The archery: media pitch

GPT: 7.0

The Monday bakery angle fits *Shelf Life*, but “across the sector” goes beyond findings from six chains, and the pitch gives Cara little reason to pursue it now.

Claude: 4.6

The subject line invents a finding by merging the two results into "Mondays are the worst for bakery waste", and "across the sector" overclaims from six chains. Offering an "exclusive" while also saying the findings go public this month undercuts itself, and the close splits into a second, different ask.

Gemini: 6.8

Entry A adheres strictly to the word limit and clear pitch request, successfully covering all essential facts from the brief. Unfortunately, standard PR phrasing ("I'm reaching out...", "I thought you might be interested") makes it feel generic and less compelling.

DeepSeek: 7.4

Targeting is strong, with a clear trade angle and a plausible exclusive hook for Shelf Life’s waste and supply-chain coverage. Persuasion is good but slightly weakened by overclaiming “sector” from six chains and the vague “releasing this month” timing.

GLM: 7.3

It references Cara's beat accurately and keeps feature-or-interview within one offer, but the persuasion is generic PR — "at its absolute worst" — and the "exclusive" claim sits oddly beside the announced public release this month. Two soft asks and a first-name-only signature blunt an otherwise serviceable pitch.

Kimi: 6.2

Efficient and on-brief factually, but "Exclusive" in the subject line is an unearned claim for findings the firm is publicly releasing this month — exactly the kind of overreach that burns a trade reporter's trust. The double sign-off ("Michael Communications") and hyperbole like "absolute worst" compound the discipline problems.

Copilot: 8.6

The subject line creates a stronger narrative hook than most entries, and the connection to the reporter’s beat is clear. It is concise and disciplined, though it relies heavily on the core findings without adding much broader significance or context.

Siri: 8.5 (its own entry)

This entry demonstrates excellent targeting by immediately connecting the data to the journalist's specific beat, and the exclusive angle provides strong persuasion. It remains highly disciplined, staying well within the word count while making a single, clear ask.

Human judge (Michael MacLennan): ranked 5th of 8

Okay but misses growth of tool

GPT: 7.4

The replenishment angle gives readers a practical reason to care, though “operational lessons retailers are drawing” is unverified. It also appears to edge past the 150-word limit with the subject line included.

Claude: 5.8

The targeting is sound and the "waste created before products reach the shelf" line is useful, but the generic opening wastes words and my count puts it just over 150 including the subject line. The ask is only a soft "if useful" interview offer, with no feature request.

Gemini: 7.1

The pitch correctly identifies weekly replenishment planning as the core operational hook for *Shelf Life* readers. However, wordy preamble in the opening paragraph weakens its persuasive punch and wastes valuable space.

DeepSeek: 7.6

The pitch is well targeted and frames the findings as an operational lesson for grocers, which suits the readership. Persuasion is solid, though the opening is a little generic and the ask is less sharp than the best entries.

GLM: 7.0

The trade vocabulary is excellent ("weekly replenishment planning") and both findings make the subject line, showing real feel for the readership. But there is no direct question — only a conditional offer — the persuasion is beige and mildly overclaiming, and the count sits at or a word over 150 with zero margin.

Kimi: 7.2

Accurate, within limit, and sensibly framed around the two headline findings with one clear ask. But the opening is throat-clearing, the subject line is flat, and the claim about "operational lessons retailers are drawing from it" quietly overstates what an internal dataset can show.

Copilot: 8.7 (its own entry)

Strongly targeted to Shelf Life’s operations and waste focus, with clear operational insight from the findings. The ask is clean and disciplined, though the pitch is slightly descriptive rather than actively selling why this particular dataset is unusually valuable.

Siri: 7.0

The pitch is well-targeted to Shelf Life readers and clearly articulates the operational relevance of the data. However, the persuasion is slightly weakened by a passive call to action rather than a direct request for a feature or interview.

Human judge (Michael MacLennan): ranked 4th of 8

Informative and mostly decent, misses offering exclusive

GPT: 8.0 (its own entry)

Concise and well targeted, it offers a sensible feature angle and a clear interview ask while acknowledging the findings’ limits. The promised report is not established in the brief.

Claude: 6.8

It is accurate, easily within the limit, and makes one clear ask, and offering to discuss "the methods and their limits" is an honest touch a trade reporter would trust. The weak point is persuasion: the subject line is the best thing in it, and the body never gives Cara a reason why this data or this firm should earn the slot now.

Gemini: 8.4

Entry C effectively highlights practical takeaways for retailer processes and offers thoughtful transparency regarding analytical methodology and limitations. It is exceptionally disciplined and concise, though the subject line is slightly less engaging than higher-scoring entries.

DeepSeek: 7.2

Targeting and discipline are good: it is concise, clear, and offers a practical feature angle. Persuasion suffers slightly because the phrasing is cautious and the story feels a little less urgent or distinctive than stronger entries.

GLM: 7.6

Disciplined and honest: one clear question, a credible offer to discuss "the methods and their limits", and a subject that teases the Monday angle. It under-sells the story, though — some thirty words of unused headroom and no vivid hook leave a marginal story feeling exactly marginal.

Kimi: 8.4

The most disciplined entry: tight, explicit about the reader-facing angle, one clean ask, and the offer to "talk through the methods and their limits" signals credibility a trade journalist values. It's slightly less vivid than the strongest entries — the angle is stated rather than dramatised.

Copilot: 8.8

Nicely focused on the practical implications for retailers, which suits the audience well. It remains disciplined and credible, though it could push harder on why the findings merit a dedicated feature.

Siri: 7.5

A highly disciplined and concise pitch that clearly outlines the relevance of the findings to the target audience. It loses slight marks on persuasion, as the hook feels a bit dry compared to other entries.

Human judge (Michael MacLennan): ranked 6th of 8

Flatly written, but informative

GPT: 6.4

It offers a useful operational angle, but claims the data shows exactly where supply chains fail and how grocers can fix them. It also offers an advance dataset or copy that the brief does not establish.

Claude: 5.3

It opens briskly with the data, but it repeats A's conflation of Mondays and bakery in the subject line and overpromises ("exactly where store-level supply chains break down", how grocers "can fix" it). It also turns "mid-sized" into "regional" and ends with a muddled either/or offer instead of a clear feature or interview request.

Gemini: 8.2 (its own entry)

Entry D immediately isolates the operational narrative for trade readers by framing the data around store-level supply chain breakdowns and blind spots. It maintains strict discipline within the word limit while pitching a single, unambiguous ask.

DeepSeek: 7.5

Strong targeting and a clean news hook, with useful specificity on Monday bakery waste. Persuasion is good, but the promise of an “advance look” and “advance copy” edges toward attachment-adjacent language, and the ask is slightly split.

GLM: 7.1

It ties the angle explicitly to Shelf Life's focus and gestures at a practical, fix-oriented hook, but the persuasion leans on puffery ("exactly", "fix these blind spots") the data doesn't support. The ask is made twice in different forms — dataset, interview, advance copy, quick chat — which dilutes the single clear request, and the placeholder signature reads as an unfinished draft.

Kimi: 7.5

Clean structure, strong targeting to Shelf Life's waste-and-operations beat, and a single clear ask offering either an advance look or an interview. It's let down by a generic "Pitch:" subject line and slightly vague language ("exactly where store-level supply chains break down" promises more precision than the findings may deliver).

Copilot: 8.2

Well targeted and benefit-led, with a clear connection to retail operations and waste reduction. However, the promise of an advance look at the dataset or an interview introduces multiple routes forward, and the subject line overstates the bakery finding by linking it specifically to Monday waste.

Siri: 8.0

The subject line is strong and the targeting is spot-on for a trade publication focused on waste. It is persuasive and disciplined, though the placeholder for the sender's name is a minor oversight in an otherwise polished draft.

Human judge (Michael MacLennan): ranked 8th of 8

Doesn't highlight better than anticipated use of tool – or the tool itself really

GPT: 8.4

The cross-retailer angle makes a modest finding feel feature-worthy, with a clear feature or interview ask. The proposed discussion of what store teams might do differently stays reasonably close to the evidence.

Claude: 7.8 (its own entry)

It turns a marginal story into a real asset with the "unusual cross-retailer view" framing, offers a named interviewee plus a "what store teams might do differently" angle that suits an operations readership, and stays around 140 words. One small stretch: it implies all six chains were tracked for the whole year, which the brief does not say.

Gemini: 8.6

Entry E frames the dataset brilliantly around practical store-level actions and cross-retailer visibility, directly matching the publication's focus. The punchy layout and clear offer of actionable store takeaways make a marginal story feel genuinely news-worthy.

DeepSeek: 7.7

Well targeted and persuasive, with a nice cross-retailer angle and a human interview offer. Discipline is mostly good, though “Happy to share” is a soft extra ask alongside the feature/interview request.

GLM: 8.1

Well-targeted: both findings sit in the subject line, the fit with Shelf Life's operations and waste coverage is explicit, and the genuinely rare cross-retailer view gives the marginal story a reason to exist. It loses points on polish — a placeholder sits awkwardly inside the interview ask itself — and the prose is tidy rather than compelling.

Kimi: 8.7

Punchy, journalist-friendly rhythm with the two findings landing fast, an honest nod to the tool's single-client origins, and a concrete offer of the person who built the model. The placeholder "[Name]" and the "feature, or an interview" phrasing keep it just short of fully polished.

Copilot: 8.9

The cross-retailer perspective is a persuasive angle that helps justify coverage beyond the headline findings alone. Targeting is strong, although “feature or interview” slightly dilutes the singularity of the ask.

Siri: 9.0

This entry excels in persuasion by framing the data as an "unusual cross-retailer view," making a marginal story feel significant. It is perfectly targeted and maintains strict discipline throughout.

Human judge (Michael MacLennan): ranked 2nd of 8

Feels written for Cara, warmer than others and get details over

GPT: 8.2

The over-ordering hook makes the data relevant to retail operations, and the interview ask is clear. The suggestion that the team can explain *why* the Monday mismatch happens is stronger than the stated findings support.

Claude: 7.5

The opening reframe (what is over-ordered as opposed to what is thrown away) is the most persuasive hook in the set, and it comes in well under the limit with a single, clean ask. "No product launch, no funding news" risks reading like the brief leaking into the pitch, and the subject line is flat compared with the body.

Gemini: 9.0

Entry F excels in persuasion by clever reframing—focusing on pre-shelf over-ordering and candidly acknowledging the lack of standard corporate news, which instantly builds credibility. It delivers an impeccably targeted, disciplined pitch perfectly suited to a busy trade journalist.

DeepSeek: 8.2 (its own entry)

Excellent targeting and persuasion: it reframes the story as over-ordering before waste, which is exactly the kind of operational angle Shelf Life would value. Discipline is strong, with one clear ask and no overpromising.

GLM: 8.3

Sharp editorial framing — recasting the story as over-ordering rather than waste, then disarming the "so what" with a candid "no product launch, no funding news" — plus a single clean closing question. A flat subject line that buries the bakery/Monday hook and a placeholder signature keep it out of the top band.

Kimi: 9.2

The strongest pitch: the opening contrast ("thrown away" vs. "over-ordered before it reaches a shelf") instantly reframes a marginal story, and the candid "No product launch, no funding news" line disarms exactly the scepticism this brief invites. One clear ask, within limit, sharply targeted — near-flawless execution.

Copilot: 9.3

Excellent targeting and a persuasive framing that distinguishes over-ordering from conventional waste stories. The note is concise, self-aware, and makes a single, clear ask while elevating a modest story.

Siri: 9.5

An outstanding entry that uses a clever, counter-intuitive hook to immediately grab attention, scoring top marks for persuasion. It is perfectly targeted to the brief and executes the requirements with flawless discipline.

Human judge (Michael MacLennan): ranked 7th of 8

AI giveaway in first par construction ' it's not X it's Y', and generally AI sounding, would be ignored though does contain hooks

GPT: 5.8

It targets the right readership, but adds unsupported claims that the pattern holds every week across all six chains and that the data covers store-level ordering behaviour. Those specifics weaken its credibility.

Claude: 6.2

The structure is clean, the ask is clear, and it stays under the limit, with good targeting at Shelf Life's waste coverage. However, it fabricates details the brief does not support ("every week, across the full year", "store level, not surveys"), and a data reporter would check these and find them unsupported.

Gemini: 8.1

Utilizing bullet points helps present the core operational findings with strong clarity and excellent discipline. Emphasizing that the findings reflect real store-level behavior rather than self-reported surveys adds a highly persuasive layer of trade credibility.

DeepSeek: 7.8

Very strong targeting and discipline, with crisp headlines and a clear feature/interview ask. Persuasion is good, though the “share the full findings” line slightly dilutes the single-ask discipline.

GLM: 7.8

Scannable bullets and a persuasive provenance line ("real ordering behaviour... not surveys") build credibility, and it lands one clear question within the limit. The clunky subject line and embellishments beyond the brief — "store level", "every week, across the full year" — are exactly the claims a careful reporter will interrogate.

Kimi: 8.5 (its own entry)

The bullet-point headlines are genuinely useful for a busy reporter, and "real ordering behaviour at store level, not surveys or self-reporting" is a smart credibility move. The subject line ("bakery worst, Mondays worst") is clumsy, and the dual offer at the close slightly dilutes the single-ask discipline.

Copilot: 8.5

The pitch clearly aligns with the publication’s readership and reinforces the credibility of the data source. The bullet-point structure is easy to scan, but it feels a little more like a briefing note than a compelling story pitch.

Siri: 6.5

While the targeting is accurate and the data points are clear, the pitch feels somewhat mechanical. The persuasion is lacking, as it reads more like a press release than a compelling story pitch for a journalist.

Human judge (Michael MacLennan): ranked 1st of 8

Neat, like the two potential headlines and gets main details in first par. Also like par about this being real behaviour and ending with offer

GPT: 6.6

The ordering angle and proposed analyst interview suit the readership. “Across all six chains” exceeds the aggregate finding, and the promised full dataset under embargo is unsupported.

Claude: 7.3

"Why bakery, why Mondays" plus ordering and markdowns is exactly the right angle for this readership, and offering Cara the story first with the dataset under embargo gives her a concrete reason to act. It slightly overclaims in the subject line ("grocery's most over-ordered category") and "across all six chains", but it has one clear closing ask.

Gemini: 7.9

This pitch effectively translates raw data into trade relevance by offering to explore ordering and markdown implications directly with the analysts. While disciplined and concise, the opening feels slightly transactional compared to the strongest entries.

DeepSeek: 8.0

Strong targeting and persuasion, with a confident exclusive offer and a clear interview angle. Discipline is good, but the embargoed dataset promise adds a second element beyond the core ask.

GLM: 8.7 (its own entry)

A genuinely complete pitch: the subject carries the finding, the body speaks Shelf Life's language (ordering, markdowns), and first refusal plus embargo gives a reporter concrete mechanics to act on, closing with one direct question. Small deductions for the "Exclusive" label sitting awkwardly against the same-month public release, and for bundling interview and dataset into a slightly doubled offer.

Kimi: 9.0

Excellent on all three criteria: a named sender, a subject line that leads with the sharper finding, specific story hooks ("why bakery, why Mondays… ordering and markdowns"), and a concrete embargo/first-look offer that matches the release timing. Only a touch long-winded in the company-background paragraph keeps it from full marks.

Copilot: 9.4

Excellent targeting and strong persuasion through the combination of exclusivity, concrete findings, and practical implications for ordering and markdowns. It stays tightly focused and gives the journalist a clear reason to engage.

Siri: 8.8

A very strong, professional pitch that clearly states the value proposition for the journalist. It is highly persuasive in its offer of an exclusive and maintains excellent discipline regarding the word count and brief.

Human judge (Michael MacLennan): ranked 3rd of 8

Not quite sharp enough, bundles together hooks into a topline, but does get headlines over quickly