Pain or vision

The pain engine

Cyert and March's behavioural theory of the firm has a simple idea at its centre. Every organisation has an aspiration level, what it thinks it should be achieving. While performance stays above that line, nobody looks for anything new. When it falls below, the search begins.

Henrich Greve tested this idea in shipbuilding companies ⁵. He measured investment in research and the launch of innovations over several years. When performance was below aspiration, managers invested more and agreed to launch more new things. When performance was high, they invested less and launched less.

Companies that are doing well are the ones that search least. That is where the vendor of something new has the hardest time.

Andrew Van de Ven spent years studying the adoption of innovations in hospitals ⁶. Action began when people reached a threshold of dissatisfaction. Small, gradual improvements were not enough. It took a shock, internal or external, for anyone to move.

For sellers, this has a practical consequence. A mild pain does not move an organisation. The pain has to cross a threshold, and the vendor can help make it visible, but cannot invent it.

The slack engine

In the same study, Greve found a second path. Shipyards with spare resources also invested more in research, even without problems. This is slack search. The company experiments because it can.

Slack has a limit. Nitin Nohria and Ranjay Gulati studied 264 departments in two multinationals ⁷. The relationship between slack and innovation had the shape of an inverted U. Too little slack killed experimentation. Too much also did harm, because projects lost discipline and nobody cut them.

This is exactly the Explorer's risk. It has money for pilots and launches several. Many die for lack of criteria, not lack of merit.

The adoption figures by company size fit here. In every official source we consulted, large companies adopt first. They have more slack, more specialists and more capacity to absorb the risk of a failed experiment.

The third engine: pressure from outside

In 1983 Paul DiMaggio and Walter Powell described three forces that lead organisations to adopt the same practices ¹². The first is coercive: a customer, a regulator or a parent company requires it. The second is mimetic: in the face of uncertainty, people copy whoever seems to be succeeding. The third is normative: professions and schools spread what is considered good practice. The authors warn that none of this guarantees efficiency.

The spread of ISO 9000 certification is the best measured example. Isin Guler, Mauro Guillén and John Macpherson followed 85 countries between 1993 and 1998 ¹³. In 1993 there were about 27,000 certificates in 48 countries. By the end of 1999 there were more than 400,000 in 158. What best explained adoption in a country was trade relations with countries that already had many certificates. The state and foreign multinationals also carried weight. The country's technical know-how did not.

Companies got certified because their customers asked them for the certificate. Pressure from outside manufactured the pain.

Eric Abrahamson added the less pleasant side ¹⁴. Some innovations spread through fashion and the need to look modern, and some efficient ones are rejected. Diffusion research tends to assume that adopting is always good. Abrahamson called this a pro-innovation bias.

Status quo bias

In 1988, William Samuelson and Richard Zeckhauser published a study that gave a name to something every salesperson knows ¹⁵. They ran decision experiments in which the same choice was presented in two ways. In one, one of the options appeared as the current situation. In the other, all the options were new. When an option was the status quo, people chose it far more often.

Then they looked at real decisions. They analysed university lecturers' choices of health and pension plans. The pattern repeated. People stayed with what they had, even when the alternatives were better.

Amos Tversky and Daniel Kahneman measured the size of that weight ¹⁶. In their experiments, a loss had a median weight 2.25 times greater than an equal gain. In practice, a bet was only accepted if the possible gain was more than twice the possible loss.

The 9 times effect

John Gourville, of Harvard Business School, took these ideas into the launch of new products ¹⁷. The reasoning is this. The buyer overvalues what they already have, by about three times. Whoever developed the new thing overvalues what they created, also by about three times. Multiplied, the gap between the two can reach nine times.

Gourville cites a classic experiment by Jack Knetsch. He gave some people a mug and others a chocolate bar of similar value, and then offered a swap. Only 11% of those with the mug swapped it. Only 10% of those with the chocolate. They had only had the object for minutes.

For the customer, your solution does not compete with the competitor's. It competes with something they already have, that they know and that asks nothing of them.

The consequence for sellers is harsh. A solution 20% better looks to you like a clear win. For the buyer, it may not even reach the point where changing is worth it.

Indecision is not the same as preferring what exists

Matt Dixon and Ted McKenna analysed about 2.5 million recorded sales calls ³. They concluded that 40 to 60% of qualified deals are lost because the customer does not decide. They separated two different causes ¹⁸.

In 44% of cases the customer preferred to stay as they were. They were not convinced that changing was worth it. In 56% the customer wanted to change, but was afraid of choosing badly. These are two problems with opposite remedies.

The customer who prefers the status quo needs to be shown the cost of staying. With the indecisive customer, talking more about that cost makes things worse. According to the authors, using the fear of what is lost by not buying lowered the success rate in 84% of cases involving indecision. What works with the indecisive customer is reducing the risk of the choice: recommend one option, limit the alternatives, offer guarantees.

In Analysis No. 1 we saw the size of this problem. Faced with a very indecisive customer, the average salesperson wins 6% of the time.

"There was no need"

Official statistics say the same thing in other words. The European innovation survey asks companies why they did not innovate, or did not innovate more ². We calculated the percentages from Eurostat's absolute figures.

Among companies that did not innovate, 60.5% said there was no need. Only 12.5% pointed to a lack of resources. Even among those that did innovate, lack of need (38.9%) comes ahead of lack of resources (27.3%).

Eurostat also identifies a group of about 20% of EU companies with the potential to innovate, but held back by a lack of resources or by reasons of strategy, risk and return ²⁰.

In Brazil, IBGE asked industrial companies with 100 or more people why they do not adopt advanced digital technologies ¹⁹. Cost was the most cited reason (74.3%), followed by a lack of qualified staff (60.6%). Even so, 54% said they see no need.

In Europe, the Settled is the most common profile among those who do not innovate. In industrial Brazil, the Urgent without means and the Settled weigh more.

Rebuy or new purchase

In 1967, Patrick Robinson, Charles Faris and Yoram Wind proposed a way of classifying industrial purchases that is still taught today ²¹. There are three situations.

In a straight rebuy, the customer reorders what it already bought, from the supplier it already had. It is routine. In a modified rebuy, something changes: the specification, the price, the supplier. In a new task purchase, the company buys something for the first time. More people get involved, more information is sought and it takes longer.

A solution can be new to the market and reach a customer as a modified rebuy. A better version of something they already use, sold by the supplier they already know, comes in through a much wider door. The same solution, sold to someone who has never bought anything like it, is a new purchase, with all the weight that carries.

Later research found mixed support for this model. In practice, the boundaries between the three types are not sharp. The central idea held up: the newer the purchase is for the buyer, the slower and more collective the decision.

Profitability pulls, investment brakes

In 1961 Edwin Mansfield studied how quickly US companies adopted twelve major innovations in four sectors: coal, steel, brewing and railways ²². Almost all were heavy equipment that lowered costs.

The time until half the companies adopted ranged from less than a year to 15 years, with an average of 7.8. Two factors explained almost all the difference: how much the innovation paid and how much it cost to adopt relative to the size of the company.

Zvi Griliches found the same pattern with hybrid maize in the US ²³. Farmers are businesses that buy from suppliers, so the case applies. Adoption began first where it was most profitable for seed companies to enter. The speed and the point where it stopped depended on how much the change paid each farmer.

A solution that pays a lot and costs little to try moves fast. One that pays little and requires a lot of investment moves slowly, however good it is.

At country level, Diego Comin and Bart Hobijn measured the adoption of 15 technologies in 166 countries over about two centuries ²⁴. On average, a country adopted a technology about 45 years after it was invented. More recent technologies spread faster. Differences in adoption explain at least a quarter of the differences in income between countries.

The buying centre

In 1972, Frederick Webster and Yoram Wind named the roles found in almost every industrial purchase ²⁵. The user will work with the solution. The influencer helps define what gets bought, often a technical person. The buyer negotiates. The decider says yes. The gatekeeper controls the information that reaches the group, and can be an assistant, a technical person or the purchasing department.

The model is more than fifty years old, but the group has grown. Gartner today speaks of 5 to 11 people in large companies ²⁶. Forrester counts 13 internal people plus 9 outside influencers ²⁸. In purchases of generative AI solutions, the group is twice the size: 14 people against 7. The purchasing department takes part in the decision in 53% of cycles.

Each extra person lowers the probability

CEB, now part of Gartner, surveyed about 3,000 buyers for the book The Challenger Customer ²⁷. It measured the probability of a purchase happening according to the number of people involved. With one person, 81%. With two, 55%. With five or more, the probability drops to around 30%.

The explanation is intuitive. Each person brings their own priorities and fears. It only takes one to disagree for the decision to stop. This is why the Transformer, with plenty of pain and plenty of means, still fails to complete so many purchases.

Gartner shows the other side ³³. 77% of buyers say their last purchase was very complex or difficult. About two thirds of the journey is spent gathering, processing and reconciling information. When the supplier helps the group decide, the buyer is 2.8 times more likely to find the purchase easy and 3 times more likely to buy the bigger option, with less regret.

Personal value weighs more than people think

In 2013, CEB and Google, with the analytics company Motista, surveyed more than 3,000 buyers about 36 brands ²⁹. Only 14% saw enough difference between suppliers to pay more. Personal value, such as career progression, confidence and pride in the decision, had about twice the impact of business value.

Bain surveyed more than 2,300 decision-makers to build its B2B "elements of value" ³¹. Buyers said cost reduction was the most important. When what actually made them promoters of a supplier was measured, cost reduction came 27th. "Hope", an emotional element, came 4th.

In 2025 LinkedIn surveyed 750 senior buyers about five emotional dimensions of buying ³⁰. The first was the ability to defend the decision internally, even if it went wrong.

The buyer does not just buy the best solution. They buy the one they can defend if it goes wrong.

Vision can also be created

Explorers do not appear on their own. Part of the demand for new things is created by those who publish ideas. Edelman and LinkedIn surveyed 3,500 decision-makers in 2024 ³². 75% said a piece of thought leadership content led them to research something they were not considering. Of these, 23% ended up buying from the author. 86% would invite those companies to a tender. Nearly 70% of top executives questioned a current supplier because of a piece of content.

These are figures reported by the respondents themselves and published by two companies that sell exactly this service. The direction, however, is consistent with the rest. A customer's vision can be fed from outside.

The five attributes, and the three that count

Everett Rogers summed up decades of research in five attributes that make an innovation easier to adopt ⁹. They are relative advantage over what already exists, compatibility with what the company already does, complexity, the chance to try it first and the visibility of results.

In 1982 Louis Tornatzky and Katherine Klein ran a meta-analysis of 75 studies to see which ones actually counted ³⁴. Three had a consistent relationship with adoption: compatibility, relative advantage and complexity, the last with a negative sign.

More recent meta-analyses, on information technologies, add the company side. One, with 35 studies, found the significant factors to be relative advantage, compatibility, top management support, organisational readiness and competitive pressure ³⁶. Another concluded that organisational readiness was the strongest factor ³⁷. Having an internal champion for the product was not significant.

Ruud Frambach and Niels Schillewaert brought it all together in a two-level model ³⁵. At the first level, the company decides. What counts is the company's characteristics, its perception of the innovation, market pressure, its network of contacts and the supplier's effort to reduce risk. At the second, people accept or refuse to use what was bought. We come back to that second level in the next chapter.

Seeing is believing

The strongest evidence comes from recent field experiments, with random assignment of who receives what.

In Bangladesh, researchers worked with 505 leather goods factories in Dhaka ³⁸. Some received a video about efficient electric motors. Others received a motor installed on a machine, with an electricity meter. The rest received nothing. The motor used about 71% less electricity.

The most interesting result was the effect on neighbours. A factory with a neighbour, less than 500 metres away, that had received the installed motor was 19.3 percentage points more likely to adopt it. The information video added 9.9 points. Seeing the motor working next door was worth twice the explanation.

In Ghana, Morgan Hardy and Jamie McCasland worked with 417 small garment makers ³⁹. Some received training in a new weaving technique. Others received real orders from customers who asked for that technique. Training alone barely spread: only 2.1% of the trained firms taught it to others. When there was training and an order, 54.2% taught it.

Information is not enough. The technique spread when there were customers asking for it.

In China, Jing Cai and Adam Szeidl randomly assigned 2,820 young companies ⁴⁰. The owners of some met every month in small groups for a year. The others did not. The companies in the groups increased revenue by 8.1% and improved their management practices. The effects lasted at least a year after the meetings ended.

References and pilots

V. Kumar, Andrew Petersen and Robert Leone measured the value of a customer as a reference for others, at a financial services company and at a telecoms company ⁴¹. The best reference customers were not the ones who bought the most. What counted was the length of the relationship, the size of the customer, the format of the reference and the similarity to whoever was going to hear it.

Experimentation has become the norm. According to Forrester, more than 60% of buyers use some form of trial before buying, and 78% for purchases of 10 million dollars or more ²⁸. Only a little over a third plan to convert the trial into a paid contract.

The tool that stayed in the drawer

Sialkot, in Pakistan, is a football manufacturing hub. A team of economists led by David Atkin and Eric Verhoogen designed a new cutting die for these factories ¹. It made better use of synthetic leather, the main raw material, and lowered the material cost per piece by 6.8%.

In May 2012 they offered it free of charge to 35 factories chosen at random. Fifteen months later, only five were using it. The tool was free and profitable for the owner.

The researchers went to find out why. Cutters are paid by the piece. With the new die, in the first few weeks, they cut more slowly and earned less. The owner gained, the worker lost, and the worker had ways to block the change, for example by exaggerating the tool's difficulties.

In 2013 they ran a second experiment. They offered one cutter and one printer in each factory about a month's wages if they showed the owner they had mastered the new die. Adoption rose by between 27 and 32 percentage points. In the factories that accepted the offer, it rose by 48.

The solution was right and the purchase was made. What was missing was dealing with what the people using it stood to lose.

Profitable practices nobody adopts

Nicholas Bloom and colleagues ran a similar experiment with management practices ⁴². They chose family-owned textile factories in the state of Maharashtra and offered them free management consulting. Some factories received one month of diagnosis and four of implementation of 38 management practices. Others received only the diagnosis.

The supported factories increased productivity by 17% in the first year, with better quality, more efficiency and less stock. Within three years they had opened more factories.

The interesting question is why they had not done it before. According to the working paper version of the study, about 15% of the practices were unknown to the owners. A little over 45% were known, but the owners thought they were not worth it. After seeing the results, the owners' limited time and procrastination still delayed adoption. Lack of money hardly counted.

This is the portrait of a Settled company. The factories did not feel they needed it and did not know what they were missing.

What makes people use it

In 1989 Fred Davis proposed the technology acceptance model, still the most cited today ⁴³. With 152 users and four applications, he measured two perceptions. Perceived usefulness had a correlation of 0.63 with current use. Ease of use had 0.45, and acted mainly through usefulness. People use what they find useful. Ease helps when there is usefulness.

In 2003 Viswanath Venkatesh and colleagues compared eight competing models in four organisations over six months ⁴⁴. Each model explained between 17% and 53% of the intention to use. The unified model they proposed explained 69%, and 70% in a second sample. The factors are performance expectancy, effort expectancy, the influence of colleagues and managers, and facilitating conditions.

Regret

When adoption fails, the customer regrets it, and that weighs on the next purchase. In 2023 Gartner surveyed 1,503 people involved in renewals of technology services ⁴⁵. 60% said they regret almost every purchase they make. In 2024 Forrester found 81% of buyers dissatisfied with the supplier they chose ⁴.

For sellers, the work does not end at the signature. A customer who regrets the purchase does not renew and does not serve as a reference for the next one.

Europe

The European innovation survey shows that 51.4% of EU companies with 10 or more people had innovation activity between 2020 and 2022 ². Germany came in at 63.4%. Portugal came in at 47.4%. By INE's measure, which counts only companies that actually introduced an innovation, it was 44.7% ¹¹. In the following survey, for 2022–2024, Portugal fell to 42.5%, the lowest figure since 2018 ⁵⁸.

In artificial intelligence, the EU went from 8% of companies using it in 2023 to 20% in 2025 ¹⁰. Denmark leads with 42%. Germany is at 26%. Portugal rose from 7.9% to 11.5%, below the European pace.

In cloud the gap is growing ⁴⁶. In 2021, Portugal was about 7.6 points below the European average. In 2025 it is 14 points below: 38.7% against 52.7%.

In Portugal, 41.5% of companies with 10 to 49 people innovated, against 79.1% of those with 250 or more ¹¹. In a country of small companies, there are proportionally more Settled companies and more Urgent ones without means.

United States

Every two weeks the Census Bureau asks companies whether they used AI in the previous two weeks ⁴⁷. In May 2026, the answer was 19.8%. Among companies with 250 or more workers, 37%. Among those with up to four, less than 20%. The information (39.7%) and finance and insurance (33.9%) sectors lead.

An earlier Census study, covering about 850,000 companies, shows the same concentration ⁴⁸. In 2018, less than 6% of companies used any of five AI technologies. Weighted by employment, the figure rose to just over 18%, because those using it were the large ones.

Asia

Industrial automation is where Asia stands out most ⁴⁹. South Korea has 1,220 robots per 10,000 manufacturing workers, Singapore 818, Germany 449, Japan 446 and the US 307. In 2024 China installed more than half of the world's industrial robots ⁵⁰.

On AI in companies, the Asian data are less comparable. In Korea, a report based on the national business survey indicates 9.2% use in 2024, with 18% among large companies and 6.2% among small ones ⁵². In Singapore, 62.5% of non-SME companies use AI, against 14.5% of SMEs ⁵³. In Japan, 18.4% of companies with 100 or more workers have introduced IoT or AI systems ⁵⁴. For China we did not find an official rate of companies using AI.

Brazil

IBGE measures innovation in industrial companies with 100 or more people ⁵¹. In 2023, 64.6% of companies innovated, the second fall in a row, after 70.5% in 2021. Among those that innovated, the most cited obstacles were economic instability (44.2%), limited internal resources (42.1%) and competition (41.4%).

In digital technologies, change was fast ¹⁹. The share of industry using AI went from 16.9% in 2022 to 41.9% in 2024. 77.2% use cloud and 30.5% robotics. Even among those adopting, 78.6% point to high cost as a barrier.

The picture of Brazilian industry has many Urgent companies: businesses that want to change and feel held back by cost and a lack of staff.