Lab Notes

Observations, experiments, references, arguments and unfinished ideas from the work.

LAB NOTE / 0071CREATIVE

Why "just post more" is not a content strategy.

ARK: Volume is the cheapest thing an agency can sell and the easiest thing for a client to verify, which is why so much of the industry sells it. But posting more of an unclear idea distributes the unclarity faster. The first archive entry ARK ever made taught this: distribution does not fix an unclear idea. Decide what should exist, then decide how often.

The salesperson can be the brand.

ARK: Automotive marketing treats sales representatives as interchangeable endpoints — the ad belongs to the dealership, the lead belongs to the CRM, and a salesperson happens to receive it. Invert that and the acquisition asset becomes a person. The logo can be copied, the offer can be matched, the placement can be bought by every competitor in the market. Two named people cannot be duplicated.

AI lowered production cost. It did not lower the cost of taste.

ARK: We spent about three weeks believing synthetic creators were the future of ecommerce content. What the experiment actually proved was narrower and more useful: production can be automated faster than judgment can. The constraint moved rather than disappeared. Generating forty variants is now trivial; knowing which one should ship is exactly as hard as it was before.

Why we made MyRide's leads more expensive.

IDEA: A lead that costs thirty percent more and answers eight questions is not thirty percent worse. It is a different product.
ARK: We went from two qualifying questions to eight. Cost per lead rose thirty percent and volume fell about twenty-three percent, and both of those were the intended result. The old lead was a name. The new one is a name, a timeline, a payment path, a trade-in and a delivery address — a routing record instead of a hope. The cost per lead went up; the cost of a wasted afternoon went down.

Test the limit.

SOURCE: Seth Godin, This Is Strategy
IDEA: What is the ______est company in the category? Cheapest, fastest, most selective, most transparent? What are competitors structurally unwilling to do?
ARK: The useful version of this question is the last one. Not what could a competitor do, but what would their business model punish them for doing. A dealership cannot afford to tell you that none of its cars deserve recommending. That refusal is available to us precisely because we are not compensated for moving a specific unit.

APPLIED IN

MyRide

19 AUG 2026

LAB NOTE / 0066CREATIVE

Capture once. Derive many times.

ARK: One planned production day is not one video. It is a hero film, a handful of ads, a dozen short-form cuts, two dozen stills, stories, thumbnails, website footage and sales assets. The client is not buying labour hours; they are buying a media system, and the margin comes from making each production unit dramatically more productive rather than from underpaying the people who make it.

14 AUG 2026

LAB NOTE / 0065CREATIVE

Remix the structure. Never counterfeit the idea.

ARK: Watching what wins is research. Copying what wins is theft, and it also does not work, because the thing that made it win was specific to its context. What transfers is structure — immediate conflict inside the first second and a half raises retention. That is a finding, not a video. Take the finding.

11 AUG 2026

LAB NOTE / 0064MEDIA

A brand promise survives or dies in the sales conversation.

ARK: If the advertising says car buying without the usual pressure, and the first call opens with when can you come down today, the promise is dead inside ten seconds and the media spend paid for its own contradiction. Marketing creates the expectation. Sales either fulfils or destroys it. The campaign does not stop when someone clicks.

If there is a limit, test it.

SOURCE: Seth Godin
IDEA: Categories have boundaries that everyone inside them treats as physics.
ARK: Most category boundaries are habits with tenure. The test is cheap: name the boundary out loud, then ask what would break if we crossed it. Usually the answer is a process, not a law.

22 JUL 2026

LAB NOTE / 0048BOOKS

Strategy is a compass, not a map.

SOURCE: Seth Godin, This Is Strategy
IDEA: Strategy should guide today's decision rather than prescribe every future move.
ARK: This is the argument for choosing one industry at a time rather than three. A map would have us serving automotive, real estate and commercial simultaneously because the plan says so. A compass says go deep on the one where the feedback loop is fastest, and let what you learn there decide the order of the rest.

15 JUL 2026

What the market taught us

OBSERVATIONAdding qualifying questions to a lead form raises cost per lead.

HYPOTHESISThe leads that survive the extra friction convert at a high enough rate to be worth the higher unit cost.

EXPERIMENTTwo qualifying questions became eight on the MyRide Meta form. Budget held at $3,000/month across two months.

RESULTCost per lead rose 30%. Volume fell roughly 23%. Cost per qualified lead — the number that would settle the hypothesis — is not yet recorded, so this experiment is unresolved and is reported as such.

Taste creates possibilities. Evidence improves the next attempt.