The Campaign Manager: How a Stranger Becomes a Buyer

Ryan Kovacs gets Camille’s messaging brief on Thursday afternoon of week four and has until the following Wednesday to turn it into a plan. It is two pages: one sentence, three proof points, two word lists. What it does not contain is a single date, a single dollar, or any indication of how a plant operations director in Ohio who has never heard of Cadence Systems ends up on a call with a salesperson fourteen weeks from now.

That is Ryan’s entire job, and the plan he writes this week is the document Naomi, Theo and Dana will work to for the next ten. He also writes down one problem he cannot solve, which turns out to be the most useful thing in it.

What lands on Ryan’s desk

Camille’s brief fixes what the campaign says and to whom: mid-market discrete manufacturers, roughly 14,000 of them, and the line you hear it from us, not from your customer. It also hands Ryan an awkward inheritance. That segment was chosen partly because its real competitor is a spreadsheet, which means most of these companies are not evaluating anybody and therefore not searching for anything. Ryan cannot buy demand that already exists; he has to manufacture some of it.

Delia’s arithmetic from module 1 is the other half. 2,530 leads become 760 MQLs become 190 opportunities become $9m of pipeline, on $1.2m — of which $38,000 is already committed to agent tooling across the research, content, creative and reporting desks. That leaves about $1,162,000 and a blended target of $474 a lead.

The launch date is not his either. It is tied to a trade show booked and paid for before any of this started, which happens in week fourteen whether or not Cadence is ready.

What a campaign manager actually does

A campaign manager does not write copy, design ads or build landing pages. They decide what gets made, in what order, for whom, with what money, by when — then spend most of the week chasing the dependencies between those things. The output is a plan, and its real function is to make a hundred small decisions in advance so that ten people do not each make them differently in week nine.

The distinction separating a good campaign manager from an expensive one is between activity and progression. Activity is what was produced and published. Progression is whether anybody moved.

The software on Ryan’s desk

The plan is a model before it is anything else: channels down the side, spend and cost per lead across, and a total that has to reach 2,530 leads. Its value is not the forecast — which will be wrong — but that it makes the trade-offs explicit, so a decision to move money later is an argument about assumptions rather than instinct.

The ad platforms are where the model becomes real money, and the discipline is that the two must match; a plan the platforms contradict is decoration. Asana holds the critical path, including the dependency Ryan flags and cannot solve: everything must clear legal review, and the estimate for that step is based on a volume that no longer applies. And the pacing dashboard is what makes the reserve usable — you cannot reallocate what you have not noticed.

The four decisions

The journey before the channels

Most campaign plans are a list of channels with a budget beside each. That is a purchasing document dressed as a strategy, and it produces a predictable failure: the same offer, in the same words, aimed at people in completely different states of mind.

Ryan plans the sequence first. A stranger becomes a buyer by passing through five states, and each is defined by what the person knows, not by what Cadence sent them.

Unaware. The plant director does not think they have a problem with a name; they think supply chain is difficult, which it is. The job is not to sell a product but to make the cost of the status quo visible — exactly the positioning Camille chose — and to ask for nothing in return.

Aware. They now believe this can be managed rather than endured. The job is to make it urgent enough to spend attention on, usually by attaching it to something they already fear.

Interested. They would like it solved. The job is to give them something concrete enough to show a colleague, because in a mid-market manufacturer nobody buys $48,000 of software alone.

Evaluating. They are comparing options — including, most often, doing nothing. The job is removing risk: how long implementation takes, what it needs from their systems, who else has done it. This is where the six weeks and the missing SAP connector belong, published rather than discovered.

In conversation. They are talking to Curtis’s team, and marketing’s job is to make that conversation shorter by having already answered the four questions a rep would otherwise spend a call on.

What makes this more than a diagram is that each state needs a different job done, not a different advert. The commonest waste in business marketing is running “book a demo” at all five stages: it converts respectably at the last two, fails silently at the first three, and the failure is misread as a channel problem. The channel was fine. The request was absurd.

Because Camille’s segment is defined by not looking, most of the 14,000 sit in the first two states — which is expensive, since creating awareness costs more per lead than capturing intent. Ryan’s plan is therefore front-heavy, and he says so in writing, because in week eight somebody will ask why so much money is going somewhere that produces no leads.

The budget split, with real arithmetic

Ryan holds back a reserve before allocating anything — that is the next decision — which leaves roughly $988,000 to commit, still expected to produce 2,530 leads. The plan’s operating cost per lead is therefore about $390, not Delia’s $474, because the reserve is money he intends not to spend yet.

Paid search takes $160,000 at roughly $640 a lead — about 250 leads. It is the most expensive line per lead and the smallest volume, deliberately: search only captures people already typing the category into a box, and Camille’s whole point is that most of this market is not.

Paid social, almost entirely LinkedIn, takes $330,000 at about $440 a lead for roughly 750 leads — the largest allocation, because it is the only channel that reaches people by job title and company size rather than waiting for them to raise a hand.

The trade show takes $230,000 — booth, stand build, staff, travel and the launch event — and returns perhaps 300 leads at about $767 each, the worst cost per lead in the plan. It is also where a plant director stands in front of a screen for ten minutes and says out loud what actually worries them.

Webinars take $115,000 for around 450 leads at roughly $255, and do the “interested” job better than anything else: an hour of self-selected attention with the customer’s own questions attached.

Content and email take $153,000 — gated research, the nurture sequence, syndication — for about 785 leads at roughly $195, the cheapest and largest single source of volume.

That is $988,000 and about 2,535 leads. Now the part that matters. Ryan does the same division one stage later, using the conversion rates Cadence has seen historically, and the ranking rearranges itself. The trade show’s 300 leads convert to MQL at about 55% — 165 MQLs. Paid search converts at 45%, giving 112. Webinars at 35% give 157. LinkedIn at 25% gives 187. Content and email, the cheapest leads in the plan, convert at about 18% — 141 from 785 leads. Total: near enough 760, which is the number Delia needs.

Set those two paragraphs against each other and the trap is obvious. The trade show produces the campaign’s most expensive leads and its second-largest block of MQLs; content and email produce the most leads and the fewest qualified ones. A channel delivering 900 cheap leads that never convert is worse than one delivering 200 expensive leads that do — worse not only because the pipeline does not appear, but because somebody in Curtis’s team spends three weeks phoning them, and that cost never shows up in the campaign report.

Which is why Ryan turns down a syndication vendor offering 900 additional leads at $170 each. It would improve every number on the weekly dashboard and damage the only one Delia is judged on: $6,300 per opportunity. Cost per lead is a purchasing metric. Cost per opportunity is the business.

The reserve

Ryan holds back about 15% — roughly $174,000 — with no channel attached to it.

The argument is simple and almost never wins on its own. A campaign plan is a set of hypotheses about where an audience is and what will move them, and in week three the plan starts returning evidence: one LinkedIn audience performing at half the expected cost, the first webinar filling in two days, the trade show list converting at a rate nobody predicted. A plan with no reserve cannot act on any of it. By week three most of the budget is not merely allocated but committed — sponsorships paid, insertion orders signed, a stand being built — so the only way to fund the thing that is working is to break something already bought.

Finance dislikes reserves for an equally sound reason, and it is worth stating fairly. An unallocated line looks like slack, and slack looks like a budget that was larger than it needed to be. Arthur Reyes will eventually ask why a campaign that requested $1.2m only assigned about a million of it, and the honest answer — “because I do not yet know which quarter of this will work” — sounds in a budget meeting like a confession.

So Ryan makes the reserve defensible rather than vague. It has a decision date of the end of week six. It has a rule: it goes to whichever channel has the lowest cost per MQL at that point, not the lowest cost per lead. And it has a default: if nothing has clearly outperformed by week six, it goes to the webinar programme, which can absorb money at short notice. A reserve with a date, a rule and a default is a plan. A reserve without them is a slush fund, and finance is right to distrust it.

The timeline, and the dependency nobody owns

Fourteen weeks, one fixed point. The trade show happens in week fourteen, so every date in the plan is calculated backwards from it rather than forwards from today.

Critical path is the phrase for the chain of tasks where any slip moves the end date: content late means creative late, review late, launch late. Tasks off that chain have slack — the third webinar can move a week and nothing else changes. The concept matters because most people track the tasks they own and almost nobody tracks whether their task sits on the chain, so a two-day delay on something that felt minor turns out to have cost the launch two days.

Cadence’s chain runs: messaging brief in week four, content drafts weeks five to eight with Naomi, creative weeks six to nine with Theo, legal and factual review in week ten, build and instrumentation weeks ten to twelve with Dana, then launch. Running alongside it, booth graphics have a printer deadline three weeks before the show.

And there is the dependency Ryan flags and cannot fix. Everything — every ad variant, every email, every landing page, the webinar deck, the booth panels — must clear Miriam’s review in module 9, and Ryan has scheduled one week for it, because one week is what it has always taken. That estimate comes from a world that no longer exists. Miriam’s queue in a normal quarter is about forty hours of material; this launch, with content and creative agents running, will hand her roughly three times that, in about a third of the time, every sentence fluent enough to read as though somebody has already checked it. Nobody has budgeted for it, because nobody has ever had to.

The reason it goes unowned is structural. Review is not a deliverable; it is a gate, and gates rarely appear on plans as tasks with an owner, a duration and someone accountable for the estimate. They appear as assumptions. The most dangerous item in any schedule is the step everyone assumes will take as long as it used to.

Ryan does what a campaign manager can do and no more. He writes the review week in as a named risk with a date, asks for Miriam to be at the week-five kick-off rather than meeting the material for the first time in week ten, and proposes that assets reach her in batches from week seven instead of in one lump. What he cannot do is add reviewer capacity, move the show, or make ten other people care about a bottleneck that is not yet hurting them. So he flags it, unsolved, in writing — which is worth more than it sounds, because in week ten that is the difference between a known risk materialising and a crisis nobody saw.

Where this goes wrong

The first is the channel-list campaign: a budget divided between platforms, one offer, one call to action, delivered identically to people who have never heard of the problem and people ready to buy. It produces respectable click rates and almost no progression, and every post-mortem blames the creative.

The second is optimising on cost per lead. Around week seven the cheapest channel looks like the best one, budget moves towards it, lead volume rises, MQLs flatten and opportunities fall. The campaign hits its lead target and misses its pipeline target — the most common way a marketing team is congratulated on the way to failing.

The third is the fully allocated plan, which learns something valuable in week three and has no free money to act on it, so the learning goes into the wash-up deck and is applied next year.

What Ryan hands on

Naomi Feldstein receives the content requirement in module 7: not a list of assets but a matrix of journey stage against channel, with volumes and dates, so that what is needed in week six is fifteen pieces doing the unaware job rather than fifteen product pages. Theo Alvarez receives formats, quantities and the printer deadline, which is earlier than everything else and therefore governs decisions that feel as though they can wait. Dana Whitlock receives the stage definitions, the tracking requirements and one hard rule: no paid spend begins before the instrumentation is live, because leads you cannot attribute are leads you cannot optimise, and by the time the reports look wrong the money has gone.

All three also inherit the review bottleneck, described plainly and not resolved. Ryan’s plan states how much material will reach Miriam, in which week, and how long he has assumed it will take her — stated as an assumption rather than a fact, which is the only responsible way to hand on a number you know is soft.

Designing this desk’s agent: the pacing agent

Ryan’s reserve is only useful if somebody notices, early, that a channel is underperforming. That is continuous, boring, arithmetic work — which is exactly the kind an agent is good at and a person is not.

One concrete design choice. Make the recommendation record require a downstream metric, not just cost per lead. A recommendation that cites only lead cost is rejected by the schema, because the cheapest leads are the trap this entire module warns about.

{
  "channel": "paid_search",
  "window": "week_03",
  "cpl_plan": 310,
  "cpl_actual": 214,
  "mql_rate_plan": 0.55,
  "mql_rate_actual": 0.31,
  "recommended_delta_usd": -40000,
  "requires_approval": true
}

The metric to track. Realised versus predicted. When a recommendation is accepted, did the change produce what the agent forecast? Track that over a quarter and you learn whether to trust it — which is a far better governance mechanism than deciding in advance how much autonomy it deserves.

Cheap, or actually working?

Read each situation and decide what you would do, then tap a card.