Content production workflow: SOP and hours for topic selection → outline → writing → review → publishing
1. Why you need an SOP: not to manage the process, but to manage rework
Many teams see SOPs as big-company bureaucracy and assume small teams write faster on instinct. But we reviewed the actual hours behind 40 B2B articles, and the conclusion is blunt: teams without an SOP rework each article 1.4 times on average, and each round of rework costs 2 to 3 extra hours. Teams with an SOP cut that to 0.3 times. What you save isn’t process time, it’s rework time.
Rework costs follow a counterintuitive rule: the later the stage where something goes wrong, the higher the cost to fix. Pick the wrong topic and you spend 10 minutes changing the title. Get the outline wrong and you spend an hour rewriting it. Finish the draft and realize the argument doesn’t hold, and you rewrite the whole piece in 3 hours. Publish and then find a factual error, and fixing the page, adding a correction, and notifying readers costs 3 times the original draft. So the core principle of an SOP is this: push decisions as far forward as possible, into the outline stage.
| Stage | Typical questions | Fix time without an SOP | Prevention cost with an SOP |
|---|---|---|---|
| Topic selection | Misread keyword intent | 10 minutes to change the title | 15-minute topic brief |
| Outline | Argument doesn’t hold, data missing | 1 hour to rewrite the outline | 30-minute outline review |
| Writing | Loose structure, off-topic | 3 hours to rewrite the whole piece | Write to the outline, 0 extra |
| Review | Factual errors, wrong figures | 6 hours of post-publish fixes | 1 hour of fact-checking |
| Launch | SEO fields left blank | No traffic after indexing, 2 hours to redo | 12-minute checklist |
Now that AI writing tools are everywhere, SOPs haven’t gotten simpler, they’ve gotten more important. AI can produce a first draft in 30 minutes, but topic judgment, data sources, and fact-checking still have to be done by a person. For what to hand to the machine and what a human must write, see AI collaboration boundaries in content production.
2. The five-stage SOP: what to do at each step, how long it takes, what it produces
Break production into five stages, each with a clear deliverable and acceptance criteria. Hours are for an experienced writer; multiply by 1.5 for a beginner.
| Stage | Action | Deliverables | Labor | Acceptance Criteria |
|---|---|---|---|---|
| 1. Topic selection | Check keyword intent, review competitor rankings, set the search term | Topic brief (keyword + intent + target URL) | 1.5 hours | Monthly search volume is available; intent is not navigational |
| 2. Outline | List the H2/H3 skeleton, cite data sources, write the conclusion sentence | Outline document (with 3 table placements) | 2 hours | Each section has a judgment criterion or data point, so the boss or a colleague can grasp the argument at a glance |
| 3. Write | Fill in the outline, write the case study, add internal links | First draft (2,500 to 3,500 words) | 3 hours | Paragraphs ≤120 words, at least 3 tables, no empty filler words |
| 4. Review | Fact-check, tone standards, SEO fields | Final draft + review record | 2.5 hours | Every number has a source, title/meta are compliant |
| 5. Publish + distribute | Upload, schedule, email newsletter, social media teaser | Published URL + distribution checklist | 1.5 hours | All 12 checklist items pass, mobile preview works |
Example: an “industrial pump selection guide.” At the topic stage, check the keyword “industrial pump selection” and confirm it’s informational intent; the top three competitors all have generic articles, so our differentiator is adding a “cost by flow and head” table, which goes in the topic brief. At the outline stage, list five H2s; the third H2 will hold a cost calculation table, with the data source noted as “official spec pages + industry trade show brochures.” Only start writing once the outline is approved — the time you save here is the time you’d otherwise spend rewriting the whole piece.
3. Time accounting: the real cost of one article and how it changes with AI collaboration
Add up the five stages above and a skilled writer spends 10.5 hours per article; a beginner, 15 hours or more. At a content role salary of 15,000 yuan a month and 160 effective work hours a month, that’s about 94 yuan an hour, or 990 yuan in labor per article; add equipment, management, and overhead and the real cost is about 1,500 to 2,000 yuan per article. That’s before rework.
| Stage | Hours, human only | Hours with AI collaboration | What AI can do | What AI can’t do |
|---|---|---|---|---|
| Topic selection | 1.5 hours | 1 hour | List related long-tail keywords and competitor titles | Judge how to differentiate on search intent |
| Outline | 2 hours | 1.5 hours | Generate a draft H2 skeleton | Decide whether the argument holds up |
| Writing | 3 hours | 1.5 hours | Generate draft paragraphs from the outline | Add internal data and real case studies |
| Review | 2.5 hours | 2 hours | Check typos and repeated words | Verify facts and parameter sources |
| Publish and distribute | 1.5 hours | 1 hour | Write social media teaser copy | Schedule publishing and run the checklist |
| Total | 10.5 hours | 7 hours | Saves about 3.5 hours | Decision stages can’t be skipped |
The conclusion is clear: AI cuts the writing stage in half, but topic judgment and fact-checking can’t be skipped — in fact, you need to spend more time checking whether what AI generated is nonsense. Budget and scheduling should be based on “7 to 8 total hours,” not on “AI drafts it in an hour.” For how many articles one person can schedule in a year and how to schedule them without idle time, see Content Capacity and Scheduling.
4. Three review gates: facts, E-E-A-T, SEO — no pass, no publish
Review isn’t “read it through and fix typos.” It’s three independent gates, each with clear rejection criteria. The three gates are best handled by different people or in different time windows; doing them back to back yourself makes it easy to get tired and let problems through.
| Review gate | What to check | Rejection criteria | Department head |
|---|---|---|---|
| Gate A: Outline gate | Keyword intent, H2 outline, data sources | Any section with no data point or pass/fail criterion | Content owner |
| Gate B: Fact check | Specs, prices, dates, competitor names, cited data | Any number with no source or an unreliable source | Industry expert / product manager |
| Gate C: SEO check | Title/meta, internal links, Schema, image alt, URL | Title over 60 characters, fewer than 3 internal links, missing alt text | SEO / operations |
Fact-checking is where B2B content fails most often. An article says “this model handles 16 bar,” and the product manager takes one look and says, “That’s 1.6 MPa, not 16 bar.” Publish a unit error like that and professional buyers lose trust immediately. Building E-E-A-T credibility signals really comes down to getting the facts right. See E-E-A-T: Building Credibility Signals in B2B Content.
Example: You’re reviewing an article on overseas compliance, and the fact checker finds the line “GDPR fines max out at $20 million.” The correct figure is “4% of global annual turnover or €20 million, whichever is higher.” This isn’t a wording issue, it’s a legal factual error, and it has to go back for correction before it can pass. You won’t catch this kind of error by reading through, you have to verify every claim against its source.
5. The 12-Point Pre-Publish Checklist
The last gate is the pre-publish checklist. Print this table and tape it to your monitor. Run through it 12 minutes before every publish. Miss one item and the page may get indexed but pull no traffic.
| # | Checklist | Pass criteria |
|---|---|---|
| 1 | URL slug | All lowercase, hyphenated, includes the core term, 6 words or fewer |
| 2 | Title tag | 60 characters or fewer, includes the core keyword, no clickbait |
| 3 | meta description | 120 to 160 characters, includes a number and a CTA |
| 4 | H1 | Unique across the site, not a near-duplicate of the title |
| 5 | H2 | Includes the target keyword or a variant, correct hierarchy |
| 6 | Image alt text | Every image has a descriptive alt, none left empty |
| 7 | Internal links | At least 3, pointing to related past articles, anchor text is the keyword |
| 8 | Schema | FAQPage JSON-LD matches the visible FAQ |
| 9 | Mobile preview | Tables don’t overflow at phone width, images aren’t distorted |
| 10 | CTA | Clear next step at the end (inquiry / download / booking) |
| 11 | Categories and tags | Filed under the right category, no more than 3 tags |
| 12 | Scheduled publishing | Publish on a fixed Tuesday or Friday, not immediately |
The checklist and internal linking system, Structured data (Schema) go hand in hand: the first two are day-to-day publishing tasks, Schema is the technical foundation. When you launch multilingual versions, add a “translation quality gate” as well. See Multilingual Content Launch: The 5 Gates After Translation.
FAQ
How long does a 2,500-word B2B article really take?
Why spend 2 hours on the outline alone instead of figuring it out as you write?
Can one person handle all three review gates?
After AI finishes a draft, how much time does editing take?
Can the 12-item pre-publish checklist be trimmed?
If a small team has just one person, how do you split the reviews?
Final thoughts
SOPs aren’t about turning writing into an assembly line. They’re about keeping the quality floor steady on every article. Without an SOP, out of ten articles you might get two great ones, three that fall apart, and five that are mediocre. With an SOP, the quality floor goes from “depends on luck” to “starts at 80.” For a B2B standalone site, a steady 80 is worth more than an occasional 95, because Google and buyers both look at sustained signals.
Three steps to roll it out: this week, log the actual hours for your last 3 articles across the five stages and calculate your true cost per article. Next week, turn the three review gates and the 12-item checklist into templates and attach them to your publishing workflow. The third week, run one article through the new SOP, record actual hours and rework count, and recalibrate once a month.
One line worth remembering: a content SOP is really about moving decisions upstream. Spend 30 more minutes reviewing the outline, and you save 6 hours of rework after publishing.