The next stage was less dramatic than a tasting table and more dangerous to get wrong.
A cooperative lawyer placed a yellow pad between Baraka and the documents.
“Before anyone says fraud,” she said, “show me the signed paper that supports the sentence.”
Victor's name was not on every page. Crown Ember was not the only company named in the chain. Kifaru Ridge Reserve was registered through a marketing entity. Emberline Brand Services appeared on packaging and service invoices. Buyer purchase orders referred to premium origin language without explaining how the origin declaration had been assembled.
It would have been easy to draw a single red line through all of it and call the picture a conspiracy.
The lawyer would not allow that.
They built the map in layers.
First came contracts. Kiamwangi supplied coffee through Crown Ember. Crown Ember controlled export handling and several quality steps. Kifaru Ridge appeared as a premium brand in buyer-facing material. Emberline invoiced for origin labels, sample sleeves and field activation support.
Second came batches. The disputed Kiamwangi premium lot, the downgraded lot, the Kifaru shipment and Karindundu's returned coffee each received separate rows. Where a link was proven, they drew a solid line. Where it was only an inference, they used a dotted one.
Third came money. Kiamwangi's farmer payout was tied to a lower grade. The buyer price for Kifaru carried a premium. Crown Ember's accounts showed service fees associated with brand preparation, but the documents in front of them did not yet show how every shilling of the difference had been distributed.
“Do not invent the missing money,” the lawyer said.
Baraka nodded. “Then we do not need the missing money to prove the payment mismatch.”
Exactly.
Achieng, now working through buyer compliance rather than Crown Ember's internal system, identified which process exceptions had allowed brand identity to be changed late in the chain. Her signed statement matched the pattern, but she resisted an attempt to make her testimony carry more than it could.
“I can say commercial asked for these exceptions,” she said. “I cannot say why each request was made.”
The lawyer wrote that distinction down.
Karanja joined by video. He wanted to know whether the map was strong enough to cancel the seven-year agreement before the postponed vote resumed.
“Canceling for what reason?” the lawyer asked.
“The source problem.”
“Your members also have a governance reason. Read the control clauses again.”
The agreement did more than buy coffee. It gave the exporter long control over sample preparation, buyer introductions and washing-station support. A separate service-fee clause linked part of that support to the same vendor stream in which Emberline appeared.
That did not prove that the seven-year agreement had been drafted specifically to hide relabeling. It proved something narrower and commercially important: if signed, the agreement would concentrate exactly the stages that had become disputed—sample custody, labeling and buyer access—under the same exporter-controlled pathway.
Karanja looked tired. “So the bad contract can be bad without being criminal.”
“Most bad contracts manage that,” the lawyer said.
Muthoni, listening from the cooperative office, asked whether Kifaru Ridge was a real company or a fake farm.
The answer required care. The marketing entity was real. The brand was real. The production claim attached to it was the problem. Public business records did not support the volume implied by the single-estate story, and field verification had found no crop capable of producing it.
“So a real brand can carry an unreal story,” Muthoni said.
“Yes,” Baraka replied.
“That is a very expensive way to lie.”
The lawyer looked up. “That sentence stays out of my filing.”
They laughed for the first time that day.
By evening the claim set had become smaller and stronger. Buyer compliance could verify the mismatch between farmer-paid grade and premium buyer quality. It could verify commercial exception requests. It could verify a marketing/service route touching labels and samples. It could verify that the named estate did not support the declared production volume. It could not yet assign every act to every person.
That was enough for the next operational step.
An auction list scheduled for the following morning contained a lot from Karindundu. Its service code matched the same vendor pathway.
Lena requested that nobody contact the auction floor with a broad accusation. She wanted the exact lot identity, the exact buyer concern and the exact basis for a temporary hold.
Baraka closed the document map and looked at the second cooperative's code.
The next morning's auction included a Karindundu lot carrying the same brand-service code.