New Coke — Coca-Cola's 1985 reformulation of its flagship cola

1. Episode summary

By the early 1980s, The Coca-Cola Company faced a decade-long competitive erosion in the US carbonated-soft-drink market. Pepsi-Cola had been running the "Pepsi Challenge" blind-taste-test campaign since 1975, and internal Coca-Cola tracking showed exclusive Coke drinkers falling from eighteen per cent in 1972 to twelve per cent in 1982 while exclusive Pepsi drinkers rose from four to eleven per cent. Supermarket-channel share, where Pepsi indexed strongest, was narrowing steadily. In 1984, under CEO Roberto Goizueta and President Donald Keough, the company launched a secret reformulation initiative, codenamed "Project Kansas," led by marketing executive Sergio Zyman and Coca-Cola USA president Brian Dyson. After roughly 190,000–200,000 blind taste tests at a reported cost exceeding four million dollars, a sweeter, smoother formula beat both original Coke and Pepsi in sip-format preference tests by a six-to-eight-point margin. On 23 April 1985 Goizueta announced at a Lincoln Center press conference that the ninety-nine-year-old formula would be discontinued and replaced by "New Coke." Consumer reaction was immediate and severe: the company logged up to 8,000 calls per day, roughly 40,000 complaint letters, and grass-roots protest organisations including Gay Mullins' Old Cola Drinkers of America. On 11 July 1985, seventy-nine days after launch, Keough and Goizueta announced the original formula would return as "Coca-Cola Classic" alongside New Coke. The strategic question the episode turned on: did the research apparatus supporting the reformulation decision measure the dimension of consumer response that actually drove the market outcome? Scope of the case. The case scores the New Coke launch decision (April 1985) and its public-market response through the 11 July 1985 reversal — a bounded 79-day decision-and-consequence arc. Coca-Cola's recovery, the Coca-Cola Classic reintroduction, and year-end 1985 market-share movements are acknowledged in §3 as the execution context that enabled the reversal, but they are outside the scoring scope; the failure under analysis is the launch decision itself, not the company's longer-run market position.

2. Sources

Primary:

1. Roberto C. Goizueta and Donald R. Keough, Coca-Cola Company press conference transcripts and public statements, 23 April 1985 (New Coke launch, Lincoln Center, New York) and 11 July 1985 (Coca-Cola Classic reintroduction announcement, Atlanta). Keough's widely-reported remark "our boss is the consumer" and his later statement that consumer research "could not measure or reveal the depth and abiding emotional attachment to original Coca-Cola" are contemporaneously quoted in wire-service and newspaper coverage.
2. The Coca-Cola Company, Annual Report for fiscal year 1985, Atlanta, GA. Describes the new formulation as earning "superior preference ratings in consumer taste tests" with "special appeal to the youth segment," and reports the July 1985 decision to return the original formula to market under the Classic designation. Archived in the Internet Archive Coca-Cola annual-reports collection.
3. United Press International (UPI) archive wire reports, April–July 1985, including the 24 July 1985 UPI Archives feature on Gay Mullins and the Old Cola Drinkers of America organising drive. Contemporaneous reporting of consumer-response volume, bottler reaction, and protest-group activity.
4. Chicago Tribune contemporaneous reporting, spring–summer 1985, on regional reaction in the US South (including the University of Mississippi academic quoted on Coca-Cola as regional identity artefact), cited in multiple secondary syntheses.

Secondary (with justification):

1. Constance L. Hays, The Real Thing: Truth and Power at the Coca-Cola Company, Random House, 2004. Journalist's long-form investigative history of the company drawing on interviews with participants; covers Project Kansas, the Goizueta-Keough decision chain, and the reversal. Used here for reconstruction of internal decision sequence.
2. Thomas Oliver, The Real Coke, The Real Story, Random House, 1986. Contemporaneous book-length account by an Atlanta Journal-Constitution business reporter, published within a year of the episode, drawing on executive interviews; the standard journalistic reference for the research methodology and internal debate.
3. Harvard Business School, "Introducing New Coke" (HBS case 500067), John Deighton, 2005 (revised). Teaching-case synthesis of the taste-test methodology, the focus-group dissent signal that was not replicated in individual surveys, and the branding and launch-communications decisions.
4. Robert M. Schindler, "The Real Lesson of New Coke: The Value of Focus Groups for Predicting the Effects of Social Influence," Marketing Research, December 1992. Peer-reviewed analysis of the research-design gap between sip-preference taste tests and whole-brand-consumption emotional attachment.

Tertiary (flagged):

1. Britannica, "New Coke," encyclopaedia entry; Wikipedia, "New Coke." Used for frame-level cross-checking of dates, participant names, and quoted statements against primary and secondary sources above, not for load-bearing claims.

3. OTA narrative

Observe. The observation apparatus was the most extensively resourced segment of the decision chain. The company ran approximately 190,000–200,000 blind taste tests over roughly two years at a cost exceeding four million dollars, and reran the 100,000-subject first wave with a second independent research firm to verify the preference result. The apparatus produced two distinct signals. The sip-test signal was clear and robust: the reformulated product beat both original Coke and Pepsi by six to eight percentage points on blind preference. A second, qualitatively different signal surfaced in focus-group work: when told the original formula would be withdrawn, a subset of participants reacted with anger disproportionate to the taste delta, and this anger did not reappear in individual quantitative surveys asking the same question. Both signals were visible to the committee. The sip-preference signal was treated as the dispositive measurement; the focus-group emotional-attachment signal was treated as non-generalisable and not escalated. Observe is a root-cause phase in this episode. The observation task was hard at the design end — measuring what is effectively brand-identification emotional attachment against a single-sip palate test is a non-trivial research-instrumentation problem — and the company's instruments resolved the palate dimension cleanly but did not resolve the identity dimension. Observe is classified Almost-wrong at the hard end of the task-difficulty axis: the identity-attachment signal was captured in one sub-instrument and then written off rather than pursued.

Think. The reasoning step moved from "we have a better-tasting sweeter formula that beats Pepsi in sip tests" to "launching this formula in place of the original will close the Pepsi gap in the measured market." That inference required the prevailing peer-group framework in consumer packaged goods circa 1984–85: product-preference research as the primary predictor of market-share response, with brand-equity considerations secondary. Under that framework the reasoning was internally consistent. What the framework did not have priced-in was the asymmetry between product-as-beverage and product-as-cultural-anchor for a flagship brand with a ninety-nine-year continuous identity and strong regional, generational, and iconographic attachment. The correct framework — that brand-identification effects can dominate sip-preference effects for iconic consumer brands — existed in marketing academia at the time but was not the peer-group default for beverage-category reformulation decisions. Goizueta's decision to position the new product as replacement rather than addition, and the simultaneous discontinuation of the original formula, magnified the reasoning error. Think carried the transmission weight between a partially-captured observation and the launch that followed, but is not the operative root cause: the interpretive problem is better understood as a failure to pursue the dissenting observation than as a failure of reasoning from the observation actually generalised. Think was not a root cause — it was the transmission step between an Observe that captured only the palate dimension and an Act that implemented on the captured dimension.

Act. Execution of the launch itself was operationally competent within its frame. Bottling, distribution, packaging changeover, the "New!" label Goizueta personally insisted on, and the advertising campaign were delivered on schedule across the US market. The reversal was also executed quickly and competently: within seventy-nine days of launch, Coca-Cola Classic was back on shelves, bottlers ordered twenty-four million cases within forty-eight hours of the reintroduction announcement, and the company contained the damage without losing its distribution or bottler network. The press conference communications on 23 April — Goizueta describing the new flavour as "bolder, rounder, and more harmonious" and refusing to admit taste tests had driven the decision, calling it "one of the easiest decisions we've ever made" — have been widely criticised as tonally misjudged, but the misjudgement is downstream of the underlying observation failure rather than an independent execution error. Act was not the root cause; execution delivered, competently, what the preceding phases had specified, and the same execution capacity enabled the rapid reversal within the case's scoring scope (Coca-Cola Classic reintroduced on shelves within forty-eight hours of the 11 July 1985 announcement). Coca-Cola's longer-run recovery, including the post-scope-window market-share movements through year-end 1985, is execution context for the reversal rather than a re-classification of the launch episode itself.

4. Modality evidence

Direction.

The strategic choice that set the episode's trajectory was the decision to retire the original Coca-Cola formula entirely rather than introduce the reformulated product as an addition to the line. This was a specific, dated, and attributable decision: on 23 April 1985, CEO Roberto Goizueta and President Donald Keough announced at a Lincoln Center press conference that the ninety-nine-year-old formula would be discontinued and replaced. The reformulation-as-replacement posture — rather than a parallel-launch model — was a deliberate directional call that Goizueta and Keough had considered and affirmatively chosen; the company had internally weighed a dual-product approach and rejected it, partly on the grounds of bottler capacity and shelf-space concerns (Hays, The Real Thing; Oliver, The Real Coke, The Real Story). Goizueta's insistence on the "New!" label, which visibly broke from the brand's continuity signals, was a downstream execution of this directional choice rather than a separate error — it made the replacement posture legible to consumers.

The direction set by the 1984–85 Project Kansas initiative was also framed against a specific competitive threat: the Pepsi Challenge had been running since 1975 and internal tracking showed exclusive Coke drinkers falling from 18 per cent in 1972 to 12 per cent in 1982, while exclusive Pepsi drinkers rose from four to eleven per cent (Coca-Cola 1985 Annual Report; Hays). The reformulation was chosen as the response to that erosion — a directional bet that taste parity with or superiority over Pepsi in blind tests would recover market share. That frame closed off the alternative direction of competing on brand-identity grounds rather than palate grounds, which would have pointed toward a different competitive strategy. The Direction contribution to this episode is the specific, attributable, dated choice to position the reformulated product as a full replacement and to compete on taste-test terms with Pepsi.

Structure.

Decision authority over Project Kansas was concentrated within a small executive group: Goizueta and Keough held final authority, with Sergio Zyman (marketing vice president) and Brian Dyson (Coca-Cola USA president) leading the research initiative (Oliver, The Real Coke, The Real Story; HBS case "Introducing New Coke"). The research committee that designed and interpreted Project Kansas was structurally the same group that would recommend the launch decision. There was no independent review body charged with challenging the research conclusions before they reached the executive level — the funnel from research design through research interpretation to executive recommendation ran through the same programme team. This arrangement meant that the methodological assumption embedded in the research design (sip-preference as the operative consumer measure) was not independently challenged at any structural gate before the launch decision was taken.

The bottler feedback channel was structurally present but activated after rather than before the consumer launch. Bottlers were briefed at an April 22 meeting at the Woodruff Arts Center, where initial support was expressed despite mixed taste reactions among some attendees; by June 23, a formal delegation of bottlers — particularly from the South — had convened a private complaint meeting at Coca-Cola headquarters in Atlanta demanding a return to the original formula (Hays, The Real Thing; Oliver). The structural point is that the bottler channel, which represented the closest institutional proxy for consumer and regional sentiment, was consulted as an implementation briefing rather than as a pre-decision input. The sequence placed bottler concerns inside the launch cycle rather than upstream of the launch decision.

Processes.

The research process that drove the launch decision was the most extensively resourced element of the episode, but its design contained a structural limitation that the process did not surface or correct. Approximately 190,000–200,000 blind taste tests were conducted across roughly two years at a cost exceeding four million dollars; a second independent research firm reran the first wave to verify the sip-preference result (Coca-Cola 1985 Annual Report; Oliver, The Real Coke, The Real Story). The sip-preference measurement was rigorous within its instrument. The limitation was that the survey instrument did not include the question of how respondents would feel if the reformulated product replaced the original — the replacement condition that would obtain in the market was not tested during the large-scale quantitative phase (Schindler, "The Real Lesson of New Coke," Marketing Research, 1992; HBS case "Introducing New Coke").

The focus-group phase surfaced the emotional-attachment signal: when told the original formula would be withdrawn, approximately 10–12 per cent of focus-group participants reacted with anger disproportionate to the taste delta, and their presence tended to negatively skew results as they exerted peer pressure on other participants (web-search corroboration: Soda Encyclopedia / New Coke article; confirmed in Schindler analysis). The process for handling this signal did not route it back into a modified quantitative instrument. Instead, the focus-group dissent was characterised as non-generalisable — a vocal minority whose reaction was treated as idiosyncratic rather than indicative — and the quantitative sip-preference result was allowed to stand as the operative finding (Oliver; HBS case). The process break was not a failure of data collection but of data integration: the process had no mechanism to reconcile a qualitative signal that contradicted the quantitative finding; the reconciliation defaulted to the quantitative result by convention rather than by deliberate adjudication.

Capability.

The Coca-Cola research team held deep capability in beverage-preference measurement; the scale, replication discipline, and statistical rigour of the Project Kansas taste tests were at the leading edge of consumer-goods research practice for the mid-1980s peer group (Oliver, The Real Coke, The Real Story; HBS case "Introducing New Coke"). The capability gap the episode surfaces is narrower and more specific: the tools for measuring emotional brand attachment as a distinct and potentially dominant consumer variable — separable from sip-preference — existed in marketing academia by the mid-1980s but had not been adopted as a standard instrument in beverage-category reformulation research. Schindler's 1992 post-mortem in Marketing Research identifies the research design gap as a known limitation of sip-test methodology for high-identity brands, and frames it as a lesson the episode made legible precisely because the capability to measure the attachment dimension was not deployed (Schindler, 1992).

This is a capability gap at the level of the peer group, not just Coca-Cola individually. No beverage-category peer had developed a validated brand-identity attachment instrument as a standard element of reformulation research in this period; the sip-test / focus-group combination was the industry default. The capability gap is therefore one of institutional knowledge that the situation required but the research profession had not yet codified as a deployable tool. The evidence does not support a finding that a readily available measurement capability was simply ignored; it supports a finding that the capability did not exist in ready-to-deploy form for the specific use case of brand-identity measurement in a high-stakes reformulation context.

Culture.

Sergio Zyman, as vice president of marketing and the lead architect of Project Kansas, was a strong advocate for quantitative data-driven decision-making — described in contemporaneous and retrospective accounts as highly confident in what the taste-test numbers indicated, and as pushing for the complete replacement strategy rather than a parallel launch (Oliver, The Real Coke, The Real Story; Hays, The Real Thing). Zyman's advocacy created a cultural gravity within the project team toward the quantitative result and toward the most decisive implementation option. In this environment, the focus-group dissent — which pointed in the opposite direction — faced a cultural headwind: to surface it as a decision-relevant signal required overriding the consensus that had formed around the taste-test data and the replacement posture.

Goizueta himself has been characterised as cerebral and introverted, with face-to-face confrontation described as "not his strong suit" (UPI Archives, August 1982 profile; encyclopedia.com biography). This leadership-style description is consistent with an internal environment in which discomfort with the dominant interpretation of the research was unlikely to surface through direct challenge. The cultural condition that the methodology distinguishes from structural blockage is the won't rather than the can't: the formal processes for dissent were not demonstrably blocked — the research team and bottlers had access to leadership — but the normative environment created by Zyman's data-confidence, Goizueta's aversion to confrontation, and the shared competitive urgency around the Pepsi gap suppressed the signalling of a dissenting interpretation before the launch decision was taken (Hays, The Real Thing; Oliver, The Real Coke, The Real Story). The bottler June 23 complaint meeting, which occurred nine weeks after launch and six weeks before the Classic reintroduction, illustrates that the dissenting institutional voice did eventually reach leadership — but through a post-launch channel, after the decision surface had shifted from "launch or not" to "recover or not." The cultural norms present during Project Kansas produced the same outcome as suppression without requiring formal suppression mechanisms.

---