportunity to safely and flexibly accumulate large amounts of cash in an account are also quite rare, and there appear to be few other incentives to acquire the numeracy skills required to process large numbers.
How can we help this population to adapt to formal finance – and in particular to mobile money?
The solution is to build from the strengths of oral culture. Given that oral individuals can read single-digit numbers and count cash accurately, but can’t safely decode place value in large numbers, a smartphone solution might look something like the one at right, which I presented to a consultation on mobile money at CGAP in Washington DC in April, 2016.
Area 1 contains the denominations of the user’s national currency. These must be represented to accurately reflect the key elements that users rely on to count, including colour, relative size and commonly recalled motifs and images.
Area 2 includes two parts – the traditional input field in Indo-Arabic numerals, and an icon that triggers a voice receipt.
Note that a parallel cash receipt can be included either as part of Area 1 (if screen real estate is limited) or above it. For example, if a user wants to input “30,000”, s/he can tap on a “10,000” note three times and a strong red ‘3’ appears above the note. It may be preferable to include a separate area for these notes, to maximize oral usability.
Field observations indicate that oral adults rarely deal with large Indo-Arabic numerals except in the context of cash. One-off purchases usually don’t require this knowledge, but financial transactions do. For smaller numbers and small-value cash transactions, oral adults often memorize details (“I’m selling piles of tomatoes worth 500 shillings each, and if someone gives me a 5000 note and buys four piles I will have to pay them 3000 shillings back.”) This system is cumbersome, and good design seeks to alleviate this sort of burden. As Don Norman, the celebrated author of The Design of Everyday Things, has observed: “Make memory unnecessary: put the required information in the world.”
It is my belief that a cash keyboard could change digital finance, because in the year since the CGAP presentation, two additional pieces of the puzzle have started to come into place. The first is that this approach is appreciated by illiterate users in field testing. They appear to face more problems with actual numeracy and literacy than with digital literacy per se. The second, evidenced in a study with MicroSave in India (a blog on this will appear shortly) and subsequent work in Tanzania, is that many literate people also struggle to decode place value. And because those who are not fully capable sometimes make errors, often by an order of magnitude, they lack confidence in their abilities, especially in real-time financial transactions, which often involve high personal stakes and time pressure.
While the evidence on the strategic significance of place value cognition for financial inclusion must still be formalized, it is now strong enough that I would encourage any reader of this blog to test this out with their clients – particularly if you are working in a country with a low-value currency. Give them some cash, ask them to count it, and show a list of similar numbers. As literates, we tend to believe that at least they will avoid order-of-magnitude errors. In fact, without knowledge of place value, this is one of the most common forms of errors. If you find yourself replicating my results, discuss with your teams what you might do about it. Can you change the graphics on your paper and digital interfaces? If so, how?
Feel free to contact me (firstname.lastname@example.org) if you would like more information.