Walmart Jetblack: the text-message shopping concierge that did not close the math
In 2017, Walmart launched Jetblack, a text-message shopping service for premium customers. In February 2020, it shut down. The case became a footnote. The lesson inside it is still current.
Jetblack promised what many promise today on WhatsApp: send a message with what you need, the service handles it. The difference is that Walmart tested it at the shelf and measured the math.
What was announced
Jetblack was a shopping concierge over SMS, with a monthly subscription, aimed at high-income families in New York. The messaging suggested an almost magical assistant: ask, it arrives.
The press (The Information, CNBC, Business Insider) covered the launch as Walmart’s bet on conversational commerce. The number that mattered, at the time, was delighted customers.
What changed in the public record
In 2020, Walmart ended the service. Reporting revealed what sat behind the conversation: a labor-intensive operation of human agents handling each order, with reported losses in the thousands of dollars per subscriber per year.
The conversation looked automated. Behind it, expensive people doing the work the interface hid. Jetblack did not fail to delight the customer. It failed on the math behind the delight.
The decision that explains the gap
It was one. The service was scaled on experience, measured in satisfaction, with no cost-per-order criterion that closed before growing.
The conversation layer was cheap and charming. The cost lived in the human resolving the hard case, and the hard case was most of them. It is the same pattern as Amazon Go: what shows in the demo is not what sustains the operation every day.
The criterion that would have caught it earlier
One number, defined before scaling: the net cost per fulfilled order, compared against revenue per subscriber, with a review trigger if margin stayed negative for two months. That indicator would have shown in months what took three years to close.
Take the service or agent project on your roadmap. Can you say, in one sentence, the cost per resolved interaction when the case is hard? If not, you are in Jetblack’s position before 2020.
This piece rests only on the public record. We have no access to the internal operation; the reading is built on what was disclosed.
Defining that criterion alone is hard, because it crosses experience, operations, and finance, and each area defends the number it already measures. Send me the three indicators your operation already tracks on automated service. In one hour I will send back which would have predicted Jetblack’s end and which are noise. If we see a project shape, the Diagnóstico is two weeks and ends with a one-page document: three indicators, three target ranges, three review triggers.