Spring Oats is grounded in five decades of peer-reviewed social-network analysis and media economics — 39 papers across seven pillars, from the strength of weak ties to the localization of global music. Every card links to its original source. We host nothing; we cite everything.
Novel information crosses social networks through bridging ties between clusters — not through dense hubs inside one community.
Weak ties — acquaintances, loose connections — are the bridges that carry novel information between social clusters; strong-tie groups mostly recirculate what they already know.
Why it matters: The founding case for seeding bridging creators who connect communities, not creators buried inside one you already reach.
View source ↗A ten-year reassessment: weak ties help most when they act as bridges and when the recipient can act on the new information — their value is conditional, not automatic.
Why it matters: Tempers the seeding logic — bridge position matters, but so does fit and the ability to act.
View source ↗Brokers who span 'structural holes' — gaps between groups that don't otherwise connect — gain early, non-redundant information and control advantages. Network position is the asset.
Why it matters: The precise name for the creators we target: brokers across structural holes.
View source ↗Inside a large firm, managers whose networks bridged structural holes produced ideas rated more valuable, were paid more and promoted more — bridging causes access to better information.
Why it matters: Evidence that bridge position pays off in practice, not just theory.
View source ↗A 20-million-person, 5-year LinkedIn experiment proved weak ties causally drive opportunity — with an inverted-U: moderately weak ties win, the very strongest and very weakest do not.
Why it matters: Tells us to target moderate-strength bridge creators with real scene credibility, not random strangers.
View source ↗Big cascades are driven by a critical mass of easily-influenced people, not by a special few mega-influencers.
Across many simulations, large cascades are driven not by rare 'influentials' but by a critical mass of easily-influenced people; whether the network can cascade matters more than who starts it.
Why it matters: The academic case against blowing budget on one mega-influencer — build the conditions for a cascade instead.
View source ↗The original 'two-step flow': media reaches ordinary opinion leaders who relay and reshape it for their peers. Opinion leaders are everyday peers, not celebrities.
Why it matters: The 70-year-old root of micro-influencer marketing — peers move people.
View source ↗An in-product experiment separating influence from susceptibility: influential people cluster together and are themselves harder to influence — so concentrating spend on clustered big nodes hits low-susceptibility audiences.
Why it matters: Mechanistic support for seeding the susceptible periphery, not the clustered top.
View source ↗A word-of-mouth model showing the sheer abundance of weak ties lets them rival strong ties in driving adoption, even though each weak tie is individually less persuasive.
Why it matters: Bridges network theory into marketing — abundance beats individual strength.
View source ↗Adopting a song needs reinforcement from several peers at once — so wide, multi-source bridges beat long thin ones.
Behaviours that need social reinforcement from multiple sources (a 'complex contagion') spread on the WIDTH of bridges, not their length — long thin ties can even block them.
Why it matters: The core reason micro-seeding works: a song's adoption needs several credible peers, so we build wide bridges, not just far-reaching ones.
View source ↗In a controlled experiment a behaviour spread to more people and ~4x faster through clustered, redundant networks than through random 'small-world' ones; a second reinforcing signal sharply lifted adoption.
Why it matters: Hard proof that reinforcement beats raw reach — seed clusters where a listener sees the track from several sides.
View source ↗People adopt once enough others have; the outcome hinges on the whole distribution of individual thresholds, so a few low-threshold instigators decide whether something tips or stalls.
Why it matters: The formal logic of momentum — micro-seeds are the low-threshold early adopters that start the chain.
View source ↗A book-length synthesis of complex-contagion research: wide bridges, network redundancy, and how to engineer behaviour change rather than just awareness.
Why it matters: The single readable source for the simple-vs-complex distinction at the heart of our method.
View source ↗Who you seed first can swing outcomes by multiples; well-connected and bridging nodes are the highest-leverage starting points.
Across field experiments and a 200,000-customer campaign, the best seeding strategy was up to 8x more effective than the worst; seeding hubs and bridges beat random and fringe — because they participate more and reach further.
Why it matters: Direct evidence that who you seed first drives multiples of outcome — the reason Spring Oats exists.
View source ↗Formalises 'pick the k seeds that maximise spread'; introduces the Independent Cascade and Linear Threshold models and a greedy algorithm with a provable ~63% optimality guarantee.
Why it matters: The math under any engine that ranks which creators to seed for maximum reach.
View source ↗Across 163 campaigns, measurable knock-on conversions across audience groups appeared only among micro-influencers — not macro ones.
Why it matters: The most on-point recent evidence that micro-creators, not macro, generate cross-cluster spillover.
View source ↗Early engagement velocity feeds recommender feedback loops — manufacturing the momentum a new release needs to be distributed widely.
A 253-million-user Facebook experiment: feed exposure made people 7.4x likelier to share; strong ties are individually more influential, but the far more numerous weak ties carry most of the spread and the novel information.
Why it matters: The online proof that abundant weak-tie micro-seeds win the awareness stage.
View source ↗A 1.3-million-user music experiment: listening 'caught on' to a fan's friends only for already-popular artists, and the effect grew with the number of friends involved — emerging artists got no organic spillover.
Why it matters: The exact gap Spring Oats fills — we manufacture the multi-friend reinforcement that organic contagion gives only to stars.
View source ↗Adopting streaming causes large, lasting increases in both how much and how diverse a listener's music becomes, and boosts discovery of new artists.
Why it matters: Grounds the discovery surface that seeding-driven momentum pays off on. (Open access.)
View source ↗Causal estimates of playlist placement: a slot on 'Today's Top Hits' was worth roughly 19.4 million streams on average — placement, not just quality, drives outcomes.
Why it matters: Quantifies the prize our seeding aims to unlock: cracking algorithmic/editorial promotion for a release.
View source ↗Recommender systems can reduce overall diversity, concentrating demand onto already-popular items — a rich-get-richer dynamic that runs against the 'long tail' hope.
Why it matters: The bias a new release must overcome — and that early seeding is designed to flip.
View source ↗When recommenders train on data they themselves shaped, the feedback loop homogenises users and erodes usefulness over successive cycles.
Why it matters: A mechanism behind 'priming the algorithm' — early signals compound through the loop.
View source ↗Simulations show the recommender feedback loop amplifies popularity bias over time — popular items get more popular, diversity falls.
Why it matters: Empirical basis for 'early engagement leads to more exposure leads to more engagement'.
View source ↗A comprehensive map of recommender biases — selection, exposure, popularity, conformity — and the methods used to counter them.
Why it matters: The one-stop survey for citing the popularity-bias / feedback-loop literature rigorously.
View source ↗A three-year event study of playlist inclusion: proportional gains are similar for emerging and established artists, but established ones gain far more in absolute terms, with snowball/path-dependence in distribution.
Why it matters: Direct evidence on emerging-vs-established placement and momentum — our novel zone. (Working paper; confirm version.)
View source ↗The formal tools — centrality, community detection — that turn 'find the right creators' into something computable.
A survey of the social-network-analysis toolkit: degree, betweenness and eigenvector centrality; community detection; temporal and ML methods; visualisation.
Why it matters: Names the exact metrics our targeting uses — betweenness for bridges, community detection for clusters. (Open access.)
View source ↗Frames influencers as high-centrality hubs, algorithms as 'super-edges', and feed exposure as the transmission mechanism, recommending a hybrid 'test on TikTok, refine on Instagram' approach.
Why it matters: Independently mirrors the Spring Oats thesis, platform by platform. (Open access.)
View source ↗The formal definition of betweenness centrality — how much a node sits on the shortest paths between others, capturing its power to control flow between clusters.
Why it matters: The metric, and its citation, for 'a creator who bridges communities'.
View source ↗An efficient method for detecting the natural communities (clusters) inside a network via 'modularity' maximisation.
Why it matters: Operationalises the clusters our seeds are meant to bridge. (Open access.)
View source ↗An influence-weighted centrality: a node counts as central in proportion to how central its neighbours are.
Why it matters: An alternative ranking if degree and betweenness aren't enough — a creator connected to influential creators scores higher.
View source ↗The canonical theory of how new things spread: the S-curve and adopter categories — innovators, early adopters, majorities — shaped by opinion leadership.
Why it matters: Frames a release as an innovation; our micro-seeds are the early adopters who legitimise it for the majority.
View source ↗Audiences are going local; 'going global' is now trans-local bridging across national/linguistic clusters — exactly what seeding engineers.
Audiences prefer culturally proximate content — local stars, language and references — over imported media whenever credible local options exist.
Why it matters: The demand-side root of why markets cluster locally, and why bridges must respect cultural proximity.
View source ↗Maps media flows beyond the US-to-world model — including subaltern (periphery-to-core) and diasporic flows that carry non-Western content across markets.
Why it matters: Names the diasporic flow Spring Oats rides to bridge between national clusters.
View source ↗Globalising and localising forces are simultaneous: the global is realised through local adaptation, and local cultures actively shape what becomes 'global'.
Why it matters: The frame that dissolves the paradox — global platforms carrying local content is the expected state, not a contradiction.
View source ↗Introduces the 'cultural discount': content rooted in one culture loses value in another because audiences can't identify with its style, language and references.
Why it matters: Names the barrier our bridging — and diaspora seeding — is built to lower.
View source ↗An empirical refutation of the 'long tail': digital amplified blockbusters rather than diluting them; the niche tail stays 'long and lonely'.
Why it matters: Half of the reconciliation — concentration intensifies, and combined with localization it produces local superstars.
View source ↗Documents music's winner-take-all economics: the top 1% of performers' share of concert revenue rose from 26% (1982) to ~60%, with the top 5% taking ~85%.
Why it matters: Quantifies the concentration that, localized, becomes 'local winner-take-all' — the gap our cross-cluster seeding arbitrages.
View source ↗A network analysis of Spotify's related-artist graph found international-label acts get cross-border, genre-based recommendations while local-label and independent acts are paired by country of origin. (Case study; treat as a structural tendency.)
Why it matters: Evidence that streaming graphs cluster independents at home, leaving few cross-cluster pathways — the gap we supply.
View source ↗A reproducibility study showing whether recommenders favour or suppress local music flips with the dataset and model settings, and that 'local music' is hard to even label — so the literature is unsettled.
Why it matters: The honest caveat we keep next to the claim above — a structural tendency, not a proven law.
View source ↗