ITS SM-CDT cheat sheet

Platform archetypes: social network (reciprocal follows), microblog (asymmetric, chronological or algorithmic feed), media-sharing (visual-first), forum/community (threaded, topic-based), messaging (private, high-intimacy).

Engagement rate formula (by reach): (likes + comments + shares + saves) ÷ reach × 100. By follower count: same numerator ÷ followers × 100. State which denominator you used.

Reach vs. impressions: reach counts unique accounts that saw a post; impressions count total displays, so impressions ≥ reach for the same content.

Algorithmic ranking signals generally fall into: predicted engagement, recency, relationship strength, content type match, and negative feedback (hide/report).

Hashtags aid discoverability and topic classification; optimal count varies by platform, and excessive tags can dilute topical relevance.

Asymmetric follow (one-way) supports broadcast reach; reciprocal follow supports conversation and community depth.

Dark social = sharing via private channels (DMs, email, copy-pasted links) that referrer data undercounts; expect under-attribution in analytics.

Platform-native formats (vertical short video, carousel, story, long-form text) outperform cross-posted repurposed assets because ranking favors native consumption patterns.

Video frame-rate/resolution quick reference: 1080p = 1920×1080; 4K UHD = 3840×2160; vertical short-form = 1080×1920 (9:16); square = 1:1; landscape = 16:9.

Aspect ratios: 9:16 stories/Reels/Shorts, 1:1 feed square, 4:5 portrait feed (higher screen real estate), 16:9 YouTube/landscape.

Safe zones: keep captions and key subjects out of the top ~14% and bottom ~20% of vertical video where platform UI overlays sit.

Three-point lighting = key (main, ~45° off-axis), fill (softens shadows, lower intensity), back/rim (separates subject from background).

Rule of thirds: divide frame into a 3×3 grid; place subjects on intersections for balanced composition.

Audio loudness target for social video is roughly −14 LUFS integrated; normalize before export to avoid platform re-compression pumping.

Export codecs: H.264/AVC for broad compatibility, H.265/HEVC for smaller files at equal quality; keep bitrate high enough to survive re-encoding.

Alt text describes image content for screen readers and is indexable; decorative images should be marked decorative rather than given filler alt text.

Response-time expectations: public complaints acknowledged quickly (often within hours), resolved asynchronously; private channels used for personal data.

Escalation ladder: acknowledge publicly → move to DM for details → route to specialist/legal/PR → document outcome. Never argue publicly.

Moderation triage categories: spam, harassment/hate, misinformation, off-topic, legal risk (copyright, privacy). Each has a distinct action path.

Community guidelines should state prohibited conduct, enforcement steps (warn, mute, remove, ban), and appeal route; publish them where members can find them.

Troll vs. critic distinction: critics raise legitimate issues and accept facts; trolls seek reaction and shift targets - respond to the former, disengage/limit the latter.

UGC rights: obtain explicit permission and credit before reposting user content; platform terms do not grant brands a license to a user's work.

Engagement prompts that work: specific questions, polls with a stance, member spotlights, and shared rituals (weekly threads) build repeat participation.

Crisis comms: single verified spokesperson, holding statement within the first hour, factual updates at fixed intervals, no speculation.

Scheduling tools queue posts via platform APIs; API rate limits and permission scopes determine which formats (e.g., stories, first comments) can be automated.

UTM parameters: utm_source (platform), utm_medium (paid/social/email), utm_campaign (initiative), utm_content (variant). Case-sensitive - standardize lowercase.

Asset workflow: brief → script/copy → shoot/design → edit → accessibility pass (captions, alt text, contrast) → approval → schedule → publish → measure.

Version control for creative assets: name files with date + version (e.g., 2025-03_reel_v03), keep a master folder separate from exports.

Caption files: SRT/VTT sidecar for video; burned-in captions survive reposts but are not editable or screen-reader friendly.

Automation vs. authenticity: auto-DMs and auto-replies should disclose automation where required and never impersonate a human response to a complaint.

Approval workflows use role-based permissions: creator drafts, editor reviews, approver publishes; audit logs record who changed what and when.

Backups: 3-2-1 rule - three copies, two media types, one off-site - applies to raw footage and brand asset libraries.

Vanity vs. actionable metrics: followers and likes are vanity-leaning; saves, shares, click-throughs, and conversions indicate intent and value.

Conversion rate = conversions ÷ clicks (or sessions) × 100. Cost per acquisition = spend ÷ conversions.

A/B testing requires one variable changed at a time and enough sample size; stopping a test early on a lucky result inflates false positives.

Attribution windows (e.g., 7-day click, 1-day view) change reported conversions - always state the window when comparing campaigns.

Sentiment analysis is directional, not precise: sarcasm, slang, and mixed-language posts cause misclassification; report confidence and sample size.

Privacy basics: collect only necessary data (minimization), state purpose, honor deletion requests, and never post identifiable individuals without consent.

Copyright: assume all found media is protected; use licensed, public-domain, or original assets, and keep license records. Fair use is a legal defense, not a blanket permission.

Digital citizenship: disclose sponsored content clearly, avoid astroturfing/fake reviews, and correct your own errors publicly rather than deleting criticism.

Funnel view: awareness (reach, impressions) → consideration (saves, shares, profile visits, video views) → conversion (clicks, leads, sales) → retention (repeat engagement, community return rate).

Engagement rate benchmark comparison is only valid within the same platform, format, and follower tier; cross-tier comparisons mislead.

Follower growth rate = (followers_end − followers_start) ÷ followers_start × 100 over the period; pair with churn to see net health.

Share of voice = your mentions ÷ total category mentions × 100; useful for competitive positioning, not for absolute performance.

UTM + platform analytics disagree because of cookie restrictions, in-app browsers, and privacy settings; treat platform numbers as relative, not absolute truth.

Dark patterns to avoid: confirmshaming, hidden unsubscribe, fake countdown timers, and pre-checked consent boxes - these are ethics and often legal violations.

Accessibility standard reference: WCAG contrast minimum 4.5:1 for normal text, 3:1 for large text; captions for all video with speech.

Data ethics test before publishing a metric: would the subject recognize and consent to how their data is used?

Reporting cadence: daily monitoring (spikes, crises), weekly tactical (post performance), monthly strategic (funnel, audience growth), quarterly review (channel mix, budget).

Metric definitions must be locked in a data dictionary so weekly and monthly reports stay comparable across teams.

Influencer disclosure rules generally require clear, prominent labels (e.g., #ad, paid partnership tag) placed before the fold, not buried in hashtags.

Consent and minors: content featuring children requires guardian consent; many platforms restrict monetization and messaging for minor accounts.

Misinformation response: correct with evidence, link primary sources, avoid amplifying the false claim in headlines or thumbnails.

Crisis metrics: track volume of negative mentions, response time, and sentiment shift; define thresholds that trigger escalation to legal/PR.

Audit trail: log moderation actions, deletions, and bans with reason codes so enforcement is consistent and defensible.

Retention schedule: keep raw analytics exports long enough for year-over-year comparison; delete personal data once its stated purpose ends.

Correlation vs. causation: a post going viral alongside a sales lift does not prove the post caused the lift; check controls and timing.

Sample bias: polls answered only by highly engaged followers overstate enthusiasm; note who is missing from the sample.

Benchmark hygiene: compare like-for-like periods (same season, same campaign phase) to avoid seasonal distortion.

Statistical significance: small accounts need larger relative lifts to distinguish signal from noise; report ranges, not single points.

Dashboard design: one primary question per dashboard; put the decision metric top-left and supporting context below.

Ethical A/B testing excludes experiments that deceive users or expose them to harm (e.g., hiding crisis information to test engagement).

Platform policy changes can break historical comparisons overnight; annotate metric discontinuities on trend charts.

Own your data: export raw records regularly because platform APIs and access tiers can change without notice.