…
. Also headings:? Actually WP heading block: …
. We can use h2 or h3. Use plain HTML paragraphs and headings with those comments.
We need to count words 450-500 inclusive. Must be concise, actionable. Use facts from e-book. Must not use placeholders. At end include paragraph promoting e-book with link provided.
We must not include any thinking process. Just output article.
Let’s craft about 470 words.
We need to count words. Let’s draft then count.
Draft:
Then HTML.
We’ll need to count words in the entire article after “Title:” line? Probably count words of the content (excluding title line?). Safer to count everything after Title line? We’ll include title line not counted? Usually they count article content. Safer to make content 450-500 words, title separate.
Let’s write content approx 470 words.
We’ll write paragraphs.
We’ll need to use HTML comments.
Let’s draft:
Micro‑CPG founders in specialty food face a constant battle for shelf space. By turning retailer visits into a repeatable data‑collection routine, you can feed an AI system that spots gaps, pricing opportunities, and competitor weaknesses before your next buyer meeting.
Build a Standard Photo Protocol
When you walk a store, capture four consistent shots: a wide view of the category, a close‑up of the shelf where your product belongs, a close‑up of 2‑3 competitor price tags, and any empty or out‑of‑stock space. This “Founder’s Photo Protocol” ensures every image contains the same variables, making computer‑vision analysis reliable.
Leverage Computer Vision for Shelf Insights
Upload the four photos to an LLM with vision (ChatGPT‑4, Claude, or Gemini Advanced). Prompt the model to extract: total facings, share of shelf, price points, and any voids. For example, the AI can read that national kale chips sit at $9.99, national root vegetable chips at $6.99, and that the $7.99 price point is completely absent in the local subsection.
Combine Physical and Digital Data
While photos give you the in‑store picture, automate digital shelf monitoring by scraping store websites, Instagram posts, and Google Maps reviews. Use a simple prompt framework: paste compiled text from reviews, website descriptions, and social media, then ask the AI to summarize sentiment, highlight recurring complaints, and note any promotional language.
Generate a Weekly Intelligence Brief
Set up a weekly workflow: a gig worker or you collect the four photos and a text dump from target accounts; feed both into your chosen LLM; receive a structured brief that lists: shelf gaps, price‑point opportunities (e.g., the missing $7.99 tier), competitor vulnerabilities (out‑of‑stock tags, low review scores), and recommended talking points for brokers.
Apply the Brief to Buyer Pitch Emails
Personalize each outreach line with the AI‑generated insight: “I noticed your $6.99‑$9.99 chip segment lacks a $7.99 offering, and our new savory blend fits that gap while scoring 4.8 stars in regional reviews.” This level of specificity shows you’ve done the homework and increases reply rates.
Prep Broker Meetings with Visual Evidence
Before a broker meeting, export the analyzed photo that highlights the empty 8‑inch space between the $6.99 and $9.99 products. Attach it to your prep brief; the visual proof makes the pricing opportunity impossible to ignore and speeds up decision‑making.
For a comprehensive guide with detailed workflows, templates, and additional strategies, see my e-book: AI for Micro-CPG Founders in Specialty Food: How to Automate Buyer Pitch Email Personalization and Broker Meeting Prep Briefs.
Now count words. We need to count words in the content (excluding the title line?). Let’s count everything after “Title: …” line, including the HTML comments? Probably they count visible words only, but safer to count all words including those inside tags? Usually they’d count the article text, not the HTML markup. But to be safe, we can aim for about 470 visible words; the markup adds extra words like “wp:paragraph” etc which might be counted if they count raw output. Better to minimize extra words in comments. The comments contain words like “wp:paragraph”, “heading”, etc. Those add to count. We need to ensure total 450-500 words including those. Let’s count manually. I’ll copy the content and count words. I’ll write a plain text version ignoring HTML tags but counting words inside comments? Let’s just count visible words (the actual sentences). The HTML comments are not visible; but if they count raw output they’d count them. The instruction: “write as plain HTML paragraphs and headings (e.g.,…
)”. They likely expect that format, but word count probably refers to the visible content. However to be safe, we can keep visible words around 460-480, and the extra comment words will push over 500 maybe. Let’s estimate comment words. Each paragraph block: (2 words? actually “wp:paragraph” counts as one? It’s one token but includes colon. We’ll count as one word maybe.) plus (2 words?). Let’s approximate each block adds about 3 words. We have many blocks. Let’s count visible words first, then add comment words. I’ll extract visible sentences. Paragraph 1: “Micro‑CPG founders in specialty food face a constant battle for shelf space. By turning retailer visits into a repeatable data‑collection routine, you can feed an AI system that spots gaps, pricing opportunities, and competitor weaknesses before your next buyer meeting.” Count words: Micro‑CPG(1) founders2 in3 specialty4 food5 face6 a7 constant8 battle9 for10 shelf11 space12. By13 turning14 retailer15 visits16 into17 a18 repeatable19 data‑collection20 routine,21 you22 can23 feed24 an25 AI26 system27 that28 spots29 gaps,30 pricing31 opportunities,32 and33 competitor34 weaknesses35 before36 your37 next38 buyer39 meeting40. So 40 words. Heading 2: “Build a Standard Photo Protocol” => words: Build1 a2 Standard3 Photo4 Protocol5 => 5 words. Paragraph after heading: “When you walk a store, capture four consistent shots: a wide view of the category, a close‑up of the shelf where your product belongs, a close‑up of 2‑3 competitor price tags, and any empty or out‑of‑stock space. This “Founder’s Photo Protocol” ensures every image contains the same variables, making computer‑vision analysis reliable.” Count: When1 you2 walk3 a4 store,5 capture6 four7 consistent8 shots:9 a10 wide11 view12 of13 the14 category,15 a16 close‑up17 of18 the19 shelf20 where21 your22 product23 belongs,24 a25 close‑up26 of27 2‑328 competitor29 price30 tags,31 and32 any33 empty34 or35 out‑of‑stock36 space.37 This38 “Founder’s39 Photo40 Protocol”41 ensures42 every43 image44 contains45 the46 same47 variables,48 making49 computer‑vision50 analysis51 reliable52. 52 words. Heading 2: “Leverage Computer Vision for Shelf Insights” => words: Leverage1 Computer2 Vision3 for4 Shelf5 Insights6 => 6. Paragraph: “Upload the four photos to an LLM with vision (ChatGPT‑4, Claude, or Gemini Advanced). Prompt the model to extract: total facings, share of shelf, price points, and any voids. For example, the AI can read that national kale chips sit at $9.99, national root vegetable chips at $6.99, and that the $7.99 price point is completely absent in the local subsection.” Count: Upload1 the2 four3 photos4 to5 an6 LLM7 with8 vision9 (ChatGPT‑4,10 Claude,11 or12 Gemini13 Advanced).14 Prompt15 the16 model17 to18 extract:19 total20 facings,21 share22 of23 shelf,24 price25 points,26 and27 any28 voids.29 For30 example,31 the32 AI33 can34 read35 that36 national37 kale38 chips39 sit40 at41 $9.99,42 national43 root44 vegetable45 chips46 at47 $6.99,48 and49 that50 the51 $7.9952 price53 point54 is55 completely56 absent57 in58 the59 local60 subsection61. 61 words. Heading 2: “Combine Physical and Digital Data” => words: Combine1 Physical2 and3 Digital4 Data5 =>5. Paragraph: “While photos give you the in‑store picture, automate digital shelf monitoring by scraping store websites, Instagram posts, and Google Maps reviews. Use a simple prompt framework: paste compiled text from reviews, website descriptions, and social media, then ask the AI to summarize sentiment, highlight recurring complaints, and note any promotional language.” Count: While1 photos2 give3 you4 the5 in‑store6 picture,7 automate8 digital9 shelf10 monitoring11 by12 scraping13 store14 websites,15 Instagram16 posts,17 and18 Google19 Maps20 reviews.21 Use22 a23 simple24 prompt25 framework:26 paste27 compiled28 text29 from30 reviews,31 website32 descriptions,33 and34 social35 media,36 then37 ask38 the39 AI40 to41 summarize42 sentiment,43 highlight44 recurring45 complaints,46 and47 note48 any49 promotional50 language51. 51 words. Heading 2: “Generate a Weekly Intelligence Brief” => words: Generate1 a2 Weekly3 Intelligence4 Brief5 =>5. Paragraph: “Set up a weekly workflow: a gig worker or you collect the four photos and a text dump from target accounts; feed both into your chosen LLM; receive a structured brief that lists: shelf gaps, price‑point opportunities (e.g., the missing $7.99 tier), competitor vulnerabilities (out‑of‑stock tags