Halo Alert at a Glance
One expedition, twenty-nine days, on a single page
The full Halo Alert record runs to two dozen files: outreach counts, complete interview transcripts, observation logs, a clustering board, three personas, an experience map, and the raw responses from three pain tests. That completeness is deliberate, because a demo you cannot check is a claim rather than evidence.
It is also discouraging if it is the first thing you see. So here is the same expedition in one page. The long version is beside it, and every row below links into it.
What this is
The concept comes from a real innovation project, most of whose artifacts are gone, so the record here was reconstructed rather than recovered. It illustrates the form of a complete expedition, not the fact of any finding about commuters. One exception: the screening matrix in Hypothesize a Good Solution is the original team’s own. The demonstration overview says more, including why generating a record like this would break the method if you did it with your own project.
Read the numbers as an honest picture of the scale of a first expedition rather than as a benchmark to hit.
The Whole Expedition
| When | Step | What they did | What came back |
|---|---|---|---|
| 19 Feb | Diverge | Generated ten candidate communities near one problem space, with orbit roles and access signals for each | A shortlist, and the discovery that the obvious group was not the only one |
| 19 Feb | Bias check | Asked which candidates were only there because they were easy to reach | Additions that would not have surfaced otherwise |
| 24 Feb | Converge | Compared candidates on coherence, neglect and access; named the emotional pull | Chose professional women commuting on foot in mid-sized cities, explicitly as a working commitment, not a final lock-in |
| 24 Feb | Access test | Three channels, real outreach: 10 LinkedIn messages, 1 coworking email, 1 forum post | 21 approached, 15 replies, 9 strong leads. Green light |
| 24 Feb – 6 Mar | Conversations | Twelve conversations across two weeks, two of them recorded and transcribed in full | Stories about route choice, check-in texts, and the cost of being late |
| 5–8 Mar | Observations | Three outings: a subway exit at 8am, the same at 8pm, a campus on a Friday night | What people did rather than said: keys in hand, street crossings, phones already out |
| 12 Mar | Clustering | Sorted the raw observations on a board | Six core themes plus edge clusters, including reassurance routines, route choice, and predictability |
| 12 Mar | Personas | Built three personas from the themes | Maya Patel, the one that carried the strongest emotional signal |
| 14 Mar | Experience map | Traced Maya’s evening commute across seven stages and four lenses | The emotional arc, and two stages where she says nothing at all |
| 16 Mar | Abduction | Generated competing explanations, checked each against the evidence, chose one | Anxiety when commuting at night, especially in unpredictable or poorly lit environments. Emotional and functional, with social overtones. Two rivals refrigerated |
| 20 Mar | Pain test | Ran three probes against three candidate pains | Dark or unlit blocks: green. Crowded exits and unexpected detours: yellow |
| 24 Mar | Brainwriting | Five people, six rounds, half an hour, including one round of deliberately bad ideas | ~90 concepts, but only two families: something you carry and something that tells someone where you are |
| 24 Mar | Recombination | Worked the pile rather than generating more | A small worn object that sends without being unlocked |
| 25 Mar | Screening | Rated candidates against pepper spray on the women’s own requirements | Two survivors: the alert ring and the alert keychain. The taser eliminated, its one advantage carried forward |
| 28 Mar | Wow factor | Both concepts to the same group, rated and probed | The ring pulled harder, because it is already in your hand |
| 28 Mar | $100 test | Eight features, a hypothetical hundred dollars | Money concentrated on two features. Nothing at all went to deterrence |
Twenty-nine days from first divergence to a validated pain, and another eight to two tested solution concepts.
Notice how the rhythm changes. Diamond 2 is measured in weeks because it waits on other people’s calendars. Diamond 3’s generation and screening took an afternoon each, because they wait on nobody.
What the Test Actually Said
Three candidate pains, three probes each, and only one converged.
| Pain | Validation | Ouch factor | Already paying | Verdict |
|---|---|---|---|---|
| Dark or unlit blocks | 8 of 10 in the past 30 days | 7.2 average, with workarounds | 40% | Green |
| Crowded exits | 6 of 10, ranked lower | 6.0, mild behaviors | 10% | Yellow |
| Unexpected detours | 6 of 10, secondary | 5.9, buffer time | 15% | Yellow |
Notice that the three probes agreed for one pain and disagreed for the other two. That disagreement is the finding, not a failure of the test.
Where It Changed Direction
Three moments where evidence overturned an expectation. These are the reason the expedition was worth running rather than reasoning through.
LinkedIn was supposed to be the cold channel. The team expected slow, formal, low-yield replies. They came back within 24 to 48 hours and were detailed. The channel they assumed was weakest carried a third of their leads.
The forum answered a question nobody asked. Replies described safety experiences outside commuting altogether: late-night walking, solo travel. The community was broader than the frame the team had brought to it.
The chosen pain was not the most obvious one. Crowding on the subway is visible, physical, and easy to describe, and it scored yellow. Dark blocks on the last stretch scored green, and the difference was behavior: people were already paying to avoid them.
What It Cost
One small team. Twenty-one people approached, fifteen of whom answered, twelve who talked at length. Three observation outings and three pain-test probes. A month of calendar time, most of it not full-time.
Two of the twelve conversations are reproduced here in full. The rest exist as notes and quotes in the clustering, which is the normal ratio: you transcribe a few completely and take the others down as they happen.
That is the honest scale of a first expedition. It is less than most people expect and more than most people do.
Reading the Full Record
Every row above links into the record that produced it. Two are worth opening even if you read nothing else: a complete conversation transcript, because it shows how unpolished a real record is, and the clustering board, because it shows scattered notes becoming themes in one place.
Diamond 3 continues past the table above, through generation and screening to two solution concepts. It stops where the original project stopped, without a smoke test, because that is where the team stopped and because a landing page deserves to be clicked rather than pictured.