People buy financial products to get something done: to save for retirement, to put aside a down payment, to leave room for the unexpected. The same product can be exactly right for one job and entirely wrong for another. The job defines the yardstick for the result: how long the horizon is, how large a loss can be borne, and what counts as done. Financial jobs are distinctive in that luck has a large effect on results, so the standard of completion can only rest on the process. A complete delivery has eight stages, from stating the goal clearly to letting the user leave at any time.
The marketing scholar Levitt (1969) often quoted a saying: people don't want to buy a quarter-inch drill; they want a quarter-inch hole. Christensen et al. (2016) developed this observation into the theory of "jobs to be done": in a particular situation a user wants to make some progress, and so "hires" a product; whether the product is good depends on whether it gets the job done. Ulwick (2005) developed it into an outcome-driven approach to innovation, which judges how well a job is done against a set of measurable desired outcomes. Financial products are no different. A stock index fund is a sound choice for retirement savings that will not be needed for thirty years, and a wrong choice for tuition due in three months. The product has not changed; the job has.
One product, different jobs
The product is fixed: a stock index fund. Switch the job the user needs done to see how the yardstick for judging it changes.
People's income is concentrated in a few decades of work, while their spending runs through their whole lives. Modigliani and Brumberg's (1954) life-cycle hypothesis and Friedman's (1957) permanent-income hypothesis both hold that people want to spread consumption smoothly over their lifetime resources: saving while working and drawing down in retirement. This gives rise to the most basic financial job: moving today's money into the future and letting it grow while it waits. There are two further kinds of job: coping with unexpected losses such as illness and unemployment, and making everyday payments. Merton (1995), taking a functional view of the financial system, held that its core functions include transferring resources across time, managing risk, and clearing and settling payments. Campbell's (2006) research on household finance found that many households do these jobs poorly: they stay out of the stock market, diversify too little, or fail to refinance their mortgages when interest rates fall. A survey by Lusardi, Schneider and Tufano (2011) found that nearly half of American respondents said they would have difficulty coming up with $2,000 within 30 days to meet an unexpected need.
"Preparing for retirement" is a goal; it cannot be carried out directly. It has to be broken down into a plan: how much to save each month and in what proportions to allocate it. The plan breaks down into decisions: which asset classes to buy and when to rebalance. The decisions are finally carried out as actions: transfers and orders. To make these decisions, the user also has to look up data and understand what is going on. Locke and Latham's (2002) research on goal setting shows that specific, clear goals guide action better than vague ones. The higher the level, the less often a job occurs and the more it is worth; the lower the level, the more often it occurs and the less each instance is worth. Most of what users do in a product happens at the bottom level, and whether those actions have value depends on whether they serve the goals above them.
User jobs in financial products fall into eight types: looking up a figure, understanding something, researching a company or a view, screening and comparing, making a plan and placing an order, ongoing monitoring, managing a portfolio against goals, and reviewing past decisions. They differ greatly in two respects. The first is frequency: looking up a figure may happen every day, while a goal-based portfolio may need adjusting only a few times a year. The second is the consequence of getting it wrong: a wrongly looked-up figure is usually easy to spot and costs little; a wrong order or a misallocated portfolio can cause losses that are hard to recover. Jobs that are frequent and low in consequence call for convenience; jobs that are infrequent and high in consequence call for confirmation, evidence and the ability to undo. There is one further difference: the first five types are completed in a single interaction, while the last three span long periods and require ongoing delivery.
Eight typical jobs: frequency and the cost of getting it wrong
The horizontal axis is how often a job occurs; the vertical axis is the consequence of getting it wrong. Filled dots are jobs completed in one go; hollow dots are ongoing jobs. Tap one to see its input, output and standard of completion.
Tap a job.
For most jobs, completion is easy to judge: the figure was found, the order was filled. Investment jobs are different: the result is revealed only much later, and luck has a large effect on it. Howard (1988), writing on decision analysis, stressed the need to separate the quality of a decision from the quality of its outcome: a good decision can meet a bad outcome, and a bad decision can, by luck, produce a good one. In the short run, bad decisions with good outcomes are quite common; only over long enough periods and enough repetitions does the quality of decisions show up in results. The standard of completion for financial jobs therefore cannot rest on returns alone; it has to look at the process: whether what should have been done was done in line with the agreed goals and constraints, whether the reasoning was clear, and whether risk stayed within the range stated in advance.
When a user hands a job to another person or a system, an information asymmetry arises between principal and agent: the agent knows what it did, and the principal often does not. Jensen and Meckling (1976) called the resulting problem agency costs; Holmström (1979) showed that the more the principal can observe, the lower the cost of constraining the agent. A complete delivery is therefore not only "getting the job done"; it also means letting the principal know what was done and whether it was done right, and keeping things under control when something goes wrong. In order, it falls into eight stages: accepting the job and stating the standard of completion, having the means to do the job, organizing the process, ensuring the result is correct, leaving verifiable evidence, being continuous and proactive when needed, keeping status visible and asking for instructions when necessary, and letting the user take control and leave. If any one of these stages is missing, the delivery is incomplete.
Monitoring a set of assets or managing a portfolio against goals are jobs with no end point; they consist of cycle after cycle, with a check completed every day, every week, or whenever some event occurs. Ongoing jobs have an easily overlooked problem: in most cycles nothing needs to be done. If a product notifies the user only when it finds a change, then "no news" can mean four entirely different things: everything is fine; the check could not be completed, for example because data were missing; something changed but was not detected; or the product is waiting for the user's confirmation. To the user these four look exactly the same, but their consequences are very different. Every cycle of an ongoing job should be accounted for, in a way that distinguishes these four states.
The job is the yardstick for the product; the standard of completion rests on the process; delivery must be complete; and ongoing jobs must account for every cycle.
First ask when the money is needed, how much, and what loss can be borne; then recommend products. The same product can be the wrong choice for a different job.
On screens for viewing and ordering, show how this action relates to the user's goal and plan.
State the goal, constraints and checks in advance, and judge afterwards by whether they were followed, not by returns alone.
Have all eight stages in place; in ongoing jobs, report the state of every cycle, distinguishing normal, blocked and awaiting confirmation.