Chapter 8The Interference Method: Narrow the Other Party's Options

1. After you can see clearly, the other party still has to understand #

On July 27, 2016, Xiaomi launched the first Xiaomi Notebook Air. The published spec for the 13.3-inch model was 14.8 millimeters thick and 1.28 kilograms. Millimeters and kilograms, standing alone, have no comparison base. The launch put the machine next to a 13.3-inch MacBook Air: Apple's published height range was 3 to 17 millimeters, and the weight was 1.35 kilograms. From the two published specs, 14.8 millimeters is about 13% less than 17 millimeters, and 1.28 kilograms is about 5% less than 1.35 kilograms (Xiaomi, 2016; Apple, 2015).

Placing products of the same size side by side gave the two sets of numbers a shared reference. The conditions of the comparison still have to be complete. The MacBook Air is a wedge; 17 millimeters is the maximum in the published height range. Xiaomi published 14.8 millimeters as overall thickness. The ratio can only describe the difference in maximum thickness in the two published specs. It cannot stand in for a test in actual use.

Reports of the same day's launch also recorded the line "thinner than a one-fen coin." The picture compared the diameter of a coin stood on edge, while "how thin a coin is" can also mean the coin's own thickness. The expression did not name the dimension of comparison, and so it is ambiguous. True numbers that rest on a mixed reference can still mislead in the overall presentation.

The Clarity Method helps an organization find gaps in judgment, borrow a verified standard, check whether it can migrate, and then, with cheap trials, expand its own choices. An outside party that has been given parameters, terms, and similar options has not, for that reason, obtained the comparison conditions a decision requires.

The Interference Method handles that distance. Narrowing does not mean hiding other plans, and still less turning the seller's answer into the only option. It means cutting irrelevant information and needless comparison, and putting first the differences a decision needs. When the conditions of fit do not hold, the recommendation should be "do not recommend this plan."

Herbert Simon, writing in 1971 about information-rich organizations, argued that information consumes the receiver's attention, and that more information therefore requires a tighter allocation of limited processing time (Simon, 1971, pp. 40–41). His object was organizational information processing, not marketing effect. The directly relevant conclusion is only this much: receiving, checking, and comparing information all take time, and adding information does not automatically add the capacity to process it.

Results of communication have to be split into at least five measures.

Attention is measured by whether a person stops and processes the information. Exposure, clicks, and dwell cannot prove understanding.

Understanding is measured by whether they can restate accurately what problem the plan solves, under what conditions it holds, and what it costs. An accurate restatement is not a purchase.

Trust is measured by whether, while the result is still uncertain, they enter a trial, a contract, or an agreed risk on the basis of the claim and its source. Acting on evidence does not mean the plan fits.

Persuasion is measured by whether choice, ranking, or a commitment to act changes after contact with the information. Change is not yet a completed sale.

Consensus is measured by whether a group of related people begin to reuse similar names for the problem, comparison standards, and judgments that can be tested. One personal choice does not make consensus.

There is no necessary conversion from one of the five to the next. A purchase record also cannot separate product utility, price, inertia, switching cost, and missing substitutes. Folding the five into "conversion" hides which link actually failed.

2. 10/10/80 is resource discipline, not a statistical law #

The Interference Method has three parts: analyzing the consensus users already hold (10), finding a suitable scene (10), and forming new consensus (80).

The three numbers set where the weight of resource goes. They are not an average computed from successful companies, and they do not require every project to log ten hours, ten hours, and eighty hours. The first two parts are used to obtain the user's actual materials of judgment and a scene that can be checked. Most of the work is left for presenting, verifying, trying, revising, and delivering.

The three parts cannot be added as if they were independent. Without the first two 10s, later output may not correspond to the user's own words, choice records, and actual scenes. Without the later 80, research and scene selection stop in meeting notes. Medicine, finance, safety, regulation, and other irreversible decisions need more investigation up front, more disclosure of risk, and more professional verification. 10/10/80 cannot replace those duties.

Figure: 10/10/80 marks where the weight of resource goes, not a precise ratio; attention, understanding, trust, persuasion, and consensus are checked separately; four behavioral standards constrain the whole process.

The Interference Method is not another name for marketing. Marketing also includes product, price, channel, and service. Communication only asks that information arrive. Persuasion checks whether choice, ranking, or a commitment to act has changed. Choice architecture arranges options, defaults, and order of presentation. The Interference Method handles a narrower problem: under information that is true and complete enough, lower the cost of recognizing, understanding, and comparing the relevant plans.

Practices that rely on concealment, coercion, or exploiting a vulnerability to bypass informed judgment count as manipulation, not as the Interference Method. The boundary is judged by four conditions: whether material facts are disclosed, whether alternatives can be compared, whether refusal is free, and whether exit is convenient.

3. First get clear how the other party judges (10) #

To analyze the consensus users already hold, collect the names they actually use for the problem, the standards they compare on, and the record of their choices. A category word the seller invented cannot stand in for those.

Existing consensus may show as affirmation, refusal, internal contradiction, or as not yet formed. Without the user's own words, a choice record, or behavioral data, "unknown" must not be written as "already agreed."

The principal-leader should look at least at five kinds of material.

The first is the name of the problem. Record the words that actually appear in interviews, searches, support tickets, and procurement files. The seller may say "AI-native organizational rebuild"; the customer's materials may say "software purchase," "process consulting," or "a layoff project." Different names point to different objects of comparison and different requirements of evidence.

The second is the standard of judgment. Record speed, total cost, risk, compatibility, data export, and conditions for an internal handoff as they appear in score sheets, RFPs, trials, and reasons for refusal. A product advantage that never entered the actual comparison cannot replace those standards.

The third is the real alternatives. Continuing the old process, staffing internally, delaying, and doing nothing all belong in the comparison, not only rival vendors. Leaving those options out inflates how much product features explain the choice.

The fourth is the material concerns. Record which changes of fact have led to refusal, delay, or a switch. Budget, implementation cycle, exit cost, safety responsibility, and limits of use all sit here.

The fifth is the trust entry. Record the raw data, peer cases, regulatory permissions, and trial results the user actually asked for or used. The name of a celebrity or an institution cannot replace the evidence itself.

These materials should come from the user's own words, actual choices, records of refusal, and alternatives already in use. They cannot come only from a persona in a conference room. Acceptance should write at the same time who fits, who does not, and which new facts would change the recommendation.

When the materials show that the plan does not fit, stop recommending it. Assuming in advance that "the plan that fits best must belong to the seller" does not meet the definition of the Interference Method.

4. Put a real difference in a scene that can be understood (10) #

The same fact, placed in different scenes, will be read differently. The scene decides which information can be shown, what background travels with it, and what path of verification is offered.

The first item is who is speaking. A seller's statement has a direct interest attached, and verification material should travel with it. A third-party review is not, by itself, reliable either. Method, sample, funding, and range of use still have to be checked. The identity of the source decides where verification starts. It does not decide whether the fact is true.

The second item is when it is said. Information should enter a procurement, evaluation, or switching node that is actually happening. It should not create a node out of fear or false urgency. When there is no corresponding decision and no act of acceptance, adding material only adds processing cost.

The third item is where it is said, and in what form. A launch is suited to showing a difference in one place. A product site is suited to verifying results in use. A technical annex is suited to stating the method of measurement and the limits. Display and verification do different jobs. One cannot replace the other.

The fourth item is how the comparison is made. Both sides should, as far as possible, use the same object, time, denominator, setting, and method of measurement. When a fully shared definition cannot be had, write the difference down. Otherwise each number may be true on its own, and the whole impression they make together may still be untrue.

A suitable scene should hold, at the same time, a real decision, disclosure of the basis, a statement of limits, and a comparison of alternatives. An unsuitable scene weakens or distorts understanding. There is no evidence for a general conclusion that "the wrong choice is zero" or that "credibility differs by an order of magnitude."

The output of the first two 10s is a record of existing consensus and a scene that can be checked. New understanding and new consensus still have to be formed by later presentation, verification, use, and delivery.

5. Forming new consensus (80): a shared language that can be tested #

Market consensus also comes from records of use, word of mouth, independent reviews, and descriptions users begin to use on their own. It is not all designed by the organization. An organization can raise how visible important facts are, and how easily they can be restated. Actual results can still overturn the account the organization puts forward. This step is therefore called "forming new consensus," not "manufacturing consensus."

Forming new consensus takes most of the resource because a positioning line only finishes the opening. The organization still has to state the claim, supply the basis, arrange a trial or a test, record where restatement drifts, repair ambiguity, and keep delivering after handoff. Repetition by itself does not produce consensus. A judgment that can be restated, checked, and revised by results may become a shared language.

Crossing the Chasm supplies a case of naming that became a shared language. In Diffusion of Innovations, first published in 1962, Everett Rogers had already discussed innovators, early adopters, the early majority, the late majority, and laggards as a system. Geoffrey Moore did not invent those five groups. In Crossing the Chasm, published in 1991, he put the weight on the break between early adopters and the early majority: the former may adopt while the product is still incomplete; the latter usually require usability, references of the same kind, and a join to work already in place (Rogers, 1962; Moore, 1991).

"The chasm" turned scattered experience into a hypothesis that can be checked: records of adoption by early adopters cannot be extrapolated straight to the early majority.

In the author's note to the third edition, Moore looked back: the book had been expected to sell about five thousand copies, and by the late 1990s it had sold more than three hundred thousand; by the 2013 revision, including foreign-language editions, it had passed six hundred thousand. He recorded readers passing it on, venture investors and founders sharing the vocabulary, and some companies putting it on a shared reading list (Moore, 2014). Those materials prove that "the chasm" was systematized and widely spread. They cannot prove that Moore coined the word without dispute, and they cannot prove that a whole industry judges in this language.

A useful shared language is not a monopoly language. It explains the facts better than the old wording, and it still allows the facts to keep revising it.

6. Truth is not enough: four standards of Treat People with Honesty #

Limited space always requires a cut. Risk comes from four practices: showing only the favorable part and omitting a fact that would be enough to change the decision; choosing an object that cannot be compared fairly; deleting laboratory conditions; placing a qualification where it is hard to find. Literal truth does not remove those risks.

The law supplies a boundary that can be borrowed. China's Advertising Law requires that advertisements not contain false or misleading content, and that the relevant information be accurate, clear, and plain. A 2022 judicial interpretation by the Supreme People's Court on the Anti-Unfair Competition Law lists, among other things, one-sided publicity or comparison, treating an unsettled matter as settled, and using ambiguous language; if these deceive or mislead the relevant public, they may constitute misleading commercial publicity. Judgment cannot rest on a word-by-word defense. It also has to look at ordinary experience, the general attention of the relevant public, the actual situation, and the overall effect.

The U.S. Federal Trade Commission's test of deception also includes material omission: a representation, omission, or practice that is likely to mislead a reasonable consumer in the circumstances, and that is likely to affect the choice, may constitute deception (FTC, 1983). The European Union's Unfair Commercial Practices Directive likewise treats hiding material information, or providing it in an unclear or untimely way, as a risk of misleading (EU, 2005). Both sets of rules check the overall presentation, not only whether each word can stand on its own.

Experimental research shows that conversion cannot serve as ethical acceptance. Luguri and Strahilevitz had subjects face a fictional identity-protection service. Under a neutral interface the acceptance rate was 11.3%; with milder dark patterns it rose to 25.8%; with more aggressive dark patterns it rose to 41.9%. After treating everyone who dropped out midway as a refusal, the three groups were still 11.3%, 25.4%, and 37.2% (Luguri & Strahilevitz, 2021). The study measured interface manipulation in a single choice. It does not stand for long-run results of a real transaction. It is enough to show that a higher acceptance rate cannot, by itself, prove a more informed choice.

In a randomized experiment with 117 people, van Berlo and Bock removed important conditions of a discount from the initial advertisement and revealed them only after the user entered the site. Compared with a control group given full disclosure, the omission group scored lower on purchase intention, on a scale of retailer ethics, and on a scale of attitude toward the retailer (van Berlo & Bock, 2023). Across several experiments, Sah and Read found no stable distinction in how subjects rated strategic and non-strategic omissions, and found that estimates of the hidden information were lenient (Sah & Read, 2020).

Treat People with Honesty therefore cannot remain a values line. It contains at least four behavioral standards that can be checked before publication.

First, the claim can be verified. Facts, numbers, quotations, and user comments have to be traceable to a source. A comparison has to state the object, the definition, the time, and the conditions. A speculation that cannot be verified should appear as a speculation.

Second, do not conceal material information that would affect the decision. Information that would be enough to change the choice of a similarly situated party is material. It cannot be omitted because space is short.

Third, the promise can be kept. The organization should hold verifiable grounds that it can deliver under the conditions stated, and should disclose uncertainty, exceptions, and the stop condition at the same time. A guarantee beyond its capacity should not be published, even if it has not yet been falsified.

Fourth, preserve the other party's capacity for informed choice. Alternatives must not be hidden, false urgency manufactured, hard-to-exit defaults exploited, or the cost of refusal and comparison raised on purpose. What is narrowed is the cost of understanding and comparison, not the right to choose.

These four check conduct and process. Results cannot be read backward into them. Repeat purchase, referral, and later satisfaction can test whether value was delivered. They cannot prove that the communication was honest. Switching cost, habit, missing substitutes, and information still unrevealed can all produce a repeat purchase. A recommendation that meets the four standards may still turn out poorly because of a change that could not be foreseen. Results are responsible for calibrating the plan. Behavioral standards are responsible for judging whether a line was crossed.

Public regulatory-event studies add a layer of consequence. Tipton, Bharadwaj, and Robertson examined deceptive-marketing events in the U.S. drug industry that the FDA made public, and found the regulatory exposure associated with about a 1% negative abnormal return; scaled at the sample median, the loss of shareholder wealth was about $86 million (Tipton et al., 2009). Jeong and Yoo, studying 101 FTC deceptive-advertising cases from 1987 to 2005, also found negative abnormal returns, and the effect did not disappear quickly (Jeong & Yoo, 2011). Those results apply to listed-company events made public by a regulator. They cannot show that honest expression brings the highest return in every industry, and they cannot be used to explain Zhong Xue Gao's later operating situation. They only show that the cost, once a crossed line is revealed, may run well beyond the fine itself.

7. Xiaomi: make the difference clear, and make the boundary visible #

Xiaomi used three moves in the launch: it chose information that can be recalculated, it chose a comparison object of the same size, and it put the numbers in the same frame.

The published spec for the 13.3-inch Xiaomi Notebook Air was 14.8 millimeters and 1.28 kilograms; the published height range of a 13.3-inch MacBook Air was 3 to 17 millimeters, and the weight was 1.35 kilograms. When the comparison condition is stated as maximum height, the two published specs can be recalculated as a difference of about 13% and about 5%. That move lowered the cost of comparing the numbers.

The coin line shows the boundary of the same practice. The picture used the diameter of a coin stood on edge; "thinner than a coin" did not name the dimension of comparison, and so it is ambiguous. The public material on hand is not enough to write that expression as illegal publicity, and there is no long-run data to prove that it harmed or raised brand trust. A careful expression has to be literally true, and the overall presentation also has to stand a check on a shared definition.

The case only proves that a real difference that can be recalculated can, through a shared reference, lower the cost of understanding. It does not prove that the copy produced sales or long-run consensus, and it does not prove that every comparison is beyond dispute.

8. Zhong Xue Gao: the penalty record shows the conduct; it cannot prove why the brand declined #

Zhong Xue Gao's penalty record and later operating results are two kinds of evidence. They cannot be joined straight into a cause.

Transcriptions of two administrative-penalty decisions by China.com.cn and 21st Century Business Herald, together with Zhong Xue Gao's public apology in 2021, corroborate two penalties. On April 3, 2019, the Huangpu District Market Supervision Administration of Shanghai issued penalty 沪监管黄处字（2019）第012018001959号, a fine of 6,000 yuan. The reports record that the case involved claims about ingredients and standards for products including "Niang Hongti," "Old-Tree Northern Matcha," and "Irish aged cheese," and that the regulator found false advertising. On August 8, 2019, the Jiading District Market Supervision Administration of Shanghai issued penalty 沪市监嘉处字（2019）第142019001068号, a fine of 3,000 yuan. A light-milk ice cream advertised "not a drop of water added" and "pure milk fragrance," while the ingredient list included drinking water, was found to be misleading false advertising (21st Century Business Herald, 2021; China.com.cn, 2022).

The two fines together are 9,000 yuan. The boundary of conduct does not depend on the size of the fine. It depends on whether the claim can be supported by the company's materials. Ingredient and standard claims that the materials cannot support crossed the line "the claim can be verified," and they also broke the information conditions informed comparison requires.

The public evidence is not enough to attribute later price disputes, media storms, or operating changes to these two penalties. Pricing, product, channel, competition, expansion, financing, and cash flow may all have taken part. The material on hand lacks a time series of consumer trust and a controlled study, and so a single cause cannot be identified. A single high-priced limited SKU, the origin of a category word, and later operating news also cannot fill that causal chain.

This pairing is not a causal experiment in operating results. Xiaomi and Zhong Xue Gao are one company sample each. Industry, stage, and constraints all differ. The pair can identify the behavioral boundary among a true comparison, a misleading overall presentation, and a false claim. It cannot estimate an average effect. Studies of manipulation, omission, and regulation supply, separately, a general mechanism and a consequence that can be quantified.

9. From market understanding to organizational context #

The end of the Interference Method is not to make everyone agree. It is to establish the conditions informed judgment requires. The organization first obtains a record of existing consensus, chooses a scene that can be checked, and then lets the claim go through presentation, restatement, trial, checking against results, and revision. The whole process is constrained by four standards: verifiable, no concealment of material information, promises that can be kept, and informed choice preserved.

The method will affect how a product is presented. It does not decide from which layer of consensus a product line should start. Product design should also cut understanding cost the user does not need. Mature demand, the radius of innovation, and reverse iteration of consensus wait for Chapter 12.

Regenic.ai can only stand as a public experiment still under way. Public chapters, versions, case sources, and a revision record supply a trail that can be inspected. They cannot prove that these words have entered the industry's daily language, and they cannot prove that any firm has completed a retestable rebuild because of them. Until outside adoption, failure records, and before-and-after results appear, it is a bench, not a success case.

Once an outside party can restate and check a judgment, sales, product, delivery, and managers still need to use the same set of facts. Turning one external expression into shared context that can be reused, corrected, and maintained inside the organization is the work of Chapter 9's consensus cycle.

What to Do Monday Morning (principal-leader view) #

Pick a plan you are about to recommend to a real customer, candidate, or partner. Do not write the copy first. Complete five checks in order.

1. Record the consensus already in place. Write the names the user actually uses for the problem, the alternatives, and the comparison standards, and which facts have led to refusal, delay, or a switch. Mark each item as coming from an interview, a behavioral record, or a hypothesis. A hypothesis cannot pose as a user fact.
2. Write one informed recommendation. In one sentence, say whom the plan fits and what problem it solves. In a second sentence, say whom it does not fit, and under what condition you would advise the other party to choose another plan.
3. Choose a scene that can be checked. Decide who will explain, when they will explain, and whether a demo, raw data, a customer site, or a technical annex will support it. For third-party material, record method, interest, and range of use in the same way.
4. Run the four checks of Treat People with Honesty. Verify the source of the claim, disclosure of material information, the conditions on which the promise can be kept, and the channels for comparison, refusal, and exit. If any item fails, change the plan or the information first. Do not cover the gap with wording.
5. Record the results separately. Record how many people stopped and processed, how many restated accurately, how many took a next step on the evidence, how much choice or commitment changed after contact with the information, and whether a group already reuses the same name and standard. Do not let clicks or closed deals stand in for all five.

Narrowing is used only to lower the cost of understanding and comparison. It may not delete material information, feasible alternatives, or the capacity to refuse or leave. The last judgment always belongs to the other party.

Chapter Acceptance Self-Check (against chapter acceptance standards) #

1. Claim restatable in one sentence ✓: 10/10/80 is resource allocation, not a statistical law; without concealing material facts, help the other party see and understand the plan that fits them; do not decide for them.
2. Nine-section spine and whiteboard figure ✓: Xiaomi opening and five result splits → 10/10/80 → existing consensus → scene → form new consensus / Crossing the Chasm → four ethical standards and quantitative support → Xiaomi → Zhong Xue Gao → Chapter 9 / 12 interface. Figure is three resource steps, five separate checks, and four rails; funnel and credibility ladder are out.
3. Evidence caliber ✓: Xiaomi and Apple from published specs, ~13% and ~5% recalculated; coin line from same-day launch reports, not written as illegal publicity or as a cause of long-run trust. Zhong Xue Gao from China.com.cn, 21st Century Business Herald, and the 2021 apology, two document numbers and 6,000 / 3,000 yuan fines; later operating distress not joined as a single cause. Moore from 1991 and the 2014 author's note; Rogers 1962 for the five groups. Simon 1971 pp. 40–41 for attention allocation only. Luguri & Strahilevitz 2021; van Berlo & Bock 2023; Sah & Read 2020; Tipton et al. 2009; Jeong & Yoo 2011 as general mechanism and quantified consequence, not as an explanation of Zhong Xue Gao.
4. Case contract ✓: Xiaomi for a difference that can be recalculated becoming visible, coin line as a boundary warning; Zhong Xue Gao for conduct reported in the penalties, not brand decline; Crossing the Chasm for naming as shared language, not invention of the five groups or a monopoly of industry language. Noodle shops, Zhang Xue Motor, Jobs four-cell, author fundraising, and "they see only the solution you want" are out.
5. Ethical boundary ✓: Treat People with Honesty as four pre-publication checks; conversion, repeat purchase, and referral cannot prove honesty; manipulation, dark patterns, and taking away choice are not part of the method.
6. Chapter boundary ✓: internal shared context to Chapter 9; product path to Chapter 12; Regenic.ai as an unverified public experiment.
7. Fluency ✓: rewritten in English voice from the Chinese authority; current prose-standard.