Where we start
Session 4 measured what posting more buys, and ended on its quietest number: across the whole site, the score an answer earned fell by roughly nine tenths in ten years, whatever any one person did. Tonight starts there. The running case is still Ines Calloway, a constructed mid-career data analyst who forecasts demand for a regional grocery chain; her readers, outposts and map below are constructed for teaching, and yours will differ. You bring the list of where your readers already read, from session 4's homework.
★ What you walk out with tonight. A one-page platform map: your two audiences, where each already reads, one owned home, no more than two rented outposts with one format each, the rule that every outpost points home, and a written exit plan for the day an outpost closes.
Whose land is the audience on?
6 min liveThe useful question is not "which platform is best" but "if this platform vanished tonight, what would I still have?" Ask it about each place you publish and the answer sorts your work into two piles without any further theory.
LiveOwned, rented, and what each costs you3 min▶
The idea has a source, and the source is narrower than the slogan. In "Sharecropping the Long Tail" (Rough Type, 19 December 2006, rereleased in 2011) Nicholas Carr described what he called digital sharecropping: platforms hand out the tools for free and keep ownership of what their users make with them. His target was who captures the value of user work. He did not frame it as "owned versus rented audience"; that phrasing is a later marketing gloss built on top of his point, and it is worth knowing which part is his.
Sonia Simone carried the idea to working professionals in "Digital Sharecropping" (Copyblogger, 28 July 2015): building only on social platforms is building "on someone else's land", and she credits the term to Carr. That is the version most practitioners now repeat.
What each costs. Rented ground is cheap to start and comes with readers already standing on it; the price is that reach, layout and the list of who follows you belong to the landlord. Owned ground is slow, empty on day one and needs you to bring readers to it; the price buys you the contact list and the address. Most working reputations need both, which is why the build-along asks for one home and a small number of outposts rather than a choice of one.
Self-studyWhat "owned" still depends on3 min▶
Nothing online is owned the way a house is. A domain is rented from a registrar by the year. An email list sits with a provider who can change prices or terms. A personal site runs on someone's hosting. So "owned" here has a working definition: you can move it. If the registrar, host or email provider changed tomorrow, you could take the address, the pages and the list of subscribers to a competitor and your readers would barely notice.
Test each thing you call owned with two questions. Can I export it in a standard format today, without asking anyone? Does the address that readers know point to me, not to the provider? A list you cannot export, or a site at an address that contains the provider's name, is rented ground with nicer furniture.
Keep the renewal date of your domain and the export button of your email list somewhere you will see them. The two most common ways people lose owned ground are a lapsed renewal and a provider that closed before anyone downloaded the list.
A consultant whose whole presence was one platform profile is a familiar situation: years of posts, a healthy following, every enquiry arriving through that one inbox. Then the platform changes its rules, how posts are shown, what links are allowed, or which accounts are visible, and the enquiries thin out with no way to write to the people who used to read. The work was real and the reputation was real. The way back to those readers was the platform's, and it always had been.
Rented ground can shrink under everyone
6 min liveSession 4's bench was built on the public data dump of Data Science Stack Exchange. Most of its numbers compared creators with each other. One number compared the site with itself across years, and it is the clearest picture of rented ground moving that real data offers.
LiveWhat the bench's quietest number says4 min▶
Across every account on the site, the mean score an answer earned was 7.36 in 2014, 3.86 in 2015, 3.62 in 2016, 3.46 in 2017, 2.53 in 2018, 1.73 in 2019, 1.36 in 2020, 0.99 in 2021 and 2022, and 0.82 in 2023. Answers per year peaked at 7,779 in 2019, when the score per answer was already under a quarter of where it started.
Read it as a fact about the ground, not about the people. A careful answer written in 2023 earned a fraction of what the same answer would have earned in 2014, and nobody posting there could have changed that by posting more, posting better, or posting at a different hour. The attention that made the site rewarding moved somewhere else, and the creators standing on it went down with it together.
Now set it beside Kevin Kelly's condition from session 2: a small niche works only if "they must pay you directly", which is to say you hold the relationship, not an intermediary. A score on someone else's site is the opposite case: the relationship sits with the platform, so when the platform's attention fades, yours fades with it. The defence is not a better post. It is a way back to your readers that does not run through the landlord.
Self-studyWhy one site is not every platform3 min▶
Be exact about what one site can show. Stack Exchange is search-driven question and answer: readers arrive from a search engine with a question, there is no follower graph, and "reach" means votes on an answer rather than an audience that sees your next post. A feed platform, where people follow accounts and an algorithm picks what they see, may rise and fall on a completely different curve. The decline here is real and measured; it is one site's curve, not a law for every platform.
Two more cautions. Score accumulates over time, so older answers have had longer to collect votes; some of the gap between 2014 and 2023 is age, not just fading attention. And the dump is a snapshot from April 2024, so the 2024 figure covers only the first quarter. What survives both cautions is the point that matters tonight: the rate a whole site paid for an answer moved by a large amount, and individual creators had no lever on it.
What the data does not give you is a number for any other platform's future. Nobody has published a verified one, and no reach curve for a feed platform is offered here.
The format follows the reader
5 min livePlatform advice usually starts with the platform: what it rewards this year, what length, what time. Start from the other end. Your readers already read somewhere; the format that fits is the one they already consume there; and every one of those places should carry a road back to your owned home.
LiveThree questions before choosing a platform3 min▶
1. Who exactly is this for? Name the audience from your session 2 line, not "people in my industry". Ines has two, and they do not read the same things.
2. Where do they already read, by their own account? Ask three of them what they read last week for work. Their answers beat any ranking of platforms, because a ranking describes the average user and you only need your readers. Ines's planners named trade newsletters, a professional network and conference talks; her data peers named code hosts, blogs and Q&A sites (constructed answers).
3. What do they consume there, and can you make it well? The format follows the place: a planner skimming a feed needs one chart and one fix in a few paragraphs; a data peer on a code host wants something they can run. If you cannot make the fitting format well and repeatedly, that outpost is not yours yet, however busy it looks.
Notice what is missing: best time to post, ideal post length, what the algorithm favours this quarter. None of those numbers has a verified source that holds across platforms and years, so none is given here. The reader questions do not expire.
Self-studyOutposts point home3 min▶
An outpost is a rented place you publish on deliberately, for an audience that is already there. It earns its keep only if some of the people who meet you there can find their way to ground you own. So every outpost gets one rule: each piece ends with one link home, to a page on your domain or a sign-up for your list, never a scatter of links.
Make the home worth arriving at. The page a planner lands on should hold the longer version of the case note they just read and a way to hear about the next one. The page a data peer lands on should hold the full notebook. A home that just repeats the profile gives nobody a reason to subscribe.
Turning one piece of work into the case note, the notebook and the home page version is a production skill of its own: drafting, repurposing, and where AI writing tools help and where they flatten your voice. That craft lives in learn-ai-content-with-phoebe; tonight is only about which ground each piece stands on.
Ines asks three planners and three data peers (constructed). Not one planner mentions the code host where she has spent two years posting; two of the three name a trade newsletter she had never considered. Her data peers, meanwhile, ignore the professional network almost entirely. Her old plan, to post the same thing everywhere, would have put planner material in front of peers and peer material in front of nobody. Six short conversations changed the map more than any platform guide could. Your conversations will say something different, which is the point of having them.
A one-page platform map
★ 22 min · everyone buildsSix steps, one page, with Ines's constructed map as the worked example. Use a plain document; the map should fit on a single screen so you can check it in under a minute. Session 6 builds the routine that keeps it running, so this map is the thing that routine serves.
List your two audiences. Take them from your session 2 positioning line. Two is a ceiling, not a target; one is fine. Name each as a group of people with a job, not as a market.
Find where each already reads, with evidence. Ask three people from each audience one question: what did you read last week for work, and where? Write down their words. A guess about where people read is the most common error on these maps.
Choose one owned home. One address you can move: a site on your own domain, an email list you can export, ideally both joined. Write the export test next to it so you know it passes.
Choose at most two rented outposts. Only places named in step 2. If a platform you already use was not named by anyone, it is not an outpost for this map, however much time you have spent there.
One format per outpost, and the link home. The format your readers already consume in that place, one you can make well and repeatedly. Then write the single link every piece will end with.
Write the exit plan. Imagine each outpost closed tomorrow. What would you export, from where, and how long would it take? Anything you cannot export is a risk you have now named. Do the first export tonight if you never have.
Save it as platform-map_YYYY-MM-DD, next to your dated search audit. When session 6 repeats the audit, the map tells you which new results you should expect to see and on which ground they should sit.
Before session 6
30 min- Have the six conversations. If you guessed in step 2 tonight, ask the three people per audience this week and correct the map with their words.
- Run the export you have never run. Download whatever your main outpost lets you take, and your email list if you have one, and note what was missing from the file.
- Add the link home to the bio of each outpost and to the end of your next piece there. One link, the same one everywhere.
- Count your current places. List every platform where you have an active profile. Anything not on the map gets a one-line pointer to your home, not new posts.
Three questions
Clear all three and session 5 gets stamped in your passport.
1 · What was Nicholas Carr's digital sharecropping point, in his 2006 post?
Carr's target was who captures the value of user work. The audience framing is a later gloss; Copyblogger's 2015 piece is where many professionals met it, and it credits the term to Carr.
2 · On Data Science Stack Exchange the site-wide mean score per answer fell from 7.36 in 2014 to 0.82 in 2023. What does that show?
The number describes the ground, not the people on it. A also misses that older answers had longer to collect votes, and B turns one site's curve into a law nobody has measured.
3 · Why should every piece on a rented outpost end with one link to your owned home?
It is Kelly's direct-relationship condition in practical form. B and C are claims about algorithms with no verified source, and one clear link beats a scatter of them.
What this session teaches, and where it came from
Every fact on this page, and its verification tier, is recorded in the course's source map.
Session 5 cheat sheet · pin this
You have a map. Session 6 builds the system that keeps it going: a cadence you can actually keep, chosen on the bench's evidence that cadence is a volume dial rather than a growth lever; an idea bank so the next piece is never a blank page; and the re-audit, where you search your own name again and compare it with the dated first page from session 1. Bring your search audit and tonight's map.