Rules
Part of Why platform trends converge on the same handful of shapes
Platform trends guide: Track updates without the noise
Platform trends guide to tracking algorithm changes, emerging trends, and outages with named sources, discovery tools, and a weekly half-hour routine.
A platform trends guide is easy to start and almost impossible to keep. For teams watching TikTok, Instagram, YouTube, X, Facebook, or Reddit, the work is unbounded by default, and its reward is invisible.
Most people read everything for two weeks and then stop, or read nothing and find out late. This page describes a watch that fits in half an hour a week, with explicit rules about what enters it, what leaves it, and when you may do nothing.
What to take away
- A watch is defined by what it excludes. Without a written exclusion rule it grows until it is abandoned.
- Sort what you read by who wrote it and why, not by how interesting it is. Provenance predicts usefulness better than subject does.
- The output of a watch is a short written record, not a feeling that you are informed. If nothing is written down, the watch did not happen.
The intake rule
Something enters the watch only if it could change a decision you have already made. That is the whole rule, and it is strict enough to exclude most of what crosses your desk.
Does it enter the watch?
Could it change a decision you have already made?
In scope: add to the watch
Entertainment: leave it out
Applying it requires knowing your own decisions, which is why the first session is not reading at all. Write down the five or six standing commitments your operation runs on: where your audience comes from, which surfaces you publish to, which tools you depend on, what format you produce, and what you have decided not to do.
That list is the dependency inventory seen from your own side. Anything that could disturb one of those is in scope. Anything else is entertainment, which is fine, but it is not the watch.
Sources, ranked by what they can carry
| Source type | Good for | Weak at | Read it |
|---|---|---|---|
| Operator's own documentation and policy pages | What the operator commits to in public | Timing, weighting, anything internal | On a schedule |
| Operator's engineering or research writing | Mechanism, vocabulary, direction | Currency; often describes an older system | When it appears |
| Standards and government guidance | Obligations that apply to you | Anything product specific | Twice a year |
| Trade press | Knowing that something happened | Why it happened, and whether it matters | Skim, never quote |
| Practitioner posts | Hypotheses and early symptoms | Everything else | Weekly, as a queue |
| Aggregators and newsletters | Coverage | Judgment, which has been done for you | As an index only |
Almost everything in the middle rows is gray literature: material published outside formal channels, often unreviewed and undated. A discussion of source reliability is a reminder to record where such material came from. That is not a reason to ignore it, but it is a reason to check its provenance.
Use named sources to make the table actionable:
- For platform announcements, check Meta Newsroom and Instagram Blog for Facebook and Instagram, TikTok Newsroom, YouTube Help and Inside YouTube, and X Engineering.
- For trend discovery, try Google Trends, TikTok Creative Center, Exploding Topics, or Glimpse. Trade coverage from Platformer, Social Media Today, The Information, and Nieman Lab can add context, but treat it as reporting rather than an operator announcement.
- For outages, check the platform’s own status information where available and use Downdetector as a signal to investigate, not confirmation by itself.
- For research access, check the TikTok Research API, Reddit API, and X API, along with each platform’s current eligibility and access rules.
Before paying for a trend tool, record its quoted price and the staff time it saves; keep it only if that tradeoff suits your operation.
Sources ranked by what they carry
Source type
- Operator docs
- Public commitments
- Engineering writing
- Mechanism, direction
- Standards, government
- Obligations
- Trade press
- That it happened
- Practitioner posts
- Hypotheses, symptoms
- Aggregators
- Coverage
Good for
- Operator docs
- Timing, weighting
- Engineering writing
- Currency
- Standards, government
- Product specifics
- Trade press
- Why, whether it matters
- Practitioner posts
- Everything else
- Aggregators
- Judgment
Weak at
- Operator docs
- On a schedule
- Engineering writing
- When it appears
- Standards, government
- Twice a year
- Trade press
- Skim, never quote
- Practitioner posts
- Weekly, as a queue
- Aggregators
- As an index only
Read it
- Operator docs
- Engineering writing
- Standards, government
- Trade press
- Practitioner posts
- Aggregators
The order is deliberate, and it is roughly the reverse of how most people spend their attention. Practitioner posts are the most engaging and the least reliable; the reasons a practitioner claim decays explain why they belong in a queue of hypotheses rather than in a record of facts.
Government material matters when a platform change affects your obligations, not as an algorithm feed. In the United States, check the Federal Trade Commission and the Federal Trade Commission Act for consumer-protection issues. Follow the EU Digital Services Act or Ofcom’s UK guidance only when your audience or operations make those jurisdictions relevant, and keep regulatory notices separate from claims about ranking.
The weekly half hour
Split it in three.
The weekly half hour
- Ten minutes on the queuediscard, note, or investigate
- Ten minutes on noted itemsone sentence each
- Ten minutes on one investigation, only one
Ten minutes on the queue. Skim what accumulated. For each item, one of three verdicts: discard, note, or investigate. Most items are discarded, and if fewer than four in five are, your intake rule is too loose.
Ten minutes on the noted items. For each, write one sentence saying which of your standing commitments it touches and what would have to be true for it to matter. The three durability tests are the fastest way to write that sentence. Most items die here, which is the point.
Ten minutes on one investigation. Only one. If two things need investigating, the second waits a week, and if it still needs investigating then, it was real.
What leaves the watch
Items leave three ways, and naming them stops the list growing forever.
Three ways items leave
- Resolved: it happened, you acted, move to the record
- Expired: ninety days with no further signal
- Absorbed: two items are the same item, merge them
Resolved. The thing happened, you did something, it is now part of how you operate. Move it to the record.
Expired. Ninety days passed with no further signal. A change that mattered would have produced something by then, so an item with nothing behind it goes out.
Absorbed. Two items turn out to be the same item wearing different words. This happens constantly and merging them is most of the value of writing things down at all.
The record
One file, appended to, never rewritten. It is the same discipline as the change register kept for product changes, and the two files can be one file if that suits you. Each entry carries the date you noticed, the source, the commitment it touches, and what you decided. That is four fields and it takes a minute.
Four fields per record entry
- Date you noticed
- Source
- Commitment it touches
- What you decided
The reason for the record is not the record. It is that in six months somebody will assert that a change happened, and you will be able to check whether you saw it at the time. Without that, every argument about what changed is an argument between two memories.
A watch must survive the week when nothing happens, which is most weeks, and doing nothing and writing "nothing" is a legitimate output. It avoids manufacturing significance just because the calendar had a slot.
When something does happen, watch your own accounts rather than the announcement to learn how much of the change reaches you. TikTok’s For You recommendations, Instagram’s ranking of Reels, and YouTube recommendations are distinct algorithm areas to monitor; a BookTok spike is a trend example to check against discovery tools. Treat a Facebook and Instagram login outage as a service incident, not an algorithm change, and compare outage reports with the platforms’ own status information.
Common questions
How many sources is enough?
Fewer than you think, and the number matters less than the mix. One operator page per surface you depend on, one standards source for your obligations, and one practitioner queue is a complete watch. Adding a fourth newsletter adds coverage of the same events.
What if I miss something important?
You will, occasionally, and the watch is not insurance. Its purpose is to reduce the number of things that surprise you and to shorten the gap when one does. Anything that would be catastrophic to miss belongs in an inventory with an owner, not in a reading list.
Should the watch cover competitors?
Only as a symptom source. Watching another operation change tells you that something moved and nothing about what. You see their output and never their inputs, the same limit that makes outside evidence about ranking so weak.
Is a half hour really enough?
For a small operation, yes, provided the intake rule is doing its job. The hours people currently spend are mostly spent on items that would have been discarded in ten seconds under a written rule.
