Product Marketing: Who Is This For, and What Do We Promise?

On Monday of week four Camille Duarte has three documents on her desk and one sentence to write. Joel’s customer needs report, every finding carrying a confidence grade. Wes’s product truth document, with its approved facts and its forbidden verbs. And Delia’s mandate, which drew a boundary around the audience before anybody had evidence about anyone.

By Thursday she has to hand Ryan, Naomi and Theo something they can plan a $1.2m campaign on. Almost everything those three produce over the following ten weeks will be an expression of the sentence she writes this week — and none of them will have time to check whether it was the right one.

What lands on Camille’s desk

Joel’s report is organised by need, and every finding carries a grade. The one flagged above the others is high confidence, carried by three differently-biased sources: the buyer’s real fear is being blamed for a shortage nobody saw coming, and their real want is never to be surprised in front of their boss again. Moderate findings Camille may use but not lead with; low findings she must test first — and everything Joel has on distributors and wholesalers is low, because he interviewed nobody there.

Wes’s product truth document sets the outer wall: Supply Signal watches 1.4m suppliers for nine risk signals and flags them earlier than a person would, it does not predict, implementation takes six weeks, and there is no native SAP connector until the quarter after launch.

Delia’s boundary is the third input and the one Camille handles most carefully. Delia named mid-market discrete manufacturers in North America, $50m to $500m of revenue, about 14,000 companies — before any evidence existed, as a market size and an instinct. Camille’s job is to test that boundary rather than inherit it, because an assumption gets reopened every time somebody senior feels uneasy, and a decision with evidence behind it does not.

What a product marketing manager actually does

A product marketing manager sits between what was built and who it is sold to, and the job is almost entirely one of subtraction. Engineering builds something capable of several things; sales would like to sell it to everybody. Product marketing decides which capability, aimed at which buyer, in which words — and writes it down in a form that survives being handed to people under deadline pressure who were not in the room.

The output is a document, not a campaign. Camille will not write an advert this quarter. She writes the sentence every advert has to express, and the list of sentences that may never be written at all.

The software on Camille’s desk

The segment decision is made in a spreadsheet, because the three candidates differ on numbers that have to sit side by side: how many companies, what they pay, how long they take to buy, and — the column people forget — how well the product actually serves them. Competitive intelligence tools track what rivals say and publish, though Camille’s most important competitor appears in no such tool, because the spreadsheet a plant manager already maintains has no marketing department.

The messaging brief lives beside Wes’s truth document, deliberately. And the last screen is where positioning succeeds or quietly dies: sales enablement is the place the sentence becomes what a rep actually says on a call. A brief that never reaches the deck is a document, not a position.

The four decisions

Choosing the segment

Three groups could plausibly buy Supply Signal, and Camille works through all three on paper rather than letting the answer be whoever lobbies hardest.

Enterprise manufacturers, above $1bn of revenue, number about 900. Sales wants them, and the arithmetic is seductive: contracts several times the size of $48,000 mean perhaps sixty opportunities rather than 190 to reach $9m. What kills it is time. Enterprise buying runs on eleven-month cycles and formal RFPs — security review, references at scale, a scored questionnaire two hundred rows long that asks about SAP integration near the top. A fourteen-week launch cannot produce qualified pipeline from a process that has not finished its first stage by then. This is not a segment Cadence should never sell to; it is one that cannot be reached inside this campaign’s clock.

Distributors and wholesalers number about 22,000 and are cheap to reach. Two things sink them. Contract value is roughly half, which is where the cheapness stops being an advantage: at around $24,000 a deal, $9m of pipeline needs 380 opportunities, some 1,520 MQLs and over 5,000 leads, and $1.2m cannot buy 5,000 leads of any quality. The second problem is worse. Supply Signal watches the nine things that go wrong before a component supplier fails a production line; a distributor’s exposure is spread across thousands of SKUs and hundreds of vendors, and most of the nine signals do not mean downstream what they mean upstream. Joel’s evidence here is entirely low confidence — not bad news, but an admission that nobody at Cadence knows what these buyers feel.

Mid-market discrete manufacturers, the 14,000, are where Camille lands. The pain is real and recent, the buyer is identifiable, the contract value matches the funnel Delia already built, and — decisively — it is the only one of the three Joel has high-confidence evidence for.

What makes this a decision rather than a preference is the second half of it. Camille writes down what she is giving up: the enterprise logos that would make the launch look impressive, and 22,000 companies she will not spend a dollar on. Choosing a segment is mostly choosing what to forgo, and a choice recorded without its cost is not a choice, it is a hope.

Positioning against the real competitor

Anika’s research came back with two named rivals. Arbor Risk is the enterprise incumbent: credible, expensive, slow to deploy. Nodefield is the cheap startup: fast, shallow supplier coverage, undercutting on price. Both are real, and both are the wrong thing to organise a campaign around, because the competitor taking most of this market is a spreadsheet maintained by somebody with forty other jobs. Most of the 14,000 are not evaluating anybody. They are carrying on.

Positioning against a rival and positioning against nothing are different arguments. Against Arbor Risk, Camille is comparing — deeper coverage, faster to live, a price a plant director can approve without a board paper. Against nothing there is nothing to compare with, and the instinct to say “we are better” has no object. The spreadsheet has advantages no product can match: free, already installed, needs nobody’s approval, no implementation. And it has never once been blamed for anything, because when it fails the failure is attributed to the supplier.

So the argument changes shape. You are not claiming to be better than the status quo. You are claiming the status quo has a cost the buyer is already paying and has never counted. That cost is not hypothetical, and Joel has it in the buyer’s own words: the expedited freight, the line standing idle, the Monday meeting where a plant operations director was asked why and had no answer. None of it sits on a budget line. It is absorbed into “supply chain is difficult”, which is the most expensive sentence in manufacturing. Camille writes it as a theme with verbatims behind it, never as a statistic — Joel forbade turning eight of twelve interviews into a percentage, and an invented one would reach Miriam in module 9 as an unsupported claim.

The value proposition

Joel handed over a feeling. Camille has to hand on a promise, and the two are not the same object.

Here is the weak version, which writes itself and which every company in the category has already published: Supply Signal gives you complete visibility of supply chain risk, so you are never caught out again. It sounds strong. It is the category sentence — Arbor Risk’s ground, argued with a fraction of Arbor Risk’s budget. It promises what the product cannot keep, because seven of nine is not never. And no event would demonstrate a failure of “complete visibility”, which is exactly why a buyer who has been sold three products in five years does not believe it.

Here is the version Camille writes: Supply Signal watches your suppliers for the nine things that go wrong before one of them fails you, and warns your team weeks before it reaches the line — so you hear it from us, not from your customer. Live in six weeks.

Every clause is bounded by something Wes wrote. Watches and warns, never predicts. Nine signals, an approved fact. Weeks before it reaches the line is a claim about lead time, which is what the buyer actually wants — the difference between rerouting an order and stopping a plant. Live in six weeks puts the implementation limit inside the promise rather than in a footnote a salesperson delivers in week six of a sales cycle. And the last clause is falsifiable in the only way that matters: if a customer learns about a supplier problem from their own customer first, Cadence has broken its promise and can be told so. That is uncomfortable, and it is precisely what makes the sentence worth believing.

Note what Camille does not promise. “Never surprised in front of your boss again” is the buyer’s want, and a want is not a warrantable claim — Supply Signal missed two of nine at Calder Thermal, and a promise that cannot survive a bad quarter gets quoted back at you during the first one. You may name the fear. You may not guarantee its absence. The fear goes in the headline; the promise goes in the sentence underneath it.

What she writes down so others cannot drift

The last decision is about form, and it determines whether the other three survive ten weeks of production.

The messaging brief is two pages. At the top, the one sentence. Beneath it, three supporting points, each with its proof and its sample size welded on: breadth of watching (1.4m suppliers, nine signals — Wes’s document, dated, his name against it); lead time (in a one-quarter pilot, Calder Thermal saw Supply Signal flag seven of nine supplier disruptions at least three weeks before they hit production — n = 9, one customer, one quarter); and time to value (live in six weeks, no native SAP connector until the following quarter). Then the words to use — watches, flags, warns, lead time, hear it first — and the buyer’s own nouns: line down, expedited freight, the Monday meeting.

Then the part that looks like pedantry and is not: the words to avoid, each with a line of reason. Predicts — false, and Wes refused it. Prevents, guarantees, never — promises the product cannot keep on a bad day. Complete visibility, end-to-end, total — the category sentence, on the incumbent’s ground. Real-time, instant, in a day — contradicts a six-week implementation. Up to three weeks early — reads as what a buyer can expect, and needs evidence the pilot cannot bear.

That list saves the campaign twice, both times later. In module 7 the content agents will produce hundreds of variants, and a generative system asked for thirty headlines reaches for the strongest available adjective unless the strong adjectives are excluded by name; a brief listing only what is permitted quietly authorises everything it fails to mention. And in module 9 Miriam receives roughly three times a normal quarter’s material. A forbidden-words list turns much of that review from a judgement on every sentence into a search.

Where this goes wrong

In real companies the segment decision is never formally made. Sales pushes for enterprise because the logos look better on a board slide, partnerships pushes for distributors because a channel deal is available, and the campaign ends up aimed at all three — which means it is aimed at nobody, because the message that works for a plant director in Ohio is not the message that works for a procurement committee at a $3bn manufacturer. The budget spreads thin and every channel underperforms slightly, which is far harder to diagnose than one channel failing loudly. Then in week ten somebody senior asks why enterprise is not in the campaign, and because nobody wrote down what was given up and why, the argument has to be had again from scratch, under deadline.

The second failure is positioning by committee. Six people review the sentence, each adds the thing they care about, and out comes something containing visibility, AI, end-to-end and platform. It offends nobody, says nothing, and competes with the incumbent on the incumbent’s chosen ground.

What Camille hands on

The messaging brief goes to Ryan Kovacs in module 6, Naomi Feldstein in module 7 and Theo Alvarez in module 8, and each inherits a different constraint from it.

Ryan inherits the segment and the awkward consequence of the positioning: his audience is defined by not looking. He cannot buy his way to 2,530 leads on category search terms, because most of the 14,000 are not searching for a category they have not decided they need. His journey has to create the awareness that search would otherwise capture.

Naomi inherits both word lists as literal material for agent prompts, and the rule that every claim must trace back to a proof point here with its sample size intact. Theo inherits something subtler: the brief constrains pictures as well as words. A crystal ball, a forecast curve arching into next quarter, a dashboard showing a problem already solved — each says predicts without using the word, and an image is not exempt from a claim merely because it contains no sentence.

The line all three work inside is short enough to remember, which is the only test a positioning statement really has to pass: you hear it from us, not from your customer.

Designing this desk’s agent: the comparison agent

Camille needs a comparison across three competitors and a status quo, built from evidence rather than from what the sales team believes. The hazard is specific: a comparative claim about a named competitor is a legal artefact, not a marketing one.

One concrete design choice. Separate their claim from our verified position in the schema, so the two can never merge in a draft. A competitor’s own page is evidence of what they say, never of what they do.

{
  "competitor": "Nodefield",
  "dimension": "implementation_time",
  "their_claim": "live in a week",
  "their_claim_source": "https://...",
  "our_position": "six weeks",
  "our_evidence_id": "impl-avg-2026h1",
  "comparative_claim_allowed": false
}

The metric to track. The share of proposed differentiators that survive both product and legal review. A high rejection rate is not the agent failing — it is the agent doing its job upstream of two people whose time is scarcer than compute.

Strong, or just confident?

Read each line and decide whether Camille would let it through, then tap a card.