campaign-plan

Campaign plan

Most campaign plans are written forwards: we'll spend $200k, here's what we hope happens. That shape is unfalsifiable and therefore easy to approve, which is why so many of them get approved and so few of them work.

Working backwards inverts the burden. Start from the number the business has already committed to, derive what has to happen at every stage for it to land, and the plan either closes or visibly does not — before anyone spends money.

The workflow

1. Get the real target and the real conversion rates

The target is whatever revenue or pipeline number marketing is accountable for. Take it as given; the point of this exercise is not to negotiate it.

Subtract what is already in flight. If deals are already open against the same period, the campaign is accountable for the gap they will not close, not the whole number. Run the open pipeline through the engine first:

PYTHONPATH="${CLAUDE_PLUGIN_ROOT}" python3 -m gtmkit.pipeline pipeline.csv \
  --stages stages.json --target 4000000 --closed 650000 \
  --period-end 2026-12-31 --as-of 2026-10-06

The shortfall at the expected landing becomes the funnel's --target-revenue, and the shortfall at the low end of the landing range is the stress case in step 4. The stage probabilities need sources just like the conversion rates below, and the engine refuses a CRM default declared as a fact. The format is in the pipeline format reference.

Conversion rates are where plans go wrong, and there are only three honest sources for them, in descending order of trustworthiness:

1. The user's own historical data, segmented the same way the campaign is.
2. The user's own historical data, unsegmented — usable, but say so, because blended rates hide that enterprise and self-serve convert differently by an order of magnitude.
3. Nothing — in which case you are building a scenario, not a forecast, and the output must be labeled that way.

Never supply industry benchmark conversion rates as though they were the user's. The variance between companies in the same category is larger than the difference the campaign is trying to make.

2. Establish the audience ceiling before anything else

Ask: how many organizations or people are actually reachable in this segment, through this channel, in this timeframe?

This single number kills more plans than any other, and it is the one most often skipped. A plan requiring 22 million impressions against an addressable audience of 200,000 people is not aggressive, it is arithmetically impossible — and it is far better to discover that in planning than in month two.

3. Run the math

PYTHONPATH="${CLAUDE_PLUGIN_ROOT}" python3 -m gtmkit.funnel \
  --target-revenue 4000000 \
  --acv 45000 \
  --stage "impression:0.002" \
  --stage "visit:0.08" \
  --stage "mql:0.25" \
  --stage "sql:0.35" \
  --stage "opp:0.28" \
  --cpm 22 \
  --audience-ceiling 40000000 \
  --sales-cycle-days 96

Each --stage is name:rate, where the rate is the probability of moving to the next stage. The final stage converts to closed-won, or to --win-rate if you model the close separately.

The command exits non-zero when the plan exceeds the audience ceiling, so it can gate a CI check or a planning script.

4. Stress the plan before presenting it

Run it three times: at the user's stated rates, at rates 25% worse, and at the best rates they have ever achieved. The spread between those three is the real finding.

If the plan only closes at the best rates the team has ever achieved, it does not close. Say that plainly, then offer the levers rather than just the verdict.

5. Name the lever, not just the gap

When a plan does not close, there are exactly five things that can change, and naming which one you are recommending is what separates a planner from a reporter:

Lever	When it is the right answer	What it costs
Raise ACV	Pipeline is adequate, deals are small	Longer cycles, narrower segment
Improve a conversion rate	One stage is visibly worse than peers	Time; usually a quarter before it shows
Add a channel	Ceiling-bound, not rate-bound	New CAC to learn from scratch
Extend the timeline	Cycle length exceeds the period	Misses the committed date
Lower the target	The others are exhausted	The conversation nobody wants

Pick the one with the best ratio of impact to elapsed time, and state the second choice too — planning conversations go better when the recommendation has a visible alternative.

6. Work the calendar backwards

A 96-day sales cycle means Q4 revenue requires pipeline created in Q3. This is obvious and routinely ignored, producing plans that commit to revenue in a period whose pipeline window has already closed.

Put the pipeline-creation deadline in the plan as a dated milestone, not as a note. If that date is already in the past, the plan needs a different shape entirely and you should say so in the first paragraph.

Budget allocation across channels

When splitting spend across channels, allocate against marginal CAC rather than blended CAC. A channel with $400 blended CAC that degrades to $1,200 at triple the volume is not a $400 channel at the volume you are planning.

Ask for the historical relationship between spend and volume per channel. If it does not exist, cap incremental spend in any single channel at roughly 1.5x its proven level and route the rest elsewhere. Saturation is the most common reason a plan that penciled at the portfolio level misses in execution.

What to hand over

A campaign plan that will survive review contains:

1. The target net of what open pipeline is expected to close, and the deal count it implies.
2. Required volume at each stage, with the source of every conversion rate.
3. Spend, blended CAC, and CAC as a share of first-year ACV.
4. The audience ceiling and the headroom against it.
5. The pipeline-creation deadline, dated.
6. What happens if the two weakest rates come in 25% below plan.
7. The one lever you would pull first if it does.

Point seven is the one that gets the plan approved.

Common failures

Blended rates across mixed segments. Enterprise and self-serve converting at one blended rate produces a plan that is wrong for both.

Counting MQLs as the goal. MQL volume is an input. A plan that hits MQL target and misses revenue has failed, and structuring the plan around the intermediate metric is how that happens.

Ignoring sales capacity. 317 opportunities against 4 reps and a 96-day cycle is not a marketing plan, it is a marketing plan and a hiring plan. Check the opportunity count against the coverage the team can actually work.

Treating CAC as efficient without knowing gross margin. CAC at 12% of ACV sounds excellent and may not be, depending on margin and retention. State the comparison rather than implying it.