Rebirth 04: Starting from Taking an Unconventional Approach

Chapter 159 Engineering Reality

Wen Jie wiped his sweat and checked if all the rings were back to normal operation. He brought up the monitoring panel and saw that the last line of red had turned green.

"The queue has been cleared." Wen Jie breathed a sigh of relief. "The interface callback in Guangdong has returned to normal, and all backlogged subscriptions have been confirmed. The current total number of subscriptions is... 770,000."

Han Fei nodded, looking at the crowd gathered in the conference room. They had been mentally prepared for this conversion rate, but the number still far exceeded their expectations. The tense faces from just a few minutes ago were now infused with new vitality by this number, as if they had been revived from the ashes.

770,000. Two million push notifications, a conversion rate of 38.5%. This was done while the system nearly crashed.

"Brother Fei," Wen Jie said, a smile involuntarily appearing in his eyes, "we made it through."

Zhou Ming let out a short laugh: "President Han, although this incident wasn't Qingniao's fault, it served as a wake-up call for us. The architecture of synchronous requests has its limits when faced with explosive growth. The asynchronous queue was usable today because an interface was provided in the MVP version. This little incident isn't necessarily a bad thing. Since Qingniao will only grow bigger in the future, now that we've identified the weakness in time, we should solidify the peak-shaving and valley-filling mechanism completely, thus avoiding problems in more important push notifications in the future."

"Yes," Han Fei said, "Zhou is right. Some pitfalls, you'll never know they're there until you fall into them once. And only once you know them can you avoid them in the future. Wen Jie, write this into next week's technical plan."

"Okay, I'll get on it as soon as possible," Wen Jie said.

"Also," Zhou Ming said, "Wen Jie reacted very quickly today. This kind of emergency response ability is more valuable than writing a hundred lines of elegant code. This was a rare and valuable practical training opportunity."

Everyone burst into laughter.

Chen Yuanshan returned to the secure conference room from the technical department. Liu Zhenhua, Fang Jing, and Zhao Guoqiang were discussing the minor setback and the astonishing conversion rate. He was excited and delighted; words couldn't describe his feelings. Liu Zhenhua and Fang Jing also wore bright smiles. However, it was clear that not everyone was satisfied with the result. He had just entered the office when he heard Zhao Guoqiang make a disheartening remark.

"Mr. Liu, we need to review this incident. If Han Fei had chosen to push the data in batches instead of dumping all two million messages at once, would the Guangdong system have crashed? Of course not. His reckless approach was tantamount to gambling with our provincial pilot program. Fortunately, we managed to stabilize it in the end; what if we hadn't?"

Silence fell. Liu Zhenhua and Fang Jing's smiles faded, replaced by serious expressions.

"Yeah, I wholeheartedly agree with what Guoqiang said," Chen Yuanshan said, realizing the irony as soon as he said it. "I think we should completely cancel the provincial pilot program, and it would be best to just shut down the number pool in Shanghai as well. That would be safer. Anyway, as long as we do nothing, we'll never make a mistake."

Zhao Guoqiang's expression changed, as if he had been slapped. Chen Yuanshan sat down in the chair between Fang Jing and Zhao Guoqiang, and smacked his lips loudly.

Liu Zhenhua frowned at Zhao Guoqiang: "Guoqiang, what kind of thinking is this? You only see some problems with the system, but you don't see the 770,000 new users and the 38.5% conversion rate. The golden window of opportunity during the public outcry is fleeting. If we push out content in batches for fear of problems, by the time public opinion has transitioned from the outcry phase to the fermentation phase, the public's emotions will have changed from impulsive to rational attention, and the conversion rate could be halved. Innovation isn't about conducting experiments in a sterile laboratory; it's about seizing territory amidst gunfire. This is an inevitable choice during the strategic window of opportunity. Are you trying to make innovation give way to processes? That's putting the cart before the horse. Besides, is the insufficient capacity of the Guangdong system a problem with Qingniao? Our own technical team only managed to stress test 400,000 users, while the actual peak reached 700,000. This shows that our weakness has been exposed by the market. Instead of examining the fragility of your own internal system, you blame your partners for moving too fast? If a patient takes medication as prescribed, but the medication has a problem, instead of blaming the pharmaceutical company, you blame the patient for not taking it?"

Chen Yuanshan laughed heartily, then realized he had laughed at the wrong time and shut his mouth.

"President Liu," Zhao Guoqiang said, "you said innovation requires seizing ground amidst gunfire, and I agree. But before seizing ground, shouldn't we understand the enemy's firing positions? Did Han Fei know the pressure ceiling of the Guangdong system? A decision made without knowing the ceiling isn't called bravery, it's called blind gambling. As someone in risk control, what's wrong with me pointing out this blind spot?"

"Where is the upper limit of the Guangdong system?" Chen Yuanshan leaned back in his chair, hands behind his head. "They don't even know themselves. The 400,000 concurrent user tests they conducted used virtual data; they didn't simulate real user conditions. The true upper limit can only be revealed under real traffic surges. Han Fei's full-scale push precisely helped us find this upper limit. This is valuable practical data," he paused, "not incriminating evidence."

"Shouldn't he be more cautious?"

Fang Jing took a breath: "Let me tell you, before the full rollout, Han Fei had his team conduct a three-minute test of 10,000 records to confirm that the interface was completely working. He made the optimal decision with limited information. You, sitting in headquarters, saw the complete logs afterward, so of course you felt you should have been more cautious. This is a classic case of hindsight bias. If every pilot unit had to pass your absolute security audit first, no team would dare to innovate. Han Fei understood the risk of system crashes better than you, but he was willing to exchange controllable chaos for an irreversible first-mover advantage. Moreover, the root cause of this incident was the Guangdong database locking, and it was Han Fei's rapid response that saved the business."

"I admit Qingniao reacted quickly. But a healthy system shouldn't rely on emergency measures as a backup. Don't they know that prevention is better than cure?"

"How can you defend against attacks without knowing the limits? And where should you defend?" Fang Jing protested. "According to your logic, all successful emergency responses are due to flawed prior decisions. Then there are no good decisions in the world," because any decision can encounter unexpected problems during execution. The criterion for judging the quality of a decision is not whether an unexpected event occurred, but whether it can be quickly recovered afterward. Qingniao's asynchronous queue went live in just over ten minutes, which proves the flexibility of their architecture design and the team's execution capabilities. Moreover, the responsibility lies with the vulnerability of China Mobile's system, not with Qingniao's push strategy. After the problem occurred, they not only did not shirk responsibility but also quickly went live with the asynchronous queue to cover for China Mobile. If you insist on calling this a flaw, I can only assume that you are opposing for the sake of opposing."

"This is pure luck!" Zhao Guoqiang's voice gradually rose, becoming shrill and piercing. "You're mistaking a lucky success for the correctness of your approach. But true strategy cannot be built on luck and emergency measures. If we don't investigate the problems in the decision-making process today, someone will use the same method to drag the entire project into the abyss tomorrow."

Fang Jing glanced at Chen Yuanshan, who nodded meaningfully.

"You keep emphasizing that we can't rely on emergency measures," Fang Jing said. "Then let me ask you, have you ever seen an internet company's system that wasn't developed through repeated emergency iterations? Tencent's QQ was constantly down in its early days, and Microsoft's Windows 95 system was notorious for its blue screen crashes when it first came out, but they patched it every time it crashed, which is why Windows XP can now run all over the world. If you demand perfection the first time, then this project should never go live. Even if we didn't stabilize the system in time this time, we can still do a technical review, quickly fix it, learn from the experience, and upgrade our technology. We should reflect on why Qingniao's servers were fine, but China Mobile's couldn't handle it. Your prudence is essentially a fig leaf for stagnation."

Zhao Guoqiang opened his mouth to speak, but Liu Zhenhua glared at him, and he closed his mouth again.

"Your objections are not constructive, only destructive," Liu Zhenhua said sternly. "Your objections never come with alternative solutions. You only say that it can't be implemented in its entirety, but you can't explain how much should be implemented, how many batches it should be divided into, how long the intervals between each batch should be, or what data should be used to determine the next batch. Objections without data support and without implementation details are not risk control, but merely emotional venting."

Zhao Guoqiang forced a smile in response.

"Responsibility and power must be commensurate," Liu Zhenhua continued. "If we push out content in batches as you suggest, short intervals between batches will still cause a surge in subscriptions, while long intervals will miss the window of opportunity, resulting in a halved conversion rate and failing to achieve five million in revenue within three months. At that point, would you be willing to sign the report and explain to the board that it was you who suggested abandoning the window of opportunity for public opinion? You wouldn't dare; you only dare to point fingers from your comfort zone. Because you only want the credit for reminding people, but not the cost of making decisions. Han Fei, on the other hand, dared to bear all the consequences of his full-scale push. That's the difference between an entrepreneur and a bureaucrat."

Zhao Guoqiang raised his eyebrows high.

Liu Zhenhua stood up and walked towards the door: "The strategic pilot project at the mobile headquarters is not a laboratory. We chose Qingniao because we value their willingness to charge into battle. If you can't accept this culture, you can choose not to attend subsequent strategic decision-making meetings."

"I...

""

The door slammed shut behind Liu Zhenhua. Chen Yuanshan and Fang Jing looked at Zhao Guoqiang, who was staring at the closed door with a look of astonishment.

Chen Yuanshan and Fang Jing stood up. Chen Yuanshan smiled slightly: "Old Zhao, we've known each other for twenty years. Let me give you one last piece of advice: don't let 'risk control' become your excuse for not daring to take the lead. The future of mobile isn't built by avoiding mistakes and holding on; it's built by daring to fight. If you still can't grasp this, from now on, at Qingniao's technical debriefing meetings, just listen to President Liu and don't attend. It'll save everyone the time of explaining what engineering realities are to you."

At 4:26 PM, Lu Hai sat in Han Fei's office, listening to Wen Jie recount how Han Fei had made a brilliant move late at night to save the reporter, how they seized the opportunity to push out the entire subscription, creating a subscription phenomenon, and how they finally turned the tide and skillfully resolved the system crisis.

Lu Hai blinked, his mouth slightly agape, listening in stunned silence: "My God, so much has happened in the day I've been gone. Wen Jie, what you just said about asynchronous queues... does that mean 'you wait in line, and I'll deal with you later'?"

Wen Jie frowned.

"Pretty much," Han Fei said, "but it's much faster than when you were tricked into revealing information in Guangdong last time, when your brain was processing it simultaneously."

Wen Jie opened his mouth, showing his teeth, and burst into laughter, laughing so hard that saliva dripped out.

"Lu Hai isn't processing synchronously at all," Zhang Ruiyi said with a laugh. "It's just running out of memory. When someone asks a question, it freezes for ages, and in the end, it even reverts to factory settings."

Lu Hai finally understood the joke and laughed along with Wen Jie.

"Hey, wait a minute!" Lu Hai said loudly, his brows furrowed. "I got a little confused by your tricks. Shouldn't the public opinion be growing stronger? It was on CCTV at noon, then online media were all over it this afternoon, and now it's almost evening, and it's going to be on the evening news. The attention is clearly going up. Why rush to push it out all at once this afternoon? If we wait until tonight, wouldn't more people see the news, and the conversion rate be higher?"

Wen Jie blinked and looked at Han Fei, as if he felt that Lu Hai's question had hit the nail on the head.

Han Fei glanced at Zhang Ruiyi and smiled slightly: "Lu Hai asked a good question. You think public opinion is a straight line that keeps rising, but in reality it's a parabola, and a very steep one at that."

"What do you mean?"

"You didn't experience the whole process today, so you might not understand," Zhang Ruiyi said. "Let me ask you a question: how long do you think it takes for users to go from being 'moved' to 'subscribing' after seeing the news?"

Lu Hai thought for a moment: "Uh... a day? Half a day?"

Zhang Ruiyi shook her head: "No. It was only a few hours, or even less. This afternoon, CCTV's 'News 30 Minutes'..."

A brief news report about the Yang Zheng incident was broadcast. Although it was only twenty seconds long, it was on CCTV. Tens of millions of people across the country turned on their TVs at that time and saw keywords like "late-night rescue" and "official praise." At that moment, users' emotions were at their purest—perhaps moved, angry, or even admiring. They weren't thinking about analyzing industry regulatory loopholes. They only knew that the publisher's president had saved someone, the authorities had given him praise, and they wanted to support him.

"You mean, things are different at night?"

"Yes, by evening more media outlets will be following up, and the sources and angles of the reports will increase. Some will delve into the black market supply chain, some will discuss journalist safety, some will invite experts to analyze regulatory gaps, and some will even start questioning whether Qingniao is using this incident to sabotage itself. Users' attention will be diverted, and their emotions will be diluted by various voices. When they see the text message again, they won't be thinking, 'This company is great,' but rather, 'Who is right and who is wrong?' or 'Who should I believe?' Each question will decrease users' willingness to act, making them hesitate for even a second when replying to subscribe. And that one second is enough to halve the conversion rate."