Problem shaping and reframing
Problem shaping refines the statement until a problem-solving process can be applied to it. Reframing restates the same problem in another context, that of another actor, another objective or another time scale, which brings out solutions the original wording kept out of sight. Both operations work on the statement itself, where root cause analysis takes the statement as given and looks for what produces it. The deliverable is a written statement, with the stakeholders it concerns, the approach envisaged to answer it and the comparison with the wording it replaces.
Goal
Shaping and reframing work on the problem statement itself, at the moment when it can still change. An analysis team usually receives a problem already worded by a management team, a steering committee or the minutes of a meeting. It takes it as an input. The process it then applies is sometimes faultless without anyone having checked that it bears on the right problem. The IIBA guide names that risk: a predisposition to examine the available data without paying enough attention to the problem.
Donald Schön set out the distinction in 1983 in The Reflective Practitioner: to set a problem is to name what we will attend to and to frame the context in which we will attend to it; solving one takes place inside a frame already fixed. He held setting to be the harder of the two and the one most often skipped. In a survey of 106 senior executives at 91 companies in 17 countries, Thomas Wedell-Wedellsborg reports that 85% consider their organisation poor at diagnosing problems.
The deliverable is a written and adopted statement, one page long. Its value is measured by the answers that it opens and that the other frame closed off, including those that call for no data analysis at all.
Usage
When to use it
- Clean analysis, decision at a standstill: the process is rigorous, the results are sound and nothing gets decided.
- Statement written by a single actor: the finance department worded the problem and nobody else was consulted.
- Solution named in the statement: "we need a predictive churn model" describes an answer and leaves the question unspoken.
- Causal chain that ends out of reach: the five whys run into a factor the organisation cannot correct.
- Diverging definitions among stakeholders: questioned separately, two people describe two different problems.
- The required data does not exist: the problem as stated calls for data nobody collects.
- Before committing to a long analytical study: shaping costs hours, the study costs months.
When not to use it
- Stable statement, unknown cause: move to root cause analysis.
- Incident in progress: run the incident procedure and keep the reframing for the lessons learned.
Description
Two operations on one statement
Shaping narrows. It starts from a statement too broad or too vague to be handled and refines it until a problem-solving process can be applied to it. The test is mechanical: does the statement name the observed effect, its size, the population affected and the period? Can you say which piece of data or which observation would settle the question? As long as the answer is no, the statement needs more work. Shaping leaves the context in place and gains precision inside it.
Reframing shifts. It restates the same problem in another context and changes the set of admissible solutions. The IIBA guide takes the example of an insurer setting a premium: seen from the insurer, the problem is a cost calculation; seen from the policyholder, with their lifestyle, their demographic situation and their medical history, it is a different problem, one that calls for different data and different answers.
The two operations depend on each other. A reframing produces a new statement, in turn too vague, which then needs shaping. Shaping carried through to the end sometimes shows that the precision gained bears on a secondary question, which calls for a reframing. The work stops when an explicit condition is met: a statement a problem-solving process applies to, carried by someone who signs it.
The IIBA guide attaches three limitations to the technique: reframing conflicts with the stakeholders' perception of the problem; it depends on the creativity and the knowledge of a single person; a new frame loses what the analyses already carried out had established.
The boundary with root cause analysis
Root cause analysis takes the statement as given and looks for what produces it. Its two usual techniques, the five whys and the fishbone diagram, are reductive in the sense the IIBA guide gives the word: they descend from an effect towards its causes by elaborating the context, without challenging it.
Shaping and reframing work one level up, on the sentence root cause analysis takes as its input. The two families combine in that order: settle the frame, then look for the cause inside it. A five whys chain that ends on a factor beyond the organisation's reach is the usual signal that the statement should have come first.
"The costs of the Vaud portfolio exceed the premiums collected by 6.4%."
Root cause analysis
the statement does not move
rise in hospitalisations
an ageing portfolio
renegotiated hospital tariffs
a cause to correct
Shaping
same context, more precise statement
…of which 4.1 points come from adults on the standard model, over the past twelve months.
a statement a method can be applied to
Reframing
different context, different statement
Some Vaud policyholders use hospital emergency departments for situations a primary care practice handles.
a non-analytical answer becomes admissible
Creative thinking sits elsewhere: it multiplies the answers to a settled question, whereas reframing changes the question. A third term invites confusion: the framing effect described by Tversky and Kahneman in 1981. It is a decision bias, in which the same option presented as a gain or as a loss changes risk taking. It bears on how a choice is presented, where the technique described here bears on how a problem is stated.
Challenging the context
The context is the set of circumstances that influence the problem, that the problem influences or that help to understand it: industry, perceptions, assumptions, established practices, processes, objectives and regulation. Shaping analyses it, adding symptoms, root causes, constraints and risks. Reframing challenges it. The IIBA guide draws the consequence: assumptions, perceptions and business objectives may have to be restated.
A context is challenged by a small number of named moves, each of which produces a candidate statement. Thomas Wedell-Wedellsborg gathered seven reframing practices in the Harvard Business Review, three of which bear directly on the statement. Changing the objective means asking why the intended result is the one intended, then treating the answer as the statement. Looking for positive exceptions asks who, in the same organisation or the same industry, does not have this problem and by what arrangement. Asking what is missing from the statement brings up the factor everyone knew about and nobody had written down.
Practice adds three more. Changing the actor whose point of view defines the problem is the most productive: the insurer's problem differs from the policyholder's. Changing the unit of measure of the result shifts the levers: a problem stated in francs and the same one stated in turnaround time do not call for the same answers. Changing the horizon separates what can be fixed this quarter from what can be fixed over three years.
The statement template
The IIBA guide gives the deliverable a four-part shape, as valid for the original statement as for the reframed one. The problem statement describes the problem or the opportunity in a few sentences. The impacted stakeholders are those the problem affects or influences. Naming them assigns each of them their share of the problem. The high-level description of the solution or the key needs gives, for a data analytics problem, the approach retained, the data required and the type of problem: descriptive (what happened), predictive (what will happen) or prescriptive (what to do). The comparison with the alternative framing says what the other frame allowed, what it closed off and what tipped the decision.
The fourth part is the one people skip and the one that makes the document worth having: without it, the reader receives one more statement without knowing that it replaces another or why.
Running the work
Obtain the right to reopen the statement. The problem belongs to someone, and questioning it without a mandate reads as a challenge to authority. Wedell-Wedellsborg puts that legitimacy first among his practices. The wording that gets through frames the work as a check of an hour or two, whose possible outcome is that the original statement stands.
Collect each stakeholder's definition in writing and separately. This is the step with the best return for its cost; Wedell-Wedellsborg makes it his third practice. A definition written before the session is fixed before the first speaker can pull the others towards their own view, and the divergences between two texts are the raw material of reframing. One to three sentences per person is enough, with the same question put to all of them: what is the problem?
Name the context each definition implies. Every text assumes an actor, an objective, a unit of measure and a horizon that it leaves unstated. Writing them down beside the text is the condition for challenging them.
Produce the candidate statements as a group. The dependence on one person's creativity is corrected by collective work, a brainstorming session or a stakeholder workshop, which the IIBA guide recommends. Go through the six moves one at a time, writing down each candidate without judging it. Five to eight people produce about ten in an hour.
Write up the two or three candidates retained, following the template, including the original statement, which remains a candidate like the others and sometimes wins. This is where most of the attractive reframings die: writing out the data required and the impacted stakeholders reveals that a frame demands data that does not exist or the agreement of an actor out of reach.
Decide with the owners of the initial statement. Without their buy-in the work produces nothing: this is where the conflict with the stakeholders' perception is settled.
Replace the statement where it lives: the requirements specification, the initiative record, the ticket, the project charter. A reframed statement added next to the old one leaves two problems in circulation. The old one wins, because it is already in the reference documents.
The traps
The reframing brought in ready-made
The analyst reframes alone, then brings the new statement to the session. Whoever wrote the original hears a public correction and defends their text, whatever its quality. The order that works reverses the two: the owner of the statement writes their definition first, in their own words, and the candidates are produced in front of them from that text.
Established facts lost when the frame changes
Previous analyses established facts: a seasonal pattern, a correlation, a segment that behaves differently. A new frame that starts from scratch loses them. The comparison line of the template says what the old frame had established and where those findings reappear in the new one.
One actor's bias taken for a reframing
The frame of the most influential participant replaces the organisation's, and the work has served to supply an opinion with a method. Collecting the definitions in writing and separately, before any session, leaves a record of what each person thought before hearing the others.
Shaping that stops at the available tool
The statement narrows until it matches the model the team knows how to build or the dashboard it knows how to produce. The stopping criterion for shaping is that a problem-solving process applies, without requiring that it already be in place. A statement that calls for a skill the organisation does not have is still a correct statement: it opens an investment decision, which is handled separately.
The work with no stopping condition
A group that has developed a taste for reframing produces new frames indefinitely, each frame calling for a wider one. The limit is announced at the start of the session: a fixed duration, at most three candidates written up following the template, a decision taken the same day by a named person.
AI considerations
The first useful application is producing candidate frames, where the technique depends most on individual creativity: a language model given the written statement and the six named moves for challenging the context returns about ten reformulations in a few minutes, two or three of which are worth discussing.
The second is making explicit the context a received text implies. Asking a model which actor, which objective, which unit of measure and which horizon a statement assumes without saying so is reading work, mechanical and tedious, where human attention drops off on the third document. The third is setting diverging definitions side by side: ten definitions written separately group by underlying frame faster on a machine than on a whiteboard.
Three reservations hold. A model trained on general text reproduces the dominant framing of the field: asked how a pricing problem is usually stated, it returns the insurer's view, because that is the one the literature writes. The way round it is to impose the move: tell the model to change the actor to one named actor, rather than asking it for a reframing. Next, stakeholder buy-in has to be negotiated: the political cost of a reframing, the identity of whoever loses ownership of the problem and what the sponsor is prepared to hear appear in no text supplied to the model. Finally, a statement nobody signs is worth nothing. A reformulation produced by a machine and presented as it stands arrives before a committee with no author, yet its function is to bind someone to the statement.
Examples
A Swiss health insurer is preparing the submission of its compulsory health insurance premiums (LAMal) for approval by the Federal Office of Public Health (FOPH). The finance department wrote the problem. The four parts of the template give the rows of the table, preceded by the context each frame retains; the comparison, the fourth part, is read across the two columns on the last row.
| Original frame, the insurer's view | Reframed, the policyholder's view | |
|---|---|---|
| Context retained | Annual pricing exercise. Premiums subject to FOPH approval, differentiated by canton and by age band | The policyholder's care pathway: insurance model chosen, deductible from CHF 300 to CHF 2'500, primary care available near where the policyholder lives |
| Problem statement | The costs of the Vaud portfolio exceed the premiums collected by 6.4%. How should the increase be spread across age bands and insurance models? | Some Vaud policyholders use hospital emergency departments for situations a primary care practice handles. Which ones are they, and which would switch to a family doctor model if the offer were put to them? |
| Impacted stakeholders | Finance department, actuarial team, FOPH | Policyholders concerned, primary care physician networks, claims management, customer service, FOPH |
| Approach and data required | Descriptive. Claims costs by age band, by deductible and by model, over three financial years | Predictive then prescriptive. Claims by type of care used, distance to the nearest practice, current model and deductible, history of model changes |
| Comparison: what this frame makes possible | Spread an increase. The amount to be covered is taken as given | Act on the cost itself. A campaign targeted on the family doctor model, already approved with its discount, is an admissible answer with no new premium approval |
The third row shows it: reframing changes who has to be in the room. The original frame is handled between two internal functions and the regulator; the reframed one involves five parties, three of them outside the company. That is where buy-in is paid for: its absence is the most frequent reason a sound reframing goes nowhere.
The fourth part of the template then writes itself in three lines. The original frame spreads an increase taken as given and is decided within the current pricing exercise. The reframed one acts on the cost itself and opens an answer that requires no data analysis, at the price of a longer study and a broader mandate. It was the one chosen because the first brings the same trade-off back the following year.
Visualisations
Two things call for a picture. One statement opens three directions and one of them has to be chosen: root cause analysis descends towards the causes, shaping narrows the statement, reframing moves it into another context. The figure shows them from the same starting statement. The comparison of the two frames is written in columns, the four-part template giving the rows and each frame a column.
A received statement is audited with three questions. Whose problem is this and who wrote it? Which piece of data or which observation would settle the question, and does the organisation hold it? Which answers does the current wording rule out? A statement that fails on one of the three calls for a round of shaping before any study.
Cost
| Phase | Level | Justification |
|---|---|---|
| Preparation | Medium | The mandate to reopen the statement is negotiated before anything else. Collecting the definitions in writing and separately means reaching every stakeholder, which takes one to two calendar weeks for a few hours of work. |
| Execution | Medium | A two-hour session produces the candidates. What costs is what follows: getting whoever owned the initial statement to adopt the new one, which often takes a second pass and a ruling from the sponsor. |
| Documentation | Low | The deliverable is one page long, in a fixed template. The one lasting discipline is to replace the statement in the documents where it lives. |
Tooling
No tool is specific to the technique. What it needs is short text, produced separately and then compared.
The written questionnaire carries the collection of definitions. Any medium will do: Microsoft Forms, a field in the ticketing system, a single-question email, provided each person answers without seeing the others' answers.
The collaborative whiteboard (Miro, Mural, Excalidraw or a wall and A4 sheets) carries the production session. The six moves are laid out there as columns, which forces the group through each one. The format survives remote work.
The tool where the statement lives decides the effect of the work: Confluence, Jira, Azure DevOps, an initiative record, a requirements specification. The work ends when the statement has been replaced there, dated and attributed to a person, with the previous version still readable in the history.
The data catalogue, where the organisation keeps one, is used to check that the data a candidate frame calls for exists somewhere, before anything is decided. An attractive frame resting on data nobody collects is paid for three months later, when the study starts.
Sources
- IIBA, Guide to Business Data Analytics, §3.15 Problem Shaping and Reframing: the definition of the two operations and their interdependence, the predisposition to examine the data before having examined the problem, the two elements (context, problem statement), the four-part template, the strengths and the limitations.
- Thomas Wedell-Wedellsborg, Are You Solving the Right Problems?, Harvard Business Review, January-February 2017: the seven reframing practices, the legitimacy to obtain before reopening a statement and the survey of 106 executives.
- Donald A. Schön, The Reflective Practitioner: How Professionals Think in Action, Basic Books, 1983: the origin of the notion of framing in professional practice and the distinction between setting a problem and solving it.
- Amos Tversky and Daniel Kahneman, The Framing of Decisions and the Psychology of Choice, Science, vol. 211, no. 4481, 1981, pp. 453-458: the framing effect as a decision bias, distinct from the technique described here.