Religion vs Strategy: A JD.com Indonesia Post-Mortem

This is an operator's note, not an analyst's. I was a director at JD.ID, running fashion, beauty and sport, social commerce, JD Enabler and JD DTC, across both the marketplace and the retail side. The source for every internal figure is the notes I wrote while the business was failing,¹ and every figure with a public counterpart is footnoted to it.

JD.ID opened in November 2015 and closed on 31 March 2023.²³⁶ The official reason was a pivot to cross-border supply chain and logistics,³ and no other reason was given.⁴

Rows sourced to 1 are my working notes; no public filings exist.

What JD should have done is the easy question. The harder one: why did a company with that much money and that much data keep doing the same thing for seven years?

The answer is in the people, and it comes in layers: the goal, the people, the model, the timing, the mechanics, the choice, and what JD did next in Europe.

The goal is the criterion the whole system is judged against, so it's the first thing to check.³¹ The official answer at JD.ID was 20% of the market while staying true to the JD model.¹ That's two goals, and in Indonesia they were incompatible.

When two goals collide, the survivor is the one that matches what the people in charge like doing, and the company kept the model. The 20% market share goal was decoration.

Humans do what they like doing and then build the story that explains why it was right. The liking isn't random: people get good at what they enjoy, get promoted for what they're good at, and arrive in the executive room as the best version of one craft, and by then the craft is the person.

The people running JD, in Beijing and in Jakarta, were retailers and merchandisers, buying stock from brands, negotiating margin, setting prices, running warehouses. So in Indonesia they did retail and logistics, and the strategy deck came afterwards to explain why Indonesia needed retail and logistics.

Asking those people to pivot to a marketplace was not asking for a different strategy, it was asking them to become beginners. Seller acquisition, performance ads and voucher economics were crafts they didn't have, and the people who had them were junior, local, and working for Shopee.

Nobody senior volunteers to become junior, so the organization defended the model with arguments that sounded like strategy. Psychologists call the mechanism escalation of commitment;³² I learned it in meeting rooms.

Every line in the deck was true somewhere, mostly in China, and a true sentence from another country is a comfortable place to hide from your own numbers.

That's what I mean by religion: a doctrine that starts protecting the people who practice it instead of testing itself against the market.

Strategy needs both: a logical model, checked against the data, and when they disagree the data wins. Religion uses only the first half: it starts from what you are and explains why the reward will come, later, the way every religion promises it.

The alternative was too competitive, too cheap, not the JD way. Logic will almost always agree with you; the data almost never will.

The left column is the internal argument as I heard it, from my notes, not a publishable document.

There are two ways to reach volume in e-commerce: the platform buys and sells, or sellers transact on the platform. Neither is better, each wins in a different structure.¹

JD's own numbers show that smartphones are the exception and fashion is the rule.

The rows on 1 are internal category shares; the public record (the brand exodus, the IPO revenue mix) points the same way without exact figures.

JD won one category with retail and was a rounding error everywhere else, in China, before it expanded internationally. Indonesia is a marketplace country in every category.¹

My feeling at the time: JD didn't quite know why it had won China, and a company unsure why it won exports the wrong half of the win.

The story JD told itself was that it won through retail: own the right stock, have the best price, deliver it fast. We even ran workshops on going back to the basics of price and assortment, pure merchandising textbook, with no discussion of why we could not have the best price and the best stock in Indonesia.¹ The story was half right.

JD had been online since January 2004, four years before Tmall existed,²² and on the direct-sales side of B2C it held around half the market for years.¹⁷¹⁸ Then Tmall launched in April 2008, took 20% of B2C in its first year, and by late 2011 had over half of it.²⁰²¹

JD's edge had never been retail, it was the job it did: putting the SKUs online when nobody else could, and that is what China rewarded with volume. Retail and marketplace are not the value, they are tools, business models you pick to solve a problem, and the problem then was getting the assortment in front of people.

My reading is that JD mistook the business model for the problem. It got where it is because twenty years ago it solved the most pressing problem in the market and generated enormous value doing it; everything since, twenty years of it, is the compounding of those magic years.

An e-commerce platform has one basic job, put the assortment online so people can buy it, and whoever does it first wins. Lazada did it in Southeast Asia in 2012, running retail because sellers weren't ready, and got paid in share.¹

Sources 1, 6, 17, 22. Selected cases that fit the claim.

Same company, same model, five years late, and the job was already done. Five years earlier it works, five years later the problem the model solved no longer existed.

The JD model promises better assortment, better price and better service, and in Indonesia none of the three showed up.¹

Source 1, one operator's sample: the China column as explained internally, the Indonesia column as I watched it.

On the distributor stack, SK-II is the clean example: the same SKU for ten years, one producer, thousands of distributors, and the best price is never at the top of the chain.¹

JD cut prices in Indonesia the way it did in China and orders exploded, but they came from physical shops and traders who move between platforms on price alone, so most of JD.ID's GMV was B2B arbitrage wearing a consumer app.¹

The country team knew, and the targets made it rational: a number that consumers can't deliver becomes reachable when you call the top 500 distributors in the country and hand them a voucher to place this month's orders on JD.¹ Not one consumer in Jakarta changed a habit.

Google and Meta don't exist in China, so JD never built the performance-marketing muscle and hired marketing as it understood it: brand CMOs, budgets a fraction of Shopee's, Tokopedia's and Lazada's, and the logic covering the gap, traffic comes from assortment and price, not ads.¹

True when nobody else has the assortment, false with three or four platforms selling the identical thing at the identical price.

The honest answer to "what should JD have done" has a time stamp. In 2015 the model was a fair bet, everything above is precise only in hindsight, and two or three years on the home model was reasonable.¹ After year three there was a real case for changing course.

I ran that pivot inside my own division. Fashion, beauty and sport are the categories where the retail model is weakest, and I moved them to the marketplace side: sellers instead of stock. The categories more than doubled.

It was branch B at division scale, but a division can change its own model and can't change what the people above it like doing.

At company level the pivot depends on the goal, which takes us back to chapter 02. Laid out as a tree, the choices were these.

Sources 1, 28, 29. Chance and cost are my hindsight reading; the tree is a reconstruction, not a document that existed.

Read that way, shutting down was the right decision: if you can't replicate your model and can't compete in one you're not comfortable with, closing is rational, and Beijing got there.

What the tree shows is the timing: the same decision was available in year three, and the four extra years on branch A cost more than $120M each. The company stayed on A, never considered C seriously, and never tried B above division level.

From inside, that is what indecision looked like: it never chose, and not choosing is a choice.

When the Chinese economy turned and Beijing asked its large companies to look after distressed firms and blue-collar wages at home, the international arm became the obvious thing to cut.¹²³

It was nobody's fault, building a small successful company in a niche was not the goal. The people in those rooms were measured against a number the market couldn't give them, and no amount of cash closes that gap.

What is success for someone is failure for someone else. A profitable 2-3% player in Jakarta is a career for a founder and a humiliation for the international division of a company the size of JD.

Three weeks ago, on 30 July, JD.com announced a voluntary offer of €4.60 a share, around €2.2B, for Ceconomy, the group behind MediaMarkt and Saturn.¹³

Deal terms as of mid-August 2025; the argument below doesn't depend on the closing.

I have no opinion on whether the regulators let it through, but I do have one on what the move is. It's JD doing what JD has always done, in Beijing, in Jakarta and now in Düsseldorf: buying from brands, negotiating margin, setting prices, managing stock.

What changed is the place and the time, and that is the whole difference between Indonesia and Europe. Consumer electronics is the one category where retail wins and the one where JD dominates at home,¹⁷²⁴²⁵ and a chain of 1,000 electronics stores is a merchandising business run by merchandisers.

In the abstract: a company does the identical thing every time, and the market and the calendar decide whether it's a win or a write-off. When the thing you like doing meets a structure that wants it, that's a coincidence, and a coincidence is about as good as it gets for a company, because the liking never changes.

Concretely, for JD, it's the smarter version of the same instinct. Instead of launching the home model into a market that had already decided what it wanted, JD looked at what it is and went looking for a structure that matches.

Knowing your craft is one lesson and knowing where it belongs is the harder one, and a decade and a billion dollars later JD has learned it. If the regulators wave the deal through, I'd expect it to work, for the same reasons Indonesia didn't.

Indonesia was the tuition: seven years, more than $120M a year, a billion in GMV built largely on distributors chasing vouchers, and 2-3% of a market that had decided what it wanted before JD landed.

What I'd do with this, the next time a market entry is on the table.

• Name the problem the home model solved at home, and check whether it still exists where you're landing. If the SKUs are already online, being first is off the table, and so is the reason the model worked.
• Write one goal. A company holding two goals hasn't chosen, and the organization will quietly pick the one its people like doing.
• Put a time stamp on the bet before launch: a fair attempt of two or three years, then the tree, with every branch priced. The same shutdown costs one write-off in year three and four more years of burn in year seven.
• Expect the pivot to be blocked by identity before it's blocked by analysis. The people who must execute the new model are the ones it demotes to beginners.
• If what the company likes doing ever matches what a market wants, take the win and call it what it is, a coincidence. The liking never changes; the place and the time are the only variables you actually control.