Concept Map

Thinking Tools1.1 Describing the SituationAlso known as Concept Mapping

Developed by Joseph Novak at Cornell from 1972 (Learning How to Learn, with D. Bob Gowin, 1984).

Arranges the ideas in a subject from general to specific and joins them with arrows labeled by relationship, producing a picture of how the pieces fit together.

A meal-kit subscription business mapped from general to specific, every link a sentence

The most general idea sits at the top; each line carries the words that join two ideas into a claim, printed at the top of the lower box. Select a box to read its claim, then select the links under the map that join one branch to another, where most of the team's insight turned out to be. Illustrative.

Links between branches:

Select a box to read the claim its link makes, or a link under the map to see two branches that touch.Meal-kit subscription. The most general concept, at the top. Read each link downward as a sentence: the business serves active customers.Active customers. Claim: the business serves active customers. The count that every other branch ends up touching.New sign-ups. Claim: active customers are added by new sign-ups. About 2,000 a month.Cancellations. Claim: active customers are lost through cancellations. About 1,700 a month, most in the first eight weeks.Recurring revenue. Claim: the business earns recurring revenue.Plan price. Claim: recurring revenue is set by plan price.First-box discounts. Claim: recurring revenue is reduced by first-box discounts, half off the first two boxes.Delivery. Claim: the business pays for delivery, its largest cost after food.Box weight. Claim: delivery costs more with box weight. Ice packs are a third of it.Carrier choice. Claim: delivery depends on carrier choice. The cheaper regional carrier arrives late one time in eight.Customers times price is revenue. The two top branches are one quantity seen twice: revenue is active customers times the average price paid.Discounts bring short stays. The link the team had not drawn: customers who join on a half-price offer cancel most often in the first eight weeks.Late boxes cause cancellations. A box that arrives late is the reason given in one cancellation in five, so the cheaper carrier costs more than it saves.

Three links from the map, each waiting for its linking words

A concept map's links carry words that turn two boxes into a sentence. The first link is done as an example; the other two start with the vague words teams often write. Pick better words for each, and the lines under the rows read each link as a sentence and say whether it makes a claim someone could check. Illustrative.

Cancellations ... active customers

First-box discounts ... cancellations

Carrier choice ... cancellations

  • Pick the linking words for a link."Cancellations reduce active customers." A proposition: one direction, and the monthly counts can confirm it."Cancellations relate to active customers." True of almost any pair of boxes, so it claims nothing anyone could test."Cancellations increase active customers." A proposition, and a false one: the words point the wrong way."First-box discounts raise cancellations." A proposition the cohort data can test: half-price joiners leave most often in their first eight weeks."First-box discounts affect cancellations." Says something happens without saying which way; nobody could prove it wrong."First-box discounts are linked with cancellations." A correlation word: it does not say which way the effect runs, so the map cannot act on it."The cheaper carrier raises cancellations, through late boxes." A proposition, and it names a concept the map was missing: late deliveries. Add it as a box between the two."Carrier choice matters for cancellations." Points at a link without saying what it is."Carrier choice lowers cancellations." A proposition, and the data says otherwise: the cheaper carrier is late one time in eight.
  • Pick the linking words for a link."Cancellations reduce active customers." A proposition: one direction, and the monthly counts can confirm it."Cancellations relate to active customers." True of almost any pair of boxes, so it claims nothing anyone could test."Cancellations increase active customers." A proposition, and a false one: the words point the wrong way."First-box discounts raise cancellations." A proposition the cohort data can test: half-price joiners leave most often in their first eight weeks."First-box discounts affect cancellations." Says something happens without saying which way; nobody could prove it wrong."First-box discounts are linked with cancellations." A correlation word: it does not say which way the effect runs, so the map cannot act on it."The cheaper carrier raises cancellations, through late boxes." A proposition, and it names a concept the map was missing: late deliveries. Add it as a box between the two."Carrier choice matters for cancellations." Points at a link without saying what it is."Carrier choice lowers cancellations." A proposition, and the data says otherwise: the cheaper carrier is late one time in eight.
  • Pick the linking words for a link."Cancellations reduce active customers." A proposition: one direction, and the monthly counts can confirm it."Cancellations relate to active customers." True of almost any pair of boxes, so it claims nothing anyone could test."Cancellations increase active customers." A proposition, and a false one: the words point the wrong way."First-box discounts raise cancellations." A proposition the cohort data can test: half-price joiners leave most often in their first eight weeks."First-box discounts affect cancellations." Says something happens without saying which way; nobody could prove it wrong."First-box discounts are linked with cancellations." A correlation word: it does not say which way the effect runs, so the map cannot act on it."The cheaper carrier raises cancellations, through late boxes." A proposition, and it names a concept the map was missing: late deliveries. Add it as a box between the two."Carrier choice matters for cancellations." Points at a link without saying what it is."Carrier choice lowers cancellations." A proposition, and the data says otherwise: the cheaper carrier is late one time in eight.
  • "Cancellations reduce active customers." A proposition: one direction, and the monthly counts can confirm it. (l1: reduce)
  • "Cancellations relate to active customers." True of almost any pair of boxes, so it claims nothing anyone could test. (l1: relate to)
  • "Cancellations increase active customers." A proposition, and a false one: the words point the wrong way. (l1: increase)
  • "First-box discounts raise cancellations." A proposition the cohort data can test: half-price joiners leave most often in their first eight weeks. (l2: raise)
  • "First-box discounts affect cancellations." Says something happens without saying which way; nobody could prove it wrong. (l2: affect)
  • "First-box discounts are linked with cancellations." A correlation word: it does not say which way the effect runs, so the map cannot act on it. (l2: are linked with)
  • "The cheaper carrier raises cancellations, through late boxes." A proposition, and it names a concept the map was missing: late deliveries. Add it as a box between the two. (l3: raises, via late boxes)
  • "Carrier choice matters for cancellations." Points at a link without saying what it is. (l3: matters for)
  • "Carrier choice lowers cancellations." A proposition, and the data says otherwise: the cheaper carrier is late one time in eight. (l3: lowers)

An example

A meal-kit company's new general manager inherited three teams who each explained the business differently: growth talked about sign-ups, finance about revenue per box, and operations about delivery cost. She asked them to build one concept map together on a whiteboard.

They listed about twenty concepts, ranked them from general to specific, and put the business at the top with three branches under it: the customers it serves, the revenue it earns, and the delivery it pays for. Every line had to carry a few words that turned the two boxes into a sentence a person could dispute. The useful part came when they drew links across branches. Growth's first-box discount was tied to the cancellations on the customer branch, since half-price joiners left fastest, and operations' cheaper carrier was tied to cancellations too, since late boxes were the reason given in one cancellation in five. The company dropped the deepest discount and moved its two largest regions back to the faster carrier.

Why it works

Lists and outlines put each idea in one place, so they hide the relations between ideas that sit in different places. A concept map forces every connection to be written as a labeled link, and a labeled link is a claim: "discounts bring short stays" can be checked, argued over, and found wrong. An unlabeled line between two boxes cannot.

Arranging concepts from general to specific gives the map a spine, and the cross-links between branches are where understanding shows. A person who can draw them sees how the parts of a subject touch; a group drawing them together finds the places where its members hold different models of the same business, which are usually the places worth investigating first.

The thinking behind it

Joseph D. Novak Joseph D. Novak, an education researcher at Cornell, developed concept mapping in the 1970s: diagrams of concepts joined by labeled links, where each pair of concepts and its link states a proposition. He used the maps to show what a learner understands and to help build meaningful learning, and he set the method out with D. Bob Gowin in Learning How to Learn (Cambridge University Press, 1984).

How to run it

  1. Write the focus question. State what the map is meant to explain, such as how the business makes or loses money.
  2. List the key concepts. Write fifteen to twenty-five nouns or short noun phrases, one per sticky note.
  3. Rank them from general to specific. Put the broadest few at the top and the most concrete at the bottom.
  4. Link related concepts. Draw a line between two concepts whenever one bears on the other.
  5. Label every link. Write the few words that make the pair a sentence, such as causes, requires, or is reduced by.
  6. Hunt for cross-links. Look for connections between branches, and write each as a labeled link like the rest.
  7. Test the claims. Read each link aloud as a sentence and mark the ones the group disputes or cannot support with data.

Where it goes wrong

The links have no words. A web of plain lines looks thorough and says nothing that can be checked. Refuse any line without a label, and rewrite vague labels like "relates to".

The map is a tree. Without cross-links, a concept map is an outline in boxes, and the insight between branches never appears. Set aside time for cross-links only once the hierarchy is drawn.

It grows without end. Sixty concepts on one board stop anyone from reading the whole map. Hold to the focus question, and start a second map for whatever falls outside it.

When to use something else

  • Mind Mapping. When the aim is to generate ideas around one topic fast, and the relations between them can wait.
  • Causal Loop Diagram. When the links are causes that feed back on each other, and the question is how the system behaves over time.
  • Issue Tree. When one question must be broken into parts that do not overlap, so each can be answered.

Where it sits in SOLVE eX

Go to the source

  • Learning How to Learn, with D. Bob Gowin, Cambridge University Press, 1984

Back to the catalog