Skip to content
Shepherd GTM

Spend

Every dollar by channel and play, next to what it produced. Cost per submission is the number. The readout says what to cut.

Recorded by hand for this demo. In production these inputs run off the card feed and CRM APIs, and every unit cost computes itself monthly.By hand
Connect card feed Connect CRM
Actual spend
$4,040
Planned $380
Submissions
29
52 meetings · 214 replies
Cost per submission
$139
actual spend / submissions
Cost per acquired broker
$238
actual spend / first-time submitting brokers
Cost per office unlock
$449
actual spend / new submitting offices
System activity
3
drafts · 0 approved · 2 briefs · 89 signals · 4 audits

Spend by channel

actual vs planned
Data
$620 / $200
Models
$240 / $80
Email
$100 / $100
AEO tool
$500 / $0
Video editor
$1,400 / $0
Paid social
$1,180 / $0
actualplanned

By play

spend against outcomes
PlaySpendMtgsSubs$/sub
Signal Engine$7205229$25
Targeting$1,18000–
AI visibility$1,90000–
Shared$24000–

Spend entries

12
  • planned
    Data · ApolloSignal Engine
    Seats and credits for ~2,000 brokers
    $200
  • planned
    Models · Anthropic, GroqShared
    Scoring, drafts, briefs, audit
    $80
  • planned
    Email · Existing senderSignal Engine
    Sending and tracking
    $100
  • planned
    AEO tool · AthenaHQ (optional)AI visibility
    Off until the 30-day baseline moves. $295/mo on Starter; free tier to test the fit first.
    $0
  • planned
    Video editor · ContractorAI visibility
    Starts with the first recording
    $0
  • planned
    Paid social · LinkedInAI visibility
    Day 61, behind the two best clips
    $0
  • actual
    Data · Apollo + Dodge trialSignal Engine
    Broker universe at 2,000 seats plus the one-state permit trial
    $620
  • actual
    Models · Anthropic, GroqShared
    Scoring, drafts, briefs, audits across all three plays
    $240
  • actual
    Email · Existing senderSignal Engine
    Three underwriter mailboxes, 60 sends a day
    $100
  • actual
    Paid social · LinkedIn Campaign ManagerTargeting
    Company + title flight at the month's target offices, held-out control
    $1,180
  • actual
    AEO tool · ProfoundAI visibility
    Turned on in month three after the baseline moved; stored baselines carried over
    $500
  • actual
    Video editor · ContractorAI visibility
    12 to 16 cuts from one founder recording
    $1,400

Outcomes

what the spend produced · the denominators for the unit costs above
  • RepliesSignal Enginesystem
    22% on 960 signal-led sends
    214
  • MeetingsSignal Enginesystem
    52
  • SubmissionsSignal Enginemanual
    Excess leads the mix; builder's risk second
    29
  • IndicationsSignal Enginemanual
    All inside 24 hours
    24
  • First-time submitting brokersSignal Enginemanual
    First-time submitters this month
    17
  • New submitting officesSignal Enginemanual
    Lockton Houston, Alliant Dallas, USI Denver among them
    7
  • New submitting officesTargetingmanual
    Targeted-flight offices out-replied the control 2.4 to 1
    2
  • Engaged brokersAI visibilitysystem
    Matched to the Play 1 universe, feeding the Intent axis
    412
  • Pages publishedAI visibilitymanual
    All twelve audit prompts covered, two refreshed
    14

Readout

what worked, what to cut, written from the numbers above

August was a $380 planned month with 77 signals, 2 drafts, 0 approvals, and zero outcomes logged, so the program did not reach a single broker.

Cost per submission cannot be computed for August. There were zero sends, zero logged submissions, and $380 of spend, so the ratio is undefined rather than bad. The first job is to log outcomes; until then the only honest figure is $380 spent for zero broker contacts.

Working

  • Signal capture is running · 77 signals collected against $200 of Apollo seats and credits for roughly 2,000 brokers. The data layer is the only part of the stack producing volume.
  • Spend discipline held · Total planned outlay $380. AthenaHQ, video editing, and LinkedIn paid social were all held at $0 pending a baseline move and a first recording. Nothing was spent on channels with no proof behind them.
  • Audit trail exists · 3 audits and 2 briefs logged on $80 of model spend. When outcomes do start landing, we will be able to trace which brief produced them.

Not working

  • Nothing was approved, so nothing was sent · 2 drafts, 0 approved. The email line item at $100 covered sending and tracking for zero sends. The bottleneck is underwriter approval time, not tooling.
  • Outcome logging is absent · No replies, meetings, or submissions recorded. This is expected when sends are zero, but there is also no logging mechanism proven to work, which means next month could look identical even if sends happen.
  • Signal to draft conversion is 77 to 2 · Under 3% of signals became a draft. Either the scoring is too tight or the drafting step is not being run against the full signal set.
  • Content engine has not started · Zero recordings means the video editor and the day 61 LinkedIn plan behind the two best clips have no input. The speed plus differentiated pricing pitch has not been put in front of any of the roughly 200 firms.

Next month

  • Set a hard approval SLA with the underwriters: every draft reviewed within 48 hours, and ship the 2 existing drafts in week one of September. · Approval is the only step between existing work and broker contact. No spend fixes this. · $0 incremental. Unlocks the $100 email line that produced nothing in August.
  • Stand up outcome logging before any new send: reply, meeting, submission, bound, per office and per broker. · Cost per submission is the number that matters and it cannot exist without a submission field. Without it we repeat August in September. · $0 incremental, inside the existing $80 model and audit budget.
  • Push drafts from 2 to 40 against the 77 signals, targeting appointed offices that have never sent a risk. · Wider growth needs volume of first contacts. Two drafts cannot test a message. · Roughly $40 to $60 more in model spend. Keeps the month under $450.
  • Record one underwriter clip on Shepherd Savings, using the DroneDeploy result of 44% fewer claims and 48% lower loss rate paired with the 24 hour indication. · It is the input the video editor and the day 61 LinkedIn plan both wait on, and it is the only pitch that pairs speed with differentiated pricing. · Triggers the contractor line, budget $300 to $500 for the first edit. Keep LinkedIn at $0 until two clips exist.

Caveat: These numbers describe a program that has not launched, not a program that is underperforming. The 77 signals and 2 briefs are setup work, and reading them as weak demand would be wrong. The real risk is treating another quiet month as a data problem when it is an approval problem inside our own team.