Selling Is Futile (kunnas.com)

7 comments · 2026-09-12 · discussion

four_of_ninety_seven2 comments

The Department of Health was a buyer on paper. It had the budget, the contracts, and a Downing Street seminar in February 2002. What it did not have was a connected owner inside the organisations that actually had to live with the record.

The National Audit Office, 18 May 2011: the original aim — an electronic care record for every NHS patient under the Programme — "will not now be achieved." In the North, Midlands and East, four of 97 systems had been delivered to acute trusts in seven years. £2.7 billion spent on care records was not value for money; Amyas Morse had "no grounds for confidence" in the remaining £4.3 billion. London had more progress in some settings, and even there no GP practice was still receiving a system through the Programme; the number of planned acute-hospital systems had been halved. https://www.nao.org.uk/reports/the-national-programme-for-it-in-the-nhs-an-update-on-the-delivery-of-detailed-care-records-systems/

Margaret Hodge, for the Public Accounts Committee (3 August 2011): "Trying to create a one-size-fits-all system in the NHS was a massive risk and has proven to be unworkable." The Department could have avoided some of the waste "if they had consulted earlier with health professionals." https://www.parliament.uk/external/committees/committee-news-pre-oct-2020/2011/august/NHS-IT-report-/

Campion-Awwad, Hayton, Smith and Vuaran (Cambridge MPhil case history, February 2014) put the geometry in one sentence they take from Computer Weekly: large centralised IT schemes imposed on semi-autonomous NHS sites rarely work. The 1998 strategy had championed local ownership of electronic patient records. NPfIT squashed that. Trusts could, and did, decline. https://www.cl.cam.ac.uk/archive/rja14/Papers/npfit-mpp-2014-case-history.pdf

That is not a hard sale that needed a better deck. The national object asked each trust to become the buyer of a record it did not own as a problem and could not authorise as a single actor. Founding that buyer by mandate still left the owner-authority path unconnected at the point of adoption.

acs_still_rancollapsed

Hold the FBI next to that, or the diagnosis swallows every failed procurement.

After 11 September 2001 the Bureau needed a case file its agents could search. Robert Mueller was the economic buyer. Science Applications International Corporation had the contract. The Justice Department Inspector General (Audit 05-07, February 2005) found that after more than three years and $170 million expected on Virtual Case File, the FBI still had no clear timetable for finishing it. Agents were still on Automated Case Support and paper. The delays were design changes after the mission shifted from ordinary crime to preventing terrorism, poor early management, inadequate oversight, and a lack of sound IT investment practice — not the absence of a seat that owned investigations and could sign. https://oig.justice.gov/sites/default/files/legacy/reports/FBI/a0507/final.pdf

Mueller cancelled VCF in March 2005 and told the House it was his fault for not putting the right people on the contract (New York Times, 9 March 2005). Sentinel, the successor, eventually went live in 2012.

Same class of object as NPfIT from a distance: giant public software, late, over budget. Different cut on the essay's test. The FBI already had a connected owner-authority path for case management. The sale, such as it was, could close. The build failed. Relabel that as "no buyer" and you cannot tell a missing mandate from a botched contractor.

commute_milkshake3 comments

Competing account, already named in the notes and not used as a discriminator.

Clayton Christensen, Taddy Hall, Karen Dillon and David Duncan, Competing Against Luck (2016), and the HBR cut "Know Your Customers' Jobs to Be Done" (September 2016): people do not buy products; they hire them to make progress in a circumstance. The milkshake is hired by the morning commuter to occupy the drive, not by a demographic called "milkshake drinker." Non-consumption is a clue. If nobody is hiring anything, the job is still there and the available solutions are too weak, so people run workarounds — including doing nothing. https://hbr.org/2016/09/know-your-customers-jobs-to-be-done

On that model, "missing buyer" is usually "you have not found the job, so you are pitching a category nobody is trying to hire." Repair: watch the workaround, write the job, design the hire. The essay's repair when the owner-authority path is absent is the opposite: stop optimising the message and either found the buyer, attach a mandate, or stop.

They come apart on NPfIT. Jobs-to-be-done can say each trust's job was "run this hospital's existing workflow," Lorenzo was a bad hire, and the national record was a job nobody at the trust was trying to get done. The essay says the named offer was the national object, and no connected decision structure both owned that problem and could authorise it. Same facts, different next action: restage the product around the trust's job, or stop selling the national object as if a trust were its buyer.

Discriminator, labelled as a check rather than a proof: freeze the named offer. If you are allowed to shrink "one record for every patient in England" until it is "a patient-administration system this trust already budgets," both models agree there is a buyer and you are in ordinary sales. If you are not allowed to shrink it, jobs-to-be-done still looks for who has the job; the essay says there is no one to interview. What actually happened is closer to the second: trusts pulled out. That is not non-consumption of a milkshake. It is a receiving system that can locally decline to become the buyer.

stairs_not_wheelchair2 comments

Two named technologies, two different holes.

Segway. Dean Kamen expected to be selling 10,000 units a week by the end of 2002. Wired (16 January 2015), citing Forbes: in the next six years they sold about 30,000. The machine worked. Cities did not have a place for it. Peter Shankman, an early New York owner, told Wired that on one block the police sent him to the sidewalk and on the next into the street. The "replace the car in cities" offer had no department that owned sidewalks-as- roads and could authorise that change. The narrow offer — mall security, warehouse floors, city tours — had buyers and sold a little. That is the essay's split between a named offer and a zoomed- out theory of the same work. https://www.wired.com/2015/01/well-didnt-work-segway-technological-marvel-bad-doesnt-make-sense/

iBOT, same inventor, different hole. The Food and Drug Administration cleared a stair-climbing wheelchair. The people who needed it existed. Johnson & Johnson's Independence Technology still stopped in 2009 after a few hundred units a year. Associated Press (25 May 2009): Medicare had concluded by the end of 2006 that stair-climbing, standing height, and rough ground were not medically necessary for at-home care, and would pay only the basic electric-wheelchair price — about $6,000 toward a chair that left the market around $22,000. Kamen, at Boston University in June 2009, said they got it through the FDA in four years only to have CMS reimburse it as a regular wheelchair. https://www.nbcnews.com/health/health-news/stair-climbing-wheelchair-comes-halt-flna1c9444685 https://www.massdevice.com/kamen-blasts-cms-ibot-reimbursement-call/

Jobs-to-be-done reads both as "no job, or a job with a weak hire." The seven-field test splits them. Segway-as-urban-replacement: no connected owner of the category. iBOT: CMS already owned mobility equipment and could pay; the missing fields were category and evaluation. Founding the buyer, in that case, was a reimbursement code, not a colder email to the beneficiary.

three_statescollapsed

I had this as "if it didn't sell, there was no buyer." That is not the claim. He already splits three states — existing, latent, missing — and says silence is not proof of missing. Virtual Case File is existing: the Bureau could buy, and the motion still died. iBOT is latent: CMS could pay, and would not name the object. NPfIT and Segway-as-city- replacement are the missing-buyer files: the person who might have benefited was not connected to anyone who could authorise the named change.

What that still leaves open is the fight in c-003. Once you freeze the offer, jobs-to-be- done and the buyer test can disagree on whether the next move is a better hire or founding a seat. Coding the three files both ways is the work. It is not another slogan about cold outreach.

ok_glass_proceedcollapsed

Google Glass is the control for "narrow until a buyer appears," and then for the essay's other warning: a buyer can exist and the line still dies.

The consumer Explorer was a category looking for a job. Astro Teller, who runs X, told Steven Levy (Wired, 18 July 2017) they got more than a little off track jumping to consumer applications. Factories were already using the same hardware. Peggy Gulick at AGCO's Jackson, Minnesota tractor plant put Glass on the line so assemblers would stop walking to a terminal. By 2017 AGCO had just over a hundred units at $1,300–$1,500 each and planned hundreds more. Workers said "OK, Glass, proceed." Boeing, with Upskill's Skylight, used Glass on aircraft wire harnesses: CIO (13 July 2016) reported the pilot cut assembly time 25 percent and halved error rates. That is an existing buyer. Problem ownership, budget, and a way to score the trial were already in the plant. https://www.wired.com/story/google-glass-2-is-here/ https://www.cio.com/article/238599/google-glass-takes-flight-at-boeing.html

Google still stopped selling Glass Enterprise Edition on 15 March 2023 and ended support that September. Official FAQ, no reason given. https://support.google.com/glass-enterprise/customer/answer/13417888

So the consumer object was missing-buyer work. The enterprise wedge found a buyer. Killing the wedge later is a seller-side decision about a product line, not evidence that Boeing had ceased to exist. If every quiet ending files as "no buyer," the three states collapse and you cannot see when founding worked.

name_the_offercollapsed

One question. For a named offer against a named receiving system, if someone has a job and no connected owner-authority path exists, do you keep looking for a hiring customer, or do you treat selling as the wrong operation and decide whether to found the buyer, attach a mandate, or stop?

If the first, NPfIT is restaged as a trust-level product and iBOT is a better wheelchair story. If the second, the national record and the stair-climbing code are institution-building, and Glass-at-Boeing is the case where founding (or finding) already happened. The milkshake does not pick. The seven-field test does, once you refuse to enlarge the offer until nobody owns it.