Content Recipes

Content Recipes

Reverse-engineered from 913,000 real corporate posts.

25 content formats that 958 marketing teams rebuilt by hand, turned into fill-in-the-blank recipes with 75 worked examples.

One PDF, one Notion database, one markdown bundle. No login, no membership, no video.

Get the library - $29 (launch price)Checkout opens shortly.Take 3 recipes freeCheckout opens shortly.

Launch price for the first 14 days or the first 100 buyers, whichever comes first. It rises to $49 14 days after launch. 30-day refund, no questions.

posts mined
913,000posts mined
companies
958companies
recipes
25recipes
worked examples
75worked examples
  • AOK
  • Commerzbank
  • Siemens Energy
  • ADAC
  • UNICEF

These companies used the platform we mined; they did not endorse this product.

It is Monday and you owe someone eight posts.

You open the document. The month is empty. You know roughly what the client sells and roughly who buys it, and none of that tells you what to put in the first box. So you scroll competitors, save three things you will never use, and lose the morning to a decision you have made two hundred times before and still make from scratch every time.

You ask an AI for ideas. You get twenty of them, all plausible, none of them wrong enough to reject and none of them right enough to run. It will happily write the post too. What it will not tell you is which of the twenty is the one this account needs this month, or when the draft in front of you is finished.

Then the client asks why this post. And the honest answer, the one you do not say out loud, is that it was the idea that arrived. There is no method to point at, no reason the format was chosen, and no number that says whether last month worked or whether the account is simply normal.

What you get

One zip and one link. No login, no membership area, no video course.

Content Recipes: The Library

The typeset PDF. All 25 recipes, each one a complete production kit: when to use it, when not to, the brief you fill before writing, the structure, the voice, the per-channel variants, the checklist and the failure modes.

The Notion duplicate

The same 25 recipes as a Notion database with filterable properties (intent, campaign shape, channel fit, effort), one page per recipe, plus a board view for planning the month. Delivered as a duplicate link.

The raw markdown bundle

A zip of the 25 recipe files. Paste a recipe straight into your own AI tool, your notes app or a client document. It is the source the PDF is built from, so nothing is held back.

The Benchmark Card

A one-page PDF of the benchmark medians, made to be screenshotted and sent to a client who thinks a post underperformed. The full table is on this page too; you get the printable version.

The 25 recipes

The whole list, with the evidence behind each entry. Ranks 1 to 12 carry counts from the dataset. The rest are labeled for what they are, and they carry no counts, because they were not counted.

Tier A: demand-ranked

Counted in the dataset and ranked by how many distinct companies built the format. Accounts is the breadth of demand; campaigns is the frequency.

#RecipeEvidenceShape
1The Event RunwayThe 6-piece sequence that fills seats before, during and after an event or webinar178 accounts / 916 stories30-day sprint
2The Human PortraitFreeOne employee, one real question, one quote that makes the company hireable138 accounts / 629 campaignsBurst (2-3)
3The Honest Job PostA hiring post that filters instead of flatters, and gets shared by the team136 accounts / 504 campaignsBurst (2-4)
4The Calendar HookFreeThe occasion post that is about the audience, not about the holiday134 accounts / 1,691 campaignsSingle post
5The Practical TipFreeOne tip, one proof, one thing they can do in ten minutes114 accounts / 532 campaignsSprint (8-10)
6The Asset DropTurning a whitepaper, study or eBook into 5 posts that gate one download112 accounts / 656 campaigns30-day sprint
7The Product Launch ArcTease, reveal, proof, objection, deadline: the five beats of a launch105 accounts / 494 campaigns30-day sprint
8The Newsletter TrailerPromoting an issue so the post is useful even to people who never subscribe92 accounts / 671 campaignsRecurring single
9The Milestone MarkerAn anniversary post that credits customers instead of celebrating yourself84 accounts / 253 campaignsBurst (2-3)
10The Episode PullPulling one 40-second idea out of a podcast so the clip stands alone72 accounts / 261 campaignsRecurring burst
11The Customer StoryA case study written as a decision, not as a testimonial57 accounts / 174 campaignsBurst (3-4)
12The Giveaway MechanicA contest whose entry condition selects buyers, not prize hunters51 accounts / 184 campaigns4-day burst

Tier B: platform taxonomy

Documented journalistic formats that the platform shipped as built-in content types. That is evidence of design intent, not of counted demand, so these carry no counts.

#RecipeEvidenceShape
13The Plain ExplainerAnswer the one question your sales team is asked every weekplatform taxonomySprint
14Behind the CurtainA process shown at the moment it is ugly, which is the only moment it is interestingplatform taxonomyRecurring single
15The Curator's PickShare someone else's work with a take that only you could addplatform taxonomyRecurring single
16The Stake in the GroundAn opinion narrow enough to be wrong, which is what makes it worth readingplatform taxonomySingle post
17The Ranked ListA listicle with a stated ranking rule, so it is an argument and not fillerplatform taxonomySingle post
18The Field ReportReportage from a place your audience cannot goplatform taxonomyBurst (2-3)
19The News BriefCompany news written so a stranger cares by the second sentenceplatform taxonomySingle post

Tier C: derived from the intent model

Built from the intent model and the value axes rather than counted in the data. Labeled as derived, and they carry no counts. The column below is the primary intent, not a number of accounts.

#RecipeEvidencePrimary intent
20The Number That LandsOne statistic, one interpretation, one consequence for the readerderivedInformation
21The Myth CorrectionName the belief, concede why it is believable, then break itderivedSharing
22The Before / AfterProof shown as a state change, with the boring middle includedderivedIdentification
23A Day InAn hour-by-hour that recruits, sells and humanizes at oncederivedIdentification
24The Open QuestionA question you genuinely do not know the answer to, asked in publicderivedFun / Sharing
25The RecapA month, quarter or year summarized as decisions, not as achievementsderivedInformation

Three of them are free: recipe 2 The Human Portrait, recipe 4 The Calendar Hook and recipe 5 The Practical Tip, complete and unabridged, with the intent primer and the full benchmark table. Recipe 2 is on this page below, so you can read one before you decide whether to hand over an email address for the other two.

Anatomy of a recipe

Every recipe has the same 14 sections, in the same order. No recipe skips one. That is what makes the library usable under time pressure: you already know where the answer is.

  1. 1Name and promiseOne sentence on what this format produces.
  2. 2EvidenceThe count and its honesty label, plus one sentence of what the count means.
  3. 3Use it when / do not use it whenTwo lists. The second one names the trap a beginner falls into.
  4. 4Intent and valueExactly one primary intent from the four, one or two value axes, and what changes if you swap them.
  5. 5Campaign shapeSingle post, 4-day burst or 30-day sprint, and how many pieces this yields.
  6. 6The briefThe fields you fill in before writing a word. Concrete fields, never 'the message'.
  7. 7StructureHeadline pattern options with a written example of each, intro pattern, the body beats, the ending.
  8. 8VoiceThe tonality pair, address form and complexity level, plus one alternative and what it costs.
  9. 9Channel variantsLinkedIn, Instagram, Facebook, X, TikTok. What changes, or an explicit 'skip' and why.
  10. 10Worked examplesThree complete, postable posts for three different business types.
  11. 11Quality checklistBinary yes/no checks you answer by looking at the draft. Never 'is it good?'.
  12. 12What good looks likeThe KPI expectation against the benchmark medians, and the number that means it is not working.
  13. 13Failure modesThe named ways this recipe goes wrong, each with the one-line fix.
  14. 14RepurposeHow to get a second and third piece out of the same work.

Read a full recipe

2. The Human Portrait

Complete, unabridged, not gated behind an email address. All 14 sections, exactly as they appear in the PDF. This is the quality bar for the other 24.

Promise: one employee, one real question, one quote that makes the company look like somewhere a person could work.

Evidence. Demand-ranked #2: 138 of 958 companies built this format, across 629 campaigns. Interviews and employee portraits are the single most broadly re-created format in the dataset. Read that as: almost every mature marketing team eventually discovers that a face outperforms a logo, and then builds a repeatable process for it.

Use it when

  • You need to look hireable (before, during or after a hiring push).
  • A person on your team has genuine expertise your audience wants and your brand account cannot credibly claim.
  • Your feed has drifted into product, product, product and engagement is flat.
  • You want content that the subject and their whole network will reshare, which is the cheapest reach you can buy.

Do not use it when

  • The person does not want to be posted. A reluctant portrait reads as reluctant, always. Never negotiate this one.
  • You want leads this week. This is not a lead format. Reaching for a portrait when the pipeline is empty is the most common misuse.
  • The only interesting thing about the person is their job title.
  • You cannot get one specific, concrete, slightly-risky sentence out of them. Without that sentence there is no post, only a bio.

Intent and value

Primary intent: Sharing ("excite people to share"). The reshare by the subject and their network is the mechanism, not a bonus. Value axes: identification (primary),knowledge (secondary, if the person teaches something). If you switch the intent to Hot lead (adding "we are hiring, apply here"), reach drops and you have built recipe #3 instead. Pick one.

Campaign shape

4-day burst, 2-3 pieces: the portrait itself (day 1), a pull-quote graphic or short clip (day 3), and optionally the subject's own first-person version on their personal account (day 4). Do not run these on the same day; they cannibalize each other.

The brief (fill before writing)

  1. Who: name, role, how long here.
  2. The one question: the single question you will actually ask. Not a list. One.
  3. The sentence: the one quote you expect or hope to get, in their words, that a stranger would repeat.
  4. The concrete detail: one specific object, number, place or moment (the whiteboard, the 4am callout, the 300-page spec).
  5. The mild risk: what this person is willing to say that a corporate account would not.
  6. Who it is for: the reader who should think "these are my people".
  7. The photo: what it shows. Working, not posing. Say where and when it is taken.
  8. What we will not say: the boundary the subject sets before the interview. Write it down and honor it.

Structure

  • Headline patterns. Best fit: "Name: quote"("Marta: 'I stopped answering support tickets at 9pm and our churn fell.'"). Alternate: question ("What does a support lead do at 4am?"). Avoid the plain factual pattern here; "Meet our new Head of Design" is the version nobody reads.
  • Intro pattern. Scene-setting. Open on the concrete detail from brief field 4, one sentence, no context. Context is the second paragraph, never the first.
  • Body beats.
    1. The concrete moment (1-2 sentences, present tense).
    2. Who this is and why they were the person in that moment (1 sentence, this is your only bio sentence).
    3. The one question, asked in the text.
    4. Their answer, in their words, as a real quote. Longest block on the page. Do not clean it up into corporate English; keep one imperfect phrase.
    5. The thing the answer implies about how the company works (1-2 sentences, and this is the only place the brand speaks).
  • Ending pattern. Tease a sequel ("Marta is writing up the tags she uses; we will post it next week"), not act-now. An act-now CTA on a portrait converts the post from Sharing to Hot lead and the subject's network stops resharing it.

Voice

Tonality pair: personal + storytelling. Address form: informal, direct-you. Complexity: layperson. Alternative pair:personal + ironic, which works when the subject is genuinely funny and fails badly when they are not - if you are unsure, they are not.

Channel variants

  • LinkedIn. The home network for this recipe; the dataset's highest engagement rate (median 4%) lives here and portraits are what it rewards. 120-200 words, the quote as its own line with white space around it, one photo. Tag the subject so their network is reachable. No link in the post body.
  • Instagram. Photo carries it. Slide 1 is the face plus the quote in large type; slides 2-4 are the story. Caption repeats the quote in the first line because that is all that shows before "more".
  • Facebook. Longer is fine here and older audiences read portraits. Same text as LinkedIn plus 2-3 more sentences of context. Strong fit for local businesses whose customers know the person by sight.
  • X. Skip the full portrait. Post the single quote plus the photo, and nothing else. The dataset's median of 3 favorites per post tells you not to spend production budget here.
  • TikTok. Only if you can get 20-40 seconds of the person actually answering the question on camera. The unpolished answer is the asset. A scripted-to-camera version of the same words performs worse than not posting.

Worked examples

Ledgerloop (B2B SaaS) - demonstrates the pull-quote as headline plus a mild, specific risk.

Priya: "We stopped calling it 'accounts receivable'. Nobody was late on purpose."

There is a whiteboard behind Priya's desk with 41 agency names on it and a number next to each one. It is a list of people who owe our customers money.

Priya runs support at Ledgerloop and spent last quarter reading every overdue-invoice thread our customers forwarded us.

I asked her what surprised her.

"Almost nobody was late on purpose. They were late because the invoice went to the wrong person and nobody followed up for nineteen days. We were building tools to chase deadbeats and the actual problem was that the email went to a designer's inbox and she went on holiday. So we stopped calling it accounts receivable. It is a routing problem."

That is why the next release does not have a bigger dunning engine in it. It has a "who actually opened this" column.

Priya is writing up the four questions she now asks every customer about their invoicing setup. Next week.

Halstead Roofing (local service) - demonstrates scene-setting plus identification for a non-office business.

What a roofer does at 4am

It is 4:10am and Danny is standing in a stranger's kitchen holding a bucket, waiting for the rain to slow down enough to get on a roof safely.

Danny has been with us eleven years and takes the emergency calls.

I asked him what people get wrong about emergency calls.

"Everyone thinks the job is fixing the roof. At four in the morning you can't fix a roof, it's dark and it's wet and I'm not going up there. The job is stopping the water getting into the ceiling and telling the family honestly that we'll be back at eight. Half of them have been up since midnight thinking the house is ruined. It isn't. Nine times out of ten it's a two-hour job in daylight."

That is why we answer the phone at night even though we cannot do the repair until morning. The call is the service.

Danny is putting together the three things to do before we arrive. We will post it before the autumn storms.

Tidepool (consumer app) - demonstrates the format when the subject is technical and the audience is not.

Sam: "Your watch doesn't know you're asleep. It's guessing, and so are we."

Sam has a spreadsheet with 1,900 nights in it. All his own.

He builds the sleep-staging model at Tidepool and has worn three trackers at once for over five years.

I asked him the question we get in every single support email: why does the app say I slept badly when I feel fine?

"Because your watch doesn't know you're asleep. It's guessing from movement and heart rate, and so are we. On a good night we're close. On a night you had two glasses of wine and fell asleep on the sofa, we are confidently wrong. I'd rather we told you that. So now when the model isn't sure, the app says it isn't sure instead of inventing a number."

That is the change in 4.2. Fewer confident numbers, more honest ones.

Sam is publishing what the app actually measures versus what it infers. Next month.

Quality checklist (all must be yes)

  1. Is there exactly one named person, not a team?
  2. Does the first sentence contain a concrete object, number, place or time?
  3. Is there a real quote of 25 words or more, in their voice, with at least one phrase you would not have written yourself?
  4. Is the bio exactly one sentence?
  5. Does the brand speak in only one place, after the quote?
  6. Is there no "we are hiring" and no application link in the post body?
  7. Did the subject read it and approve it before posting?
  8. Does the photo show them doing something rather than facing the camera?
  9. Would this post still make sense if you deleted the company name?
  10. Is there a stated next thing, so the sequel is promised rather than hoped for?

What good looks like

This is a Sharing post, so judge it on reshares first and reactions second. On LinkedIn, a portrait should beat your account's own median engagement rate; against the dataset median of 4%, a working portrait is your best post of the month, not an average one. OnInstagram, faces normally outperform: if the portrait lands at the corporate median of 51 likes it did not work for you. OnFacebook, expect reach at or above your page median of around 1,400 and reactions well above the median of 10, because the subject's own network arrives. On X, expect the usual 3 favorites and do not draw conclusions from it. If three consecutive portraits do not beat your own account's median, the problem is the quotes, not the format: you are getting job descriptions instead of sentences.

Failure modes

  • The press release portrait. "We are delighted to welcome..." Fix: delete every sentence before the concrete detail and start there.
  • The sanitized quote. Someone in comms smoothed the answer into corporate English and it now says nothing. Fix: restore one imperfect phrase; that phrase is the proof a human said it.
  • The bolted-on CTA. A hiring link at the bottom turns Sharing into Hot lead and kills the reshare. Fix: move the hiring ask to a separate post three days later (recipe #3).
  • The reluctant subject. Reads as stiff no matter how well written. Fix: pick someone else. There is always someone else.

Repurpose

  1. The quote alone as a pull-quote graphic, 2-3 days later, on the same networks.
  2. The subject posts the same story in first person on their own account, linking nothing. Their version usually outperforms yours.
  3. Three portraits with the same question become a "we asked five people the same thing" roundup (recipe #17, The Ranked List, or #25, The Recap).
  4. The promised sequel becomes recipe #5, The Practical Tip, written by the subject.

That is one of 25, and it is the same length and the same density as the other 24.

Where the data comes from

The source is a mined production database of Scompler, a German content-strategy SaaS that ran from 2015 to 2022 and is now defunct. It held 958 workspaces, 913,000 content items and 650,000 published posts across Facebook, Instagram, LinkedIn, X and others, over seven years of continuous operation.

The thing that makes it a product rather than a database is a design accident. Scompler let each team name its own campaigns and content types. So companies re-created the same formats by hand, independently, thousands of times, without ever agreeing on a name for them. Counting those re-creations gives a demand-ranked list of the formats corporate marketing actually runs. It is revealed preference at scale, not a survey and not an opinion.

Two numbers are reported per format. Accounts is how many distinct companies built it, which is the breadth of demand and the primary ranking signal: a format that 138 different companies invented independently is a real format. Campaigns is how often it was built in total, which is frequency and the secondary signal.

What this data is not

What good looks like

Real medians per post from the published-post records. Most people quit on results that were statistically normal, because nobody ever told them what normal was. Here it is, free, whether you buy anything or not.

NetworkMetricMedian per postReading
FacebookReach1,400A page post that reaches around 1.4K is normal, not a failure
FacebookReactions10Ten reactions is the middle of the market. Not a sign of bad content.
InstagramLikes51Fifty likes is a median corporate post
LinkedInEngagement rate4%The highest-engagement network in the set by a wide margin
XFavorites3Three. This is why X is a distribution channel, not an engagement channel.

Read this before you use the table. These are medians across corporate accounts 2015-2022, skewed to German-market B2B and public-sector brands, and platform reach mechanics have changed since. Use them as a floor for 'is this normal', not as a target.

Who this is for

  • The solo SMM freelancer with a handful of clients who needs a method with cited evidence, not another swipe file. You buy it once and reuse it on every client you ever take.
  • The agency junior or mid with no senior to copy from, who wants an instant playbook and a checklist that ends the 'is this good enough to send' loop.
  • The founder doing their own marketing, who mostly needs the benchmark table, because it is the thing that stops you quitting on results that were statistically normal.
  • The in-house marketer of one, who has to be strategist, writer and designer at the same time and needs a year of structured output on the shelf.

Who this is not for

  • Enterprise social teams that already have a playbook. You have this, in your own language, tuned to your own brand. You do not need ours.
  • Anyone looking for done-for-you content. Nobody writes your posts here. You get the format, the brief and three finished examples, and then you write.
  • Anyone who wants a Canva design pack. There are no templates, no graphics and no fonts in this. It is text and method.
  • Anyone who wants a guaranteed growth number. The data says which formats companies built, and what a normal result looks like. It does not promise you a result.

Pricing

One payment, no subscription. The launch price is a real, time-boxed discount: it runs for the first 14 days or the first 100 buyers, whichever comes first, and then the price is $49. There is no struck-through price on this page because there is no price we never charged.

The Library

$29one payment

Launch price. Rises to $49.

  • All 25 recipes as the typeset PDF
  • 75 worked examples, three per recipe
  • The Notion duplicate and the raw markdown bundle
  • The Benchmark Card
  • One person, personal use, yours forever
Get the library - $29Checkout opens shortly.

Studio

$99one payment

For an agency or a team.

  • Everything in The Library
  • Team and commercial license for up to 10 seats in one agency
  • Permission to adapt the recipes into client-facing deliverables under your own brand
  • The raw markdown bundle, licensed for that adaptation
  • One purchase, no seats to manage
Get Studio - $99Checkout opens shortly.

The launch price rises to $49 14 days after launch. Not sure yet? Take three complete recipes free and decide from those.

30 days, no questions, no form.

Email refund@[CONTACT_DOMAIN] with your order number and we send the money back the same day. Keep the files.

There is no exit survey, no 'tell us why', and nothing to delete. Asking people to delete files they already downloaded is theatre, and we would rather write the sentence we can actually keep.

Questions people actually ask

Is this AI-generated?

Partly, and here is exactly how. The evidence, the rankings, the framework and the benchmark medians come from the mined dataset, not from a model. The recipe structure and the exemplar recipe were written by a human. The remaining recipes and the worked examples were drafted with AI from that evidence, then curated and rewritten by a human in an editorial pass whose explicit job was to delete every invented statistic and every vague checklist item. Recipe 2 is on this page in full, ungated. That is the quality bar for all 25, and you can judge it before paying a cent.

Are the example companies real?

No. Ledgerloop, Halstead Roofing and Tidepool are three fictional standing personas: a B2B SaaS, a local service business and a consumer app. That is deliberate and it is better than the alternative. A real case study is one company's situation and you spend the first ten minutes deciding whether it maps to yours. Three invented personas across three business shapes let you pattern-match to your own client immediately, and no example is quietly a piece of someone's PR.

The brands on this page endorse the product, right?

No, and we will not imply otherwise. AOK, Commerzbank, Siemens Energy, ADAC and UNICEF were among the 958 companies whose workspaces existed on the platform we mined. They did not endorse this product, were not asked to, and have no relationship with it. They are named because the dataset's composition is part of the evidence.

Is it in English only?

Yes. English only at launch, one long document, no translated editions. The dataset is German in origin, so a German edition is the obvious next one, but it does not exist today and we would rather say so than pre-sell it.

Do I get updates?

You get every correction and revision to this edition, at the same download link, at no extra cost. If a later edition adds recipes or re-runs the analysis against new data, that is a separate product and you will be told the price, not billed automatically. There is no subscription and no card on file.

Can I use these for client work?

Yes on The Library for your own client work: the recipes are a method you apply, and posts you write with them are yours. What the personal license does not cover is packaging the recipes themselves into a deliverable you sell or hand over under your own brand, or sharing the files with a team. That is what the Studio license is for, at $99 for up to 10 seats.

What if it does not work for my niche?

Then email refund@[CONTACT_DOMAIN] and take the money back within 30 days. The honest version of the answer: these are formats corporate marketing teams built across seven years, and they are shapes rather than scripts, so they transfer better than most template packs. But the data skews to German-market B2B and public-sector brands, and if you sell to a crowd that is nothing like that, the channel notes will fit you better than the exact benchmark numbers will.

Why is it not just a list of 600 templates for $2?

Because those exist and they do not solve the problem. Volume is not scarce; nothing is easier to generate than another content idea. What is scarce is knowing which format this account needs this month, and when the draft is finished. Twenty-five formats with a decision rule, a checklist and three finished examples each is a different thing from 600 captions.

Do I need a Notion account?

No. The PDF and the markdown bundle are the product; the Notion duplicate is a convenience for people who already plan there. If you never open it you have lost nothing.

How do I get it after I pay?

Immediately, by email, from the checkout. The files come as one zip plus the Notion link. There is no login, no membership area and no waiting for a human. If the email does not arrive, write to [CONTACT_EMAIL] and we will send it by hand.

Provenance and AI disclosure

The evidence in this product comes from a mined production database of Scompler, a German content-strategy SaaS that operated from 2015 to 2022 and is now defunct: 958 workspaces, 913,000 content items, 650,000 published posts. The rankings are counts of how many distinct companies built each format. The benchmark medians are computed from the published-post records. Nothing in the numbers is estimated, modeled or rounded up for effect.

The writing was produced with AI assistance. Drafts were generated from that evidence and from a fixed recipe structure, then curated and rewritten by a human whose editorial pass had one non-negotiable rule: every number in the product traces to a single evidence file, and any sentence that needed a statistic which did not exist got rewritten so that it did not need one. Twelve recipes carry counts. Thirteen do not, and are labeled for what they are.

The companies named on this page used the platform that was mined. They did not endorse this product. The three companies in the worked examples are fictional. If you find a number in the product that you cannot trace, write to us and we will correct it publicly.

913,000 posts. 958 companies. 25 recipes.

One afternoon to learn, and you keep it. $29 at the launch price, $49 after that, and 30 days to take the money back without answering a single question.

Get the library - $29 (launch price)Checkout opens shortly.

Price rises to $49 14 days after launch. Or read the list again first.