A consulting framework is a sequence of steps that converts a business problem into a recommendation. There are hundreds of them. Most are internally coherent, professionally presented, and taught to intelligent people who apply them diligently. And the majority of engagements that use them produce a deliverable the client praises and does not implement.
That outcome is so common it has stopped being remarkable. Ask a mid-career executive how many strategy documents they have received and how many changed the trajectory of the business, and you will get a ratio that would be scandalous in any other professional service. Nobody blames the framework, because the framework did what it promised. It produced analysis. The analysis was correct.
I want to make a specific argument about why this happens, because the usual explanations — insufficient buy-in, poor change management, weak follow-through — are descriptions of the symptom presented as diagnoses. The real problem is structural, it occurs before any analysis begins, and it is almost never visible in the engagement's own documentation.
The selection problem
Start with where frameworks come from. The dominant method of framework development in business is to study organizations that achieved an outstanding result, identify what they had in common, and formalize those commonalities into a repeatable process. This is the intellectual engine behind a substantial share of the business literature, and it has an obvious defect that we mostly agree not to mention.
The defect is that you are only looking at the survivors. If you study forty companies that grew rapidly and find that thirty-one of them made an aggressive early bet on a single channel, you have learned something interesting and possibly useless. The question you cannot answer from that sample is how many companies made the same aggressive bet and disappeared. If the answer is four hundred, the bet is not a success factor. It is a lottery ticket that the winners happen to be holding.
This is not a novel criticism, and I am not the first person to make it. What interests me is the consequence for practice. A framework built on survivor analysis encodes the behaviors of successful organizations without encoding the conditions under which those behaviors work. It tells you what to do. It cannot tell you when doing it is a mistake, because the sample contains no information about that.
Compare this to how operational doctrine develops. The Incident Command System — the management structure used across American emergency response — was not built by studying successful incident responses and extracting commonalities. It was built by examining failures, identifying the specific coordination breakdowns that caused them, and designing structural controls against each. Common terminology exists because agencies using different vocabularies for the same resource have gotten people killed. Manageable span of control exists because supervisors managing too many subordinates lose track of them. Unity of command exists because personnel receiving orders from multiple supervisors execute neither set well.
That is a fundamentally different evidence base. It generalizes better, because it is derived from the requirements of operating under pressure rather than from the characteristics of entities that happened to do well.
The sequencing problem, which is worse
The selection problem is real but survivable. A framework with a biased evidence base can still be useful if applied thoughtfully. The sequencing problem is not survivable, and it is the actual reason engagements fail.
Nearly every consulting framework begins with analysis. Gather data on the problem, structure it, generate hypotheses, test them, synthesize findings, recommend action. That sequence is sound as far as it goes. The difficulty is what precedes it and is treated as already handled: the definition of the problem itself.
In practice, the problem definition arrives with the client. A prospective client makes contact and describes what they need. That description is the product of their own prior thinking, their organizational politics, their available vocabulary, and their understandable desire to appear competent while asking for help. The consultant, who wants the engagement and who has been trained to be responsive, accepts the definition and proceeds to analyze it rigorously.
Everything downstream is then correct and irrelevant. The data gathering is thorough. The hypotheses are well-formed. The synthesis is sharp. The recommendation follows from the analysis. And the engagement addresses a problem the client invented to describe a situation they had not yet understood.
A concrete case
A professional services firm of about forty people asked for help redesigning partner compensation. Senior staff were dissatisfied, the managing partner said, and the current model was misaligned with contribution. Three prior compensation redesigns had been drafted over four years. None had been adopted.
A conventional engagement takes this at face value, and there is nothing unreasonable about doing so. Compensation design is a well-defined problem with established methods. You would benchmark market rates, model contribution measures, design a formula, socialize it with stakeholders, and deliver a recommendation. The work would be competent. It would be the fourth unadopted compensation model, and the firm would conclude that consultants do not understand professional services.
What the situation actually required was a question about timing: why now? Compensation had been misaligned for four years. Something changed eleven weeks before that call, and the answer was specific. A high-performing senior associate had resigned, and in the exit conversation she said the firm offered no path to ownership. She did not mention money.
That reframes everything. Three failed compensation redesigns were three attempts to answer a question about advancement using the vocabulary of pay. They failed for a reason that had nothing to do with their technical quality: no compensation formula can answer whether someone has a future at a firm. The partners kept rejecting the models because none of them addressed the thing that was actually bothering everyone, and nobody had named the thing.
The engagement was rewritten around partnership track definition and advancement criteria. Compensation was handled later, as a downstream consequence of the track structure. The fourth compensation model was the first one adopted, because it was finally answering a question that could be answered with money.
Why the trigger is the diagnostic
The instrument that unlocked that case is not sophisticated. It is a question about timing, asked with enough persistence to get past the first answer.
The reason it works is that people tolerate problems with remarkable durability. Organizations operate with known dysfunctions for years. Individuals live with conditions they have been intending to address since the previous decade. Tolerance is the default state, and it is stable. So when someone finally acts — books the call, allocates the budget, initiates the conversation — something specific has changed. That change is the trigger, and it carries three pieces of information that the client's problem statement does not.
It dates the urgency, which tells you how much energy is available and how fast it will decay. It identifies the real stakeholder, which is the person whose position actually moved and frequently not the person on the call. And it points at the actual problem, because the trigger is evidence about what the situation is really doing to people, whereas the problem statement is evidence about how they have chosen to describe it.
None of this is available if you accept the presenting problem and begin analyzing. The trigger has to be excavated deliberately, usually against mild resistance, because clients arrive wanting to discuss solutions and experience the timing question as a delay.
The second failure: planning on assumed conditions
Suppose you get the problem right. There is a second structural failure waiting, and it is the more expensive of the two because it surfaces later.
Frameworks generally treat current-state assessment as data gathering in service of analysis. You collect what you need to test your hypotheses. What that omits is verification: establishing that the conditions the plan will depend on are actually true, independent of whether they are analytically interesting.
A regional manufacturer wanted to improve on-time delivery from roughly 82 percent, a figure that appeared in their monthly management report and that everyone in the leadership team could quote. Attempting to verify it revealed that on-time delivery was calculated against the most recently revised promise date rather than the original commitment to the customer. Every schedule slip had been quietly absorbed by the metric. Measured against original commitments, performance was 61 percent.
The problem was three times larger than the problem being managed. Any plan built on the 82 percent figure would have been calibrated to a gap of nine points when the real gap was twenty-one, and would have failed while appearing to be executed correctly. The failure would have been attributed to execution discipline.
Verification of that kind is unglamorous and clients resist it, because it looks like the consultant questioning their competence in week two. It is also where a large share of the engagement's total value is generated. An organization that discovers its core operating metric has a definitional defect has learned something more valuable than any recommendation, and it has learned it cheaply.
What a corrected sequence looks like
The remedy is not a better analytical method. Analytical methods are not the constraint. The remedy is to add two phases before analysis begins and to treat them as gates rather than as preliminaries.
- Establish the trigger. What specific, datable event caused the client to act now rather than continuing to tolerate this? Separate the presenting complaint from the underlying condition from the triggering event, in writing, in the client's own words. Produce a single-sentence engagement statement naming the trigger, the objective, and the cost of inaction. If that sentence cannot be written, do not proceed.
- Verify the current state. Establish what is true, with evidence, before proposing anything. Separate facts from constraints from resources. Reclassify stated constraints that are actually preferences and name their price. Measure and record the baseline for every metric the engagement will be judged by, from source systems rather than management reports.
Only then does analysis begin, and it begins against a verified problem and verified conditions. This is the front half of the framework I use, and the two phases are called Clarify and Comprehend. They occupy the first three to four weeks of an engagement and produce no recommendations at all, which clients find uncomfortable and which is the point.
The structure is borrowed directly from incident action planning, where it is not optional. FEMA's doctrine on the incident action planning process lists Understand the Situation as phase one and Establish Incident Objectives as phase two. You do not set objectives before you understand the situation. In an emergency, the consequence of inverting that order arrives the same day and is unambiguous. In business, it arrives in six months and gets blamed on culture.
The uncomfortable implication
A framework that begins with the trigger question will disqualify engagements. Some clients cannot name a trigger, which means the problem is being tolerated and will continue to be. Some triggers belong to a stakeholder who is not in the room and will not be persuaded by an engagement they did not commission. Some verified constraints make the stated objective unreachable within the client's horizon.
In each case the correct action is to decline, and declining costs the consultant revenue. This is, I think, the actual reason the sequencing problem persists in the profession despite being well understood by thoughtful practitioners. A framework that starts with analysis converts every inquiry into a possible engagement. A framework that starts with the trigger question converts some inquiries into a polite conversation about why this is not the right time.
The commercial incentives run against rigor here, and pretending otherwise is not useful. What I would offer is that the engagements you decline are disproportionately the ones that would have failed, and failed engagements are considerably more expensive than declined ones — not in revenue, but in the only asset a practice actually has.