The Account That Was Coherent and Incomplete
He told the story in under four minutes, and it held together. The hire had looked strong on paper. The interview had gone well. The first month produced good work. Then the second month fell apart, the client complained, and the contract ended badly.
The story had a villain, a timeline and a lesson: senior hires from big companies cannot operate without support staff.
I asked one question. What did the client actually complain about?
He paused, then reconstructed it. The complaint was about response time, not capability. The hire had been waiting on a decision from the founder for three days. The founder had been travelling and had not seen the message.
The lesson changed in that moment. The original account was not a lie. It was a model built from three mental processes that run constantly and invisibly: deletion, distortion and generalisation.
Bandler and Grinder borrowed these three terms from transformational grammar and general semantics, and made them the centre of their modelling work. They argued that every person builds a simplified internal map of experience, and that the map is produced by omitting some details, altering others and extending conclusions beyond the evidence (Bandler & Grinder, 1975, The Structure of Magic I). Alfred Korzybski had made the philosophical case decades earlier with the phrase that became the movement's shorthand: the map is not the territory (Korzybski, 1933, Science and Sanity).
Every founder runs on a map. The question is where the map was drawn badly, and how much money the bad drawing costs.
Deletion: The Mind Drops What It Cannot Use
Deletion is the most useful and most dangerous of the three. Without it, experience would be unusable. The brain receives far more sensory data than it can process, and attention selects a narrow slice for conscious handling.
Christopher Chabris and Daniel Simons demonstrated the scale of this in the famous invisible gorilla experiment. Participants counting passes between players failed to notice a person in a gorilla suit walking through the scene (Simons & Chabris, 1999, Perception). The gorilla was not hidden. Attention had deleted it.
In business, deletion shows up as the missing data point in a decision memo, the customer objection nobody recorded, the team member who stopped speaking three meetings ago. It rarely feels like blindness. It feels like a clean picture.
The NLP Meta Model treats deletion as a cue for a specific question. When a sentence omits the subject, the object or the comparison, the listener can ask for the missing piece. Bandler and Grinder catalogued these patterns and the recovery questions that go with them (Bandler & Grinder, 1975). The question is not a challenge. It is a request for the part of the map that was left blank.
Three deletions matter most in a founder's week.
- Unspecified nouns. The client said the service was poor. Which part, on which date, measured how?
- Missing referents. The team is unhappy. Which people, and what did they say?
- Deleted comparisons. This is too expensive. Compared with what alternative, at what total cost?
None of these questions requires sophistication. They require the discipline to notice that a sentence has a hole in it.
Distortion: The Mind Alters What It Keeps
Distortion changes the shape of what survives attention. Memory is reconstructive, and the reconstruction is shaped by expectation, emotion and the questions asked afterwards.
Elizabeth Loftus and John Palmer showed how a single verb altered what witnesses remembered. Participants who were asked how fast a car was going when it smashed into another car estimated higher speeds than those asked when it hit the other car, and a week later were more likely to report broken glass that had never been in the film (Loftus & Palmer, 1974, Journal of Verbal Learning and Verbal Behavior). The question changed the memory.
Frederic Bartlett had demonstrated the same reconstructive tendency decades earlier, showing that participants reshaped unfamiliar stories toward the conventions of their own culture when retelling them (Bartlett, 1932, Remembering, Cambridge University Press).
Distortion also runs through judgement. Amos Tversky and Daniel Kahneman showed that people rely on heuristics that systematically bend estimates: availability, anchoring, representativeness (Tversky & Kahneman, 1974, Science). A founder who has recently lost a deal will overestimate the risk of the next one. A founder who has recently won will underestimate it. The evidence has not changed. The weighting has.
In NLP terms, distortion includes mind reading, cause and effect claims, and complex equivalence. The recovery questions are direct. How do you know what they think? What specifically caused that? How does that action mean that intention?
The commercial cost is easiest to see in hiring and in pricing. A founder who distorts a candidate's confidence into competence pays for the error for a year. A founder who distorts a client's silence into agreement discovers the truth at renewal.
Generalisation: The Mind Extends a Single Case
Generalisation is how learning travels. One hot stove teaches caution about all stoves. Without it, every experience would be new and nothing would transfer.
The failure mode is over-extension. A single bad hire becomes a rule about an entire category of person. A single failed launch becomes a rule about the market. A single difficult client becomes a rule about an industry.
Raymond Nickerson's review of confirmation bias documented how readily people seek and interpret information that supports an existing generalisation while ignoring what contradicts it (Nickerson, 1998, Review of General Psychology). The generalisation, once formed, recruits the evidence.
NLP treats universal quantifiers as the audible signature of over-generalisation: always, never, all, none, every, nobody. The recovery question is the counter-example question. Has there ever been a time when that was not true?
The question works because it forces the map to include a case it had excluded. A founder who says we never get referrals from existing clients will usually produce one within thirty seconds of being asked. The generalisation was a feeling with a grammatical costume.
There is a strategic version of this that costs more. A founder who generalises from one market cycle will build a company for the last cycle. The rule feels like experience. It is a sample size of one, dressed as wisdom.
The Three Filters in a Strategy Meeting
Consider a strategy session where the leadership team is deciding whether to enter a new segment. The discussion sounds rigorous. In practice, three filters are doing most of the work.
Deletion shows up as the absent customer. The team reviews win rates, pipeline and competitor pricing. Nobody has spoken to a buyer in the segment for six months. The data that exists is the data that was convenient to collect, and the most important variable, whether anyone actually wants the thing, has been dropped from the room.
Distortion shows up as the story about the competitor. A single loss is narrated as evidence that the competitor is buying the market. The narration includes intent, strategy and budget, none of which anyone has verified. The team then plans against a character they invented.
Generalisation shows up as the rule from last time. We tried something like this in 2019 and it failed. The 2019 attempt had a different product, a different price and a different team, and the rule survives because nobody separates the variables.
The fix is unglamorous. Assign one person to each filter for the meeting. The deletion officer asks what we have not heard. The distortion officer asks how we know what we claim to know about other people's intentions. The generalisation officer asks which specific case the rule came from and whether it still applies.
This is a temporary role, not a personality. Rotating it prevents the officer from becoming the team sceptic, a role that gets ignored within a month. The purpose is to make the filters visible while the decision is still cheap to change.
Bandler and Grinder designed the Meta Model questions for a therapist working with one person's map. The same questions scale to a management team, provided somebody is assigned to ask them out loud. In most companies, nobody is.