Your Training Partner
Techniques Toolbox
Invented example from an e-banking team. On the left are the five data sources. In the centre, three themes each carry a negative fact, a positive fact and an effect: "Underestimated use cases", "Late end-to-end tests" and "Missed release slot". On the right, the ranked list leads to the corrected statement.

Pattern Matching

Pattern Matching is an exercise that looks for recurring themes in the data already gathered on a problem. The team groups this data by theme and notes the effect of each theme. Each theme receives positive facts as well as negative facts. The team then compares the themes with the problem statement and confirms or corrects the statement before anyone looks for solutions. George W. Anderson describes the exercise in Design Thinking for Tech (2023).

Purpose

Pattern Matching holds a problem statement up against the data the organisation already has on that problem. A pattern is a fact or a sequence of facts that repeats: the same delay from one release to the next, the same complaint from one department to the next. According to Anderson, recurring themes validate the problem or show how the team thinks and works.

Anderson places the exercise among four exercises that validate a problem. The other three are Verbatim Mapping, AEIOU and the Five Whys. The team runs it after the statement is written and before ideation. It informs a decision: does the statement hold, does it need correcting, which theme should the team focus on next?

Anderson's steps produce named themes, their effects and a consolidated list. Our format adds a theme table and a written verdict on the statement.

In case study research the same name refers to a different method. There the researcher compares an observed pattern with an expected pattern. The observed pattern is a set of measured values. The expected pattern is a precise hypothesis. The Encyclopedia of Case Study Research contrasts this method with pattern recognition, which builds a theory from the data. Anderson's Pattern Matching draws the themes from the data, then compares them with the statement.

Usage

When to use it

  • Problem statement written: test it against the data before ideation.
  • Recurring problem: delays, escalations or incidents that repeat from one release to the next.
  • Data already available: tickets, minutes, Verbatim Mapping logs and risk reviews suffice and no new session is needed.
  • Doubts about the team's view: check whether its way of seeing things keeps producing the same problems.
  • Options or principles to group: look for links less obvious than similarity, for example grouping the options by region.
  • Distributed team: a shared spreadsheet or an online whiteboard carries the exercise remotely.

When not to use it

Description

The opening questions

Anderson's first question concerns the team's view. Does the way the team sees a situation shape its response so that the same problems keep coming back? Does a problem of perception hide the actual problem?

The second concerns the shape of the pattern. In a singular pattern the same result comes back. The reason changes from one occurrence to the next. Anderson cites releases that are always late for one reason or another. In a stepwise pattern each effect triggers another. Use cases that are more complex than expected lengthen development and then testing. The release slips.

Singular pattern
Release 3.1: test server unavailable
Late
Release 3.2: late compliance approval
Late
Release 3.3: late delivery from the supplier
Late
Release 3.4: absences during the holidays
Late
The team looks for what the occurrences share
Stepwise pattern
Use cases more complex than expectedThe corrected statement targets this link
Longer development
Longer test cycles
Late release
In a singular pattern the same delay comes back with a different reason in each release. In a stepwise pattern each effect triggers the next and the corrected statement targets the first link. The chain follows Anderson's example. The reasons on the left are invented.

The answer tells the team how to correct the statement. Our reading: in a stepwise pattern the corrected statement targets the first link rather than the final delay. In a singular pattern the team looks for what all the occurrences share: often a way of planning or of committing to a date.

Participants and duration

Anderson allows for one to five people and 15 to 60 minutes per problem. A broad or complex problem takes longer. In our practice the group has at least two people. One knows the area where the problem arises. The other took no part in writing the statement. One person prepares the data before the session. The group forms the themes and gives the verdict.

The input data

Anderson warns that without good data the search for patterns becomes difficult. As sources he names documented incidents, earlier interviews and Verbatim Mapping exercises, observations, listening sessions, risk reviews and minutes.

The analyst prepares one row per fact with its source and its date.

Fact list. One row per fact. Invented example.
FactSourceDate+ / −GroupTheme
The actual effort on 4 of 6 use cases was double the estimated effort.Retrospective minutes27.02.2026−CausesUnderestimated use cases
The 2 use cases reviewed in a workshop with compliance held their estimate.Interview with a Product Owner18.03.2026+CausesUnderestimated use cases
The test campaign starts on average 4 days after the planned date: development is not finished.Release ticket09.04.2026−ConsequencesLate end-to-end tests
2 of 6 releases pushed back by a week: testing is not finished by the Thursday slot.Risk review22.06.2026−ConsequencesMissed release slot

At verdict time the team can no longer check a fact that has no source. Our rule: a theme rests on at least two distinct sources. A source is a type of data: tickets, retrospective minutes, interviews. Forty tickets count as one source. A theme drawn from a single source remains a hypothesis.

The steps

The steps are Anderson's. The guidance that goes with them is our reading.

  1. Gather the data. The team puts the data into a common form: one note, one card or one spreadsheet row per fact.
  2. Group what is alike. The team starts with broad groups. Anderson suggests pairs: positives and negatives, inputs and outputs, symptoms and effects, causes and consequences. The team keeps the pair that sheds light on the problem.
  3. Sort the groups by theme. Within each broad group the team brings together the facts that concern the same thing. When no pattern shows in the actions or their consequences, Anderson advises looking for recurring nouns (deadlines, confidence, turnover, meetings, money, schedule, escalations) or verbs (stressed, rushed, hate, think, spent, talked).
  4. Name each group. The team names each theme in the words of the data rather than those of the statement.
  5. Identify the effects. For each group the team notes what it produces. Anderson gives examples of effects: positive, negative, uncertainty, budget overruns, high quality, schedule implications, team attitude, poor focus, healthy competition, high stress.
  6. Consolidate and rank. As a group the team merges similar effects and sharpens their wording. It can also rank the themes, for example as a top-five list.

In steps 2 to 4 the team groups and names as in an affinity map. There you can see how a group moves notes on a wall and names them. Pattern Matching starts from existing data rather than from ideas written in the session. It adds the broad starting groups, the effects, the ranked list and the positive and negative facts of each theme.

Positive and negative facts in every theme

Anderson asks for positive facts and negative facts in each theme. His examples are good and poor quality, ahead of and behind schedule. In his view these extremes can bring out instructive patterns and show which groups to dig into.

In our reading the positive fact shows where the problem does not occur. If two of six use cases held their estimate, the team knows what to compare. A theme without a positive fact remains a complaint. The team looks for the counter-example in the sources before it closes the theme.

What the team delivers

Our format fits in one table with one row per theme in rank order.

Columns of the theme table. Proposed format.
ColumnContent
RankThe theme's place in the consolidated list from step 6. Our ranking rule: in a stepwise pattern the themes follow the order of the chain from the first link. Otherwise the theme that rests on the most sources comes first.
ThemeThe name the team draws from the words of the data.
Negative factsThe negative facts, each with its source.
Positive factsThe positive facts, each with its source.
SourcesThe number of distinct sources. A source is a type of data. A single source makes the theme a hypothesis.
EffectWhat the theme produces: delay, budget overrun, stress, high quality.

Below the table the team writes the verdict on the statement. The statement holds when the top-ranked themes confirm it. When the words of a theme point to another problem, the team rewrites the statement in those words. The statement stays open when data is missing. The team then notes which data to gather.

Before and after the exercise

The statement comes from problem shaping. The log of a Verbatim Mapping exercise supplies part of the data. Once a theme is confirmed, the Five Whys trace its cause. Ideation then starts from the validated statement.

Pitfalls

These pitfalls come from our practice.

Themes named after the statement. A team that knows the statement files the facts under its words. The themes then confirm the statement whatever the data says. The team names the themes after sorting. A person who did not write the statement rereads each name.

One source that drowns out the others. When the team sorts by eye, forty tickets weigh more than three interviews. The themes turn technical because the tickets are. The team counts the sources of each theme rather than its facts.

The effect mistaken for a theme. The release delay is the most visible and best documented effect. A "Delays" theme repeats the statement and teaches the team nothing. The team files this fact among the effects and looks for what comes before it.

AI considerations

A language model sorts hundreds of tickets or lines of meeting notes in a few minutes. In step 3 the analyst asks it to count the recurring nouns and verbs. The model proposes groupings. For each theme it looks for a positive fact and a negative fact in the sources. The analyst demands the source and the original line for every fact.

A model sometimes invents a fact or a quotation. Only the facts the analyst finds in the source go into the table. The model also echoes the wording of the prompt: if the statement is in it, the themes repeat it. So the analyst submits the data without the statement. The theme names, the effects, the ranking and the verdict stay with the team that knows the domain.

Tickets and minutes quote customers, employees and opinions about management. These are personal data within the meaning of the Federal Act on Data Protection (FADP). The analyst submits them to a service the organisation has approved. For any other service they first remove the names and the customer data.

Examples

The case is invented. The e-banking team of a cantonal bank shipped six releases in the first half of the year. Two of them slipped by a week. The steering committee wrote the statement "Our releases are late." The team holds it up against five sources. It first separates causes from consequences.

Statement to testInvented example"Our releases are late."Steering committee
Data (step 1)
TicketsRelease tickets
RetrosRetrospective minutes
VerbatimVerbatim Mapping log of the testers
RisksRisk review
PO interviewsProduct Owner interviews
Themes (steps 2 to 5)
Starting groups: causes and consequences
Underestimated use cases
−The actual effort on 4 of 6 use cases was double the estimated effort. Retros
+The 2 use cases reviewed in a workshop with compliance held their estimate. PO interviews
Effect Longer development2 sources
leads to
Late end-to-end tests
−The test campaign starts on average 4 days after the planned date: development is not finished. Tickets
+Automated regression runs in 2 hours. Verbatim
Effect Compressed testing2 sources
leads to
Missed release slot
−2 of 6 releases pushed back by a week: testing is not finished by the Thursday slot. Risks
+A package delivered on Tuesday makes the Thursday slot. Tickets
Effect One-week delay2 sources
Ranked list (step 6) and verdict
1Underestimated use cases
2Late end-to-end tests
3Missed release slot
Stepwise pattern
Corrected statement:"Our estimates do not cover the complexity of the use cases."
Next: the Five Whys on theme 1
Invented example. The team tests the statement "Our releases are late" against five sources. Each theme carries a negative fact, a positive fact and an effect. The themes form a chain: the corrected statement targets the first one.

Each theme leads to the next: the pattern is stepwise. The corrected statement therefore targets the first link: "Our estimates do not cover the complexity of the use cases." The positive fact of the first theme shows where to look. The two use cases reviewed in a workshop with compliance held their estimate. The team applies the Five Whys to this theme.

Visualisations

The diagram of the singular and the stepwise pattern shows what the second opening question distinguishes. The fact list shows the input data with one row per fact. The columns of the theme table give the format of the deliverable. The example diagram follows the data through to the verdict on the statement.

Cost

DimensionLevelJustification
PreparationMediumExtract the facts from several sources and note the source and date of each.
ExecutionLowOne to five people, 15 to 60 minutes per problem according to Anderson.
DocumentationLowThe theme table and the verdict fit on one page.

Tooling

A spreadsheet holds the fact list. A filter or a pivot table counts the sources of each theme. An export from the ticketing tool (Jira, ServiceNow) supplies the documented incidents. In a session the team sorts on a whiteboard with sticky notes. A distributed team uses an online whiteboard (Miro, Mural). With more than about ten long documents, qualitative analysis software (MAXQDA, ATLAS.ti, NVivo) keeps each theme linked to its passages.

Sources

  • George W. Anderson, Design Thinking for Tech: Solving Problems and Realizing Value in 24 Hours, Addison-Wesley, 2023, ISBN 978-0-13-793303-7. Hour 9 "Identifying the Right Problem", section "Design Thinking in Action: Pattern Matching for Themes", pp. 130-131: the definition, the opening questions, the participants and the duration, the steps, the positive and negative facts of each theme. Hour 3, section "Thinking and Problem Solving", p. 30: the definition of a pattern. Hour 4, section "Guiding Principles for Operating Consistency": the grouping of principles by theme. Hour 13, section "Affinity Clustering for Pattern Finding": the grouping of options by less obvious links.
  • "Pattern Matching", in A. J. Mills, G. Durepos and E. Wiebe (eds.), Encyclopedia of Case Study Research, SAGE, 2010, pp. 664-665: the method of the same name in case study research.
Pass Around
All techniques
Payback Period