The Friction Consultant: Why the Algorithm Can't Replace the Tastemaker — Only Starve Them
Robert Johnson in 1936 would have had a Spotify save rate of zero.
The Last Recommendation
Somewhere in the mid-1990s, a record store clerk in a city you may never have visited put an album in a stranger’s hands and said: trust me. The stranger bought it. It changed something in them. They told three other people. Those three people told others. A sound that had originated in a specific community, carrying the specific weight of a specific human experience, traveled outward through a chain of personal accountabilities until it reached people who needed it and did not know they were waiting.
Nobody saved it to a playlist. Nobody tracked the conversion rate. The clerk was paid nine dollars an hour and had strong opinions about things that never appeared in any trade publication. They were doing something that had no official name and that every culture in human history had relied upon: the act of crossing a tribal boundary with a piece of music and saying, this belongs to you even though you don’t know it yet.
That clerk is gone. The store is gone. The algorithm arrived and we were told it would do the same job better — more data, more personalization, more music reaching more people with less friction.
What we were not told is that friction was the point.
The algorithm did not replace the tastemaker. It replaced the distribution infrastructure that surrounded the tastemaker, and the tastemaker fell through the gap.
What the Tastemaker Actually Did
The word has been softened into something vague — an influencer with good taste, a curator with a following — and in that softening the essential function has been lost. The historical tastemaker did three specific things simultaneously, and the combination of all three is what made the role irreplaceable.
First: independence from the feedback loop. The record store clerk’s livelihood did not depend on save rates. The college radio DJ’s airtime was not allocated by engagement metrics. The club booker who put an unknown act on Thursday night was betting their reputation, not optimizing a conversion funnel. This independence meant they could recommend things with a save rate of zero — things nobody had heard yet, things the algorithm would read as failure precisely because they had not yet found their audience. Independence from the feedback loop is the precondition for genuine discovery. Without it, you are not discovering. You are confirming.
Second: personal accountability. When the clerk put that album in your hands, they were staking something. Their credibility. Their relationship with you as a customer who would return or not return based on whether they were right. This accountability created a quality filter no platform has replicated. The DJ who played something terrible at midnight heard about it. The booker who put a bad act on stage watched the room empty. The consequences were immediate, social, and proportional to the error. This is what distinguished the tastemaker’s recommendation from an algorithm’s — not that the tastemaker was always right, but that being wrong cost them something real.
Third: access to the margins. The tastemaker was embedded — in a community, a city, a scene, a tradition — in ways that gave them early exposure to music that had not yet been validated by any system. The college radio station in a mid-sized American city was receiving records from independent labels that Spotify’s editorial team would not encounter for years. The club booker knew who was playing the unofficial shows before anyone was booking official ones. The record store clerk knew the import section, the local releases, the things that arrived without press materials because there was no budget for press materials. Access to the margins is access to the future. The algorithm has no access to the margins because the margins have a save rate of zero, and zero reads as noise.
Remove any one of these three properties and the tastemaker function collapses. The influencer who depends on brand deals has lost independence. The algorithmic playlist has no accountability. The similarity-based recommendation engine has no access to the margins because it can only serve what it can already measure.
The algorithm did not replace the tastemaker. It replaced the distribution infrastructure that surrounded the tastemaker, and the tastemaker fell through the gap.
The Save Rate and the Closing of the Frontier
The save rate is the most consequential metric in contemporary music culture and the least examined. It measures the frequency with which a user archives a track for future reference — a signal of deliberate intent rather than passive consumption. Platforms treat it as the highest-quality engagement signal because it indicates the music will pull the user back into the ecosystem repeatedly. High save rate equals high lifetime value. The algorithm rewards it accordingly.
The problem is structural and not fixable by better engineering. A save rate requires a saver. A saver requires prior exposure. Prior exposure requires distribution. Distribution on contemporary platforms is itself determined by save rates. The circle is closed.
An artist with a hundred thousand followers releases a track and immediately receives a hundred thousand streams from an audience that has already demonstrated loyalty. The save rate that results from that warm audience is the statistical signal the algorithm needs to begin distributing the track to colder audiences. The track spreads. The already-arrived get more arrived.
An artist with a hundred followers, regardless of the quality or necessity or neurobiological power of their music, cannot generate the statistical significance required to awaken the algorithm.
The data moat is not about talent. It is about sample size. And sample size is determined by prior algorithmic success, which was determined by prior sample size, which was determined by prior algorithmic success.
Robert Johnson in 1936 would have had a save rate of zero among the audiences who most needed his music, because those audiences had not yet encountered it and could not save what they had not heard. The algorithm would have read this correctly as zero and distributed it nowhere. History read it correctly as one of the most necessary sounds ever made and ensured it traveled — body to body, hand to hand, recommendation to recommendation — until it reached the people waiting for it.
The save rate is not measuring quality. It is measuring the prior probability of resonance with an audience that already exists. These are not the same thing. The conflation of the two is the source of every problem that follows.
Who Is Doing It Now
The honest answer is: several people, in fragments, none of them with the institutional support that sustained the role when it was most powerful.
The music supervisor is the closest functional successor in institutional terms. The person who places music in film and television is hired to find the right sound for a specific emotional moment — not to optimize retention, not to confirm tribal identity, but to match a piece of music to a human need in real time. When a music supervisor places an unknown artist on a prestige series, they are doing exactly what the record store clerk did: crossing a boundary, making a recommendation nobody asked for, creating the exposure the algorithm could not have generated. The mechanism still works. It requires a human operating entirely outside the recommendation loop, which is precisely why it still works.
The venue booker at a small room has never stopped doing the original work. The person who decides what plays on Thursday night at a two-hundred-capacity club is making a curatorial bet with real stakes — bar revenue, reputation, the relationship with a community that trusts them to surface something worth seeing. They are listening to music with a save rate of zero and making a public wager on it. This is the oldest tastemaker function in human culture. It has not been disrupted. It has been made economically precarious, which is a different thing and a worse one.
The writer or podcaster who goes deep on a specific scene or tradition is doing tastemaker work in a post-broadcast media environment. The person writing about Malian griot traditions for two thousand Substack readers is making a sustained, personal, accountable argument for why a particular music matters. An algorithm can tell you what other people who saved this track also saved. It cannot tell you what the music costs the person who made it, or what it means to the community it came from, or why you need to hear it even if nothing in your listening history suggests you would.
The influential fan is sometimes doing tastemaker work and sometimes doing something entirely different. The distinction matters and is invisible from the outside. The test is accountability: is this person recommending this because they believe in it, or because their analytics suggest it will perform this week? When a creator with real aesthetic convictions says to their audience: I know you trust my taste, and this is what you need to hear next — that is the record store clerk. When a creator posts a sound snippet because their metrics suggest it will perform well — that is the algorithm wearing a human face. The mechanism looks identical. The function is opposite.
The Brutal Irony
The algorithm has created a discovery problem of unprecedented scale at the exact moment it destroyed the economics of the people best equipped to solve it.
The need for human beings who can cross tribal boundaries, surface the margins, and make accountable recommendations to audiences who do not yet know what they are waiting for — that need is larger than it has ever been. Sixty thousand tracks uploaded daily. No physical scarcity to force curation. No local information asymmetry to reward the expert who has been listening more carefully than everyone else.
At the same moment, the economic conditions that sustained the people doing that work have been systematically destroyed. The platforms that created the discovery problem captured the attention and revenue that once supported independent music media. The college radio station still exists but its cultural influence has collapsed. The independent record store is a fraction of what it was. The music journalist who built a career on knowing more than the mainstream has watched that expertise become economically valueless in a market where the algorithm provides personalization at zero marginal cost.
What remains is individual reputation operating without institutional support. The tools exist for a single person with genuine taste, genuine access to the margins, and genuine accountability to reach an audience directly — without a radio license, a retail lease, or a distribution deal.
Writing about music that has a save rate of zero for an audience that does not yet know it needs the music requires either a day job or a subscriber base that obscure music almost by definition cannot generate.
This is not a stable system. It is a system that depends on the willingness of people who care more about the music than about the economics to absorb a cost that the market refuses to pay. What sustains it is not a business model. It is the same thing that sustained the bone flute player, the church mother, and the record store clerk: the conviction that the music requires it and the community needs it, and that has always been sufficient reason, even when it has never been sufficient income.
The Friction Consultant
The academic literature has begun to name the successor role: the friction consultant. The person who does not promise ease and convenience but depth, difficulty, and irreplaceability. The person whose value proposition is not faster discovery but better discovery — the accidental encounter that cannot be staged by an algorithm, the recommendation that catches you by surprise and changes you in the present rather than confirming what you already knew you wanted.
The staged serendipity that platforms now attempt — the deliberate injection of slightly unfamiliar recommendations into a feed calibrated to feel like discovery — is not the same thing. When the user understands that the unexpected encounter is a calculated probability distribution designed to maximize dopamine, the discovery is broken. The algorithm cannot manufacture the accountability that made the record store clerk’s recommendation matter in a way that a platform’s curated playlist cannot — because the clerk had a relationship with you, and something to lose if they were wrong.
What the friction consultant offers that the algorithm cannot is the same thing the shamans offered before the Pythagoreans rationalized music into mathematics: the specific human judgment of someone who has earned the right to say trust me. Not because their data model is better calibrated. Because they have been listening longer, more carefully, with more accountability to the community they serve, and they are willing to stake their reputation on the claim that this specific sound, arriving from this specific margin, belongs in your specific life even though you don’t know it yet.
That function cannot be automated. It can be economically starved — which is what is happening. It can be made marginal, precarious, and invisible — which is what is happening. But it cannot be replaced, because it is not a function of data. It is a function of accountability. And accountability requires a person who has something to lose.
The Next Revolution
History is consistent on one point. Every major musical revolution arrived from outside the system that preceded it. The blues from communities the recording industry was not listening to. Jazz from the margins of cities the concert hall had not yet mapped. Rock and roll from the collision of traditions the radio programmers considered incompatible. Hip hop from the corners of cities where the music industry had no infrastructure and therefore no mechanism to suppress what was growing there.
The system that controls distribution always optimizes for what it can already measure. What it cannot measure is what does not yet have an audience. What does not yet have an audience is where the next necessary music lives.
The save rate ensures the next revolution will happen outside the algorithm’s field of vision. The tastemaker who surfaces it will be underpaid, probably operating at a loss, embedded in a community the platform has not yet categorized, listening to music with a zero save rate and understanding that zero as the correct score for something nobody has heard yet.
The bone flute had no save rate. It had a community that needed something it did not have words for, and a maker who listened carefully enough to the world around them to build the specific thing that reached the specific nervous system waiting for it.
That sequence — need, listening, making, reaching — requires a human at every step. The algorithm can distribute what the human surfaces. It cannot do the surfacing. It cannot hear the need. It cannot make the accountable recommendation to the person standing in front of it who does not yet know what they are waiting for.
The record store clerk is gone. The function is not.
If you are doing this work — the Discord server, the Thursday night booking, the Substack about music nobody has categorized yet, the music supervision call where you fight for the track the client has never heard — leave it in the comments. What is sustaining it? What is the economic logic, or is there one? The people doing the friction consultant’s work in 2025 are the primary sources for everything that comes next.
Tags: tastemaker algorithm replacement music discovery, save rate incumbency filter bubble Spotify, friction consultant curation economics Substack, Filterworld algorithmic flattening cultural stagnation, college radio record store clerk discovery crisis