At the China FMCG Conference held by New Distribution in September, we had in-depth discussions with many FMCG company executives and heads of SFA system providers about digital transformation in the FMCG industry. We reached a unanimous conclusion: The first step in digital transformation for FMCG companies is not studying digital patterns, nor making decisions based on numbers, but authentic data collection. If data loses authenticity, having data is more frightening than having no data. If the team doesn't fall in love with the terminal system, pause the digital transformation. During digital transformation, many companies focus on how to use data to study sales processes, sales trends, and formulate strategies and tactics, but overlook the most important point: data collection. If data collection goes wrong, the results of studying that data are undoubtedly disastrous, even worse than following intuition. Common problems in data collection include the following four aspects: 1. Partial representation: Data collection is localized or comprehensive only in key markets, while non-key markets are ignored. Research based on such data is undoubtedly detrimental to markets in the introduction and growth stages. 2. Falsification: Under company supervision and pressure, sales staff are forced to report false information and show only their good side, going to great lengths to make data look good. Examples include fake outlets, fake coverage rates, fake sales, fake visits, fake executions, etc. Such data is highly deceptive; leaders like to see it, but as a market reference, it can definitely lead the company astray. 3. Human intervention: When a particular market receives special attention from a leader, with human and material resources invested within their control (such as discretionary expenses not requiring headquarters approval), the market feedback appears positive, and headquarters quickly captures the data as a model for research. The results are predictable. 4. Data gaps: As the saying goes, "Know yourself and know your enemy, and you will never be defeated." If data collection focuses only on your own products and ignores key competitors, the value for future decisions will be low. Therefore, if these common problems are not resolved, do not push forward with digital transformation. But the reality is that many people would rather deliver food or packages than work in FMCG, because those jobs don't require thinking, handling complex customer relationships, or persuading seasoned store owners. So what is the profile of the remaining frontline execution staff in FMCG? 1. Age range: Multi-generational, including post-70s, post-80s, post-90s, and post-00s. In many cases, recruitment is difficult, and age restrictions cannot be imposed; often, anyone willing to work is hired. 2. Education level: The slogan is that brand owners require college degrees and distributors require high school diplomas, but in reality, they have to make exceptions. As long as they can use a smartphone and basic computer, they are hired, though a few have higher education. 3. Family background: They come from families with modest incomes and choose this job to stay close to home and care for their families. 4. Temptation level: The income of delivery drivers, couriers, and small business owners constantly tempts frontline FMCG workers to leave when they encounter difficulties or tedious tasks. How can we make these people serve the enterprise's digital transformation more happily? Or how can we make digital data collection more accurate? The answer is: laws are irreversible, and human nature cannot be violated. At this point, the terminal system used by frontline staff must not only have strong technical support but also be warm and considerate of users' emotions. Make the system understand the frontline, and let the frontline love the system. What levels constitute the enterprise's frontline staff? 1. Frontline operators: Whether for brand owners or distributors, they complete market tasks such as outlet visits, displays, and sales promotions. They are also the first step in digital data collection. 2. Frontline managers: They are responsible for checking the work of market operators and are the first line of defense for data authenticity. 3. Frontline auditors: They are responsible for verifying the authenticity, effectiveness, and compliance of market expenses, and are the second line of defense for data authenticity. For example, a company serving 1.5 million outlets needs about 10,000 frontline operators, about 1,500 frontline managers, and about 300 frontline auditors. The number of people working around the market itself is beyond the IT department's ability to supervise. Therefore, the core competitiveness of a business terminal system lies not only in how powerful and advanced the technology is, but more importantly, in how to create a warm system that frontline staff enjoy using. The first thing people love based on human nature is: simplicity and reliability. Let's take Baidu AI Cloud's usage logic as an example: 1. Make frontline operators simple and reliable: OCR text recognition for store signs can easily improve operators' work efficiency and achieve completeness of outlet data. We know that the outlets served by sales staff are dynamic—new openings, closures, and transfers. A daily task is collecting outlet information. Usually, entering one outlet's information takes 3 minutes, and the tedious input makes operators impatient. If review is not strict, a region may have dozens or hundreds of outlets all named "Blue Classic · Dream Blue." Such a system lacks care for frontline operators. With Baidu AI Cloud's technical support, the time for entering outlet information can be greatly reduced, and effective information extraction becomes more intelligent and efficient, making users happier. Another example: in densely populated outlet areas, when frontline operators use the system to capture storefront photos, how do they determine which is the "main entrance" and which is the "reference object"? With Baidu AI Cloud's support, this can be easily handled. Moreover, information collection changes from "fill-in-the-blank" to "multiple choice," making it more convenient and accurate. 2. Make frontline managers simple and reliable: Outlet deduplication uses an intelligent deduplication model based on NLP and geolocation retrieval. How do duplicate outlets arise? If we attribute it to fraud, we'd be unfairly condemning many staff. Honestly, in many cases, our frontline operators are not "intentionally fraudulent" but "unintentionally so." The reason is simple: most of their income comes from sales commissions (and the incentive for new outlets only lasts for a while), and sales require real money paid to the company, so duplicate outlets serve no purpose. Sources of duplicate outlets: a. When an old salesperson leaves, the new one is unfamiliar and re-enters the outlet. b. At regional boundaries, outlets are re-entered due to area adjustments. c. After the supervisor enters, the salesperson enters again. d. Memory lapses or fear of not passing review lead to re-entry. e. Deliberate re-entry for personal gain. So, for the 80% of salespeople with a proper attitude, the purpose of outlet deduplication is not to catch them by the tail, but to help them serve the market better. This is the warmth of the terminal system. Baidu AI Cloud's approach is three steps: a. Basic processing of outlet information: encode "outlet location" information, encode "outlet name," and normalize for judgment. b. Location retrieval: use the reconstructed address to locate on the map and search for normalized store names within 2 km. c. Similarity calculation based on deep learning—return duplicate groups. 3. Make frontline auditors simple and reliable: Vendor-specific anti-rephotography and anti-cross-location models can greatly improve recognition accuracy. If authenticity is not achieved, market expenses and personnel incentives are more terrifying when given than not given. The frontline auditors in this role are more about fighting against fraudsters, letting good money drive out bad, and purifying the market work environment. The biggest challenges they face are rephotography (flat rephotography, on-site forgery) and cross-location fraud (using photos from different outlets, misattributing them). Baidu AI Cloud's solutions are: 1. Support vendor-customized anti-rephotography models, greatly improving accuracy, safeguarding expenses and business. 2. Identify common cross-location fraud behaviors such as "different angles, slight movement, date tampering," ensuring the efficiency of market expense usage. Summary: Authentic data collection is the cornerstone of enterprise digital transformation. It requires collection by frontline operators, review by frontline managers, and verification by frontline auditors—all three are indispensable. Put yourself in their shoes: operators visit nearly 30 stores daily, managers need to review all subordinates' work, and auditors need to check all expense outlets. How long do they use the system each day? If the system doesn't think for them and improve their work efficiency, its value won't exist. What is the first step in enterprise digital transformation? Enterprise digital transformation is a big topic, but there are priorities. The core of current digitalization is twofold: one is to help increase business volume, and the other is to help improve organizational efficiency. The most urgent digitalization for the FMCG industry is channel digitalization, whose core is outlet digitalization, visit digitalization, and display & verification digitalization. The core of the above digitalization is how to truly implement it in the daily work of frontline operators. If they cannot use the terminal system happily, or even resist it, then the enterprise's digital transformation may fail like "a thousand-mile embankment collapses due to an ant nest." Finally, to answer the question: The first step in enterprise digital transformation is to use internet technology, based on human needs, to let frontline staff use a warm, simple, and reliable terminal system to happily carry out daily channel digital data collection. Click "Read Original" at the bottom for more details. Are you "watching" me?