How to split a go to market budget, and what to cut first
By Jānis Plūme, Founder, AllboundPros and Outbound Pros · 2026-08-06
Quick answer
Split go to market budget into three buckets before splitting it across channels: capacity that produces pipeline now, experiments that find the next working motion, and compounding assets that produce pipeline later at declining cost. Allocate across channels using cost per meeting held, not per meeting booked, because show rate can run near 50% where calendar discipline is broken. When budget is cut, cut experiments first, compounding assets second, and capacity last, unless capacity is already below the kill thresholds.
There are no pricing figures anywhere on this site, in either direction. This page is a model you fill in with your own costs. Any page that tells you what to spend without knowing your deal size, market, cycle length and existing infrastructure is guessing on your behalf, and confidently.
What are the three budget buckets?
Capacity, experiments and compounding assets. They exist as separate buckets because those three types of spend have completely different return profiles and completely different behaviour under a cut.
Capacity is spend that converts directly into pipeline this quarter at a roughly linear rate. Outbound sending capacity, SDR headcount, paid acquisition. You can turn it up and down and the pipeline follows within one ramp period. It is expensive and it is predictable, which is exactly what you want when you need a number by a date.
Experiments are spend on motions you have not proven yet: a new segment, a new geography, a new channel, a new offer. Experiment budget buys information rather than pipeline, and it should be sized as such. Its return is measured in decisions made, not deals closed. If you cannot name the decision an experiment will let you make, you have funded a hunch and given it a budget line.
Compounding assets are spend that produces at declining marginal cost over time. Content that gets found and cited, a product led surface, a partner ecosystem, a genuinely useful free tool. Compounding assets are the only line that gets cheaper per unit of pipeline, and they are the first thing everybody cuts, which is why so many companies have been doing go to market for five years with no accumulated advantage.
A reasonable working structure is a large majority to capacity, a meaningful minority to compounding assets, and a small fixed slice to experiments that is defended instead of raided. I will not put percentages on those, because the right split for a company that needs pipeline in six weeks is not the right split for one with eighteen months of runway, and a number here would be quoted as though it were a benchmark.
Here is the same thing as a table, because the buckets are easier to argue about when their behaviour under a cut is side by side.
| Bucket | What the line item is | What it buys | How to size it | Time to impact | What happens when you cut it |
|---|---|---|---|---|---|
| Capacity | Sending infrastructure, SDR seats, paid acquisition, agency retainer | Pipeline this quarter, at a roughly linear rate | Backwards from the required activity number, then checked against your market size and sending ceiling | One ramp period, roughly 8 to 12 weeks from a standing start | Pipeline falls within a ramp period. This is the fastest and most visible damage, which is why it is cut last |
| Experiments | A new segment, geography, channel or offer, run at honest scale | Information, measured in decisions made, not deals closed | By the number of decisions you need to make this year, not by a percentage | Two quarters before the decision is available | You lose information you would have had in two quarters. Nothing breaks now, which is why it is cut first and why that is often wrong twice over |
| Compounding assets | Content that gets found and cited, a free tool, a partner programme, data quality | Pipeline later at a falling marginal cost | By publishing cadence you can genuinely sustain, since half a cadence produces nothing | Two to four quarters, then it keeps producing | Nothing measurable for two quarters, then a permanent hole. Restarting from zero costs more than the maintenance you saved, so reduce cadence instead of stopping |
The one row to read twice is the last column of the experiments line. An experiment budget is the only line whose absence is invisible for two quarters and then explains why the company has no second working motion.
How do you calculate a realistic cost per meeting?
Cost per meeting is fully loaded channel cost divided by meetings held, over a period long enough to include the ramp. Almost every cost per meeting figure in circulation uses a different and more flattering denominator, and the differences compound.
Correction 1: held, not booked
A booked meeting that nobody attends cost you exactly as much as one that happened and produced nothing. Where calendar discipline is broken, roughly half of booked meetings die, which means a cost per meeting quoted on bookings can be double the true figure. This is the single largest correction and it is the easiest to make.
Correction 2: fully loaded
Include tooling, data, infrastructure, the salary of whoever runs it, and the fraction of your own time the channel consumes. Founder time is the most expensive input in most early companies and the one most consistently excluded from these calculations because it does not appear on an invoice.
Correction 3: include the ramp
A cost per meeting calculated on month three, after 21 days of onboarding and 4 to 6 weeks of warm up have already been paid for, is a marginal cost, not an average cost. Both are useful and they answer different questions. Use the average when deciding whether to start a channel and the marginal when deciding whether to scale one already running.
Correction 4: include the failures
If you ran four sequences and one worked, the cost of the working sequence includes the three that did not. This is where portfolio thinking rescues a channel that looks expensive per successful campaign, and it is why our derived portfolio figure of roughly 0.05% positive on sends sits an order of magnitude below our 0.5% to 1% working benchmark. The portfolio figure includes every experiment. The benchmark describes the ones we keep.
What do you cut first when the GTM budget gets cut?
Cut experiments first, compounding assets second, and capacity last, with one important exception. The order follows from time to impact: cutting an experiment costs you information you would have had in two quarters, cutting a compounding asset costs you pipeline you would have had in a year, and cutting capacity costs you pipeline in six weeks. When a budget is cut you are almost always solving for the next two quarters, so you protect the line with the shortest path to revenue.
The exception, and it is the one that matters. If a capacity line is running below the kill threshold, cut it first and cut it entirely. Under 0.5% positive on sends after real iteration is a sequence that should be killed rather than trimmed, and the same logic scales up to a channel. Trimming a failing channel by 30% produces a channel that fails 30% more slowly, which is the worst available outcome because it preserves the cost and destroys the volume that might have produced a signal. Kill it, take the budget to the motion that clears the threshold, and revisit the killed one with a different list or offer later.
Three cuts to avoid, in ascending order of self harm. Do not cut warm up or infrastructure spend on a channel you intend to keep running, because you will pay for it twice when reputation has to be rebuilt. Do not cut the reporting layer, because you are about to make several high stakes allocation decisions and you are proposing to make them blind. And do not cut the compounding asset to zero, because restarting from zero costs more than the maintenance you saved. Reduce the cadence instead of stopping it.
How do you decide which motion gets the next unit of budget?
The next unit of budget goes to whichever motion has the highest marginal return, measured with identical fractions across every motion. Identical fractions is the hard part and it is where most allocation arguments are actually stuck.
In practice this means agreeing four definitions in advance and writing them down: what counts as a meeting held, what counts as an opportunity, what counts as sourced by a motion when several touched the deal, and what period you are measuring over. Attribution across touches inside a sequence is a genuinely difficult problem and it belongs to our sibling property MultichannelPros. What belongs here is the coarser and more consequential version: which motion should own which share of revenue, and that one you can settle with a consistent rule applied honestly instead of with a perfect model.
Then apply the same threshold discipline you apply to sequences. A motion producing above its threshold gets more. A motion below its threshold after genuine iteration gets killed rather than nursed. A motion nobody is measuring gets measured before it gets another cent. Most budget allocation meetings are actually definition arguments in disguise, and they resolve in about twenty minutes once everyone is dividing by the same thing. The page on which fraction each rate uses is the reference we hand people for that conversation.
What does this look like across the group?
Inside the agency this site belongs to, where we run outbound for 36 active B2B clients across 1500+ campaigns, the allocation pattern that holds is unglamorous. Most budget sits in capacity, because clients hire us for pipeline this quarter. A fixed and defended slice sits in angle testing, which is our experiment bucket, run as high volume systematic testing across the full addressable market. And the compounding slice goes into infrastructure and data quality, which is not a growth line on anyone's spreadsheet and is the reason a campaign in month nine outperforms the identical campaign in month one.
The clients who get the most out of the channel are the ones who defend the experiment slice when a good quarter tempts them to move it all into capacity. The clients who churn are, more often than not, the ones who moved everything to capacity in month two, ran the same angle until it exhausted, and concluded the channel was declining. Group churn runs 3 to 5% monthly and that pattern is a visible share of it.
Where does this model stop?
It stops at your own cost inputs, which we do not have and will not guess at, and it misleads whenever cost per meeting is quoted without saying whether the ramp is included. Model the allocation in the pipeline model, which shows required activity by channel so you can compare what each motion has to produce for the budget you are considering. If you want the outbound line of this budget built and run rather than modelled, that is the outbound line built and run for you, month to month, with the reporting layer included.
Frequently asked questions
How do I split GTM budget across paid, content and outbound?
Split by bucket before you split by channel: capacity for pipeline now, experiments for information, compounding assets for declining marginal cost later. Then allocate inside the capacity bucket by cost per meeting held, and inside the compounding bucket by which asset your buyers actually consume. A channel level split decided before the bucket level split is how companies end up fully funded on tactics and structurally unable to build an advantage.
How do I work out a realistic cost per meeting?
Fully loaded channel cost divided by meetings held, over a period that includes the ramp, with failed experiments counted in. Then produce two versions, an average including setup and a marginal excluding it, because you need the first to decide whether to start and the second to decide whether to scale. Any single cost per meeting figure without those two qualifiers is being used to sell something.
What do I cut first when budget gets cut?
Experiments, then compounding assets, then capacity, unless a capacity line is below the kill threshold, in which case it goes first and it goes entirely. Never trim a failing channel proportionally. Trimming preserves the cost structure and destroys the volume you needed to get a clean read, which is the one combination guaranteed to teach you nothing.
How many meetings does an SDR need to book to pay for themselves?
Take the fully loaded annual cost of the seat, divide by your average deal gross profit, and that is the number of closed deals. Divide by your opportunity to close win rate for opportunities, then by your meeting to opportunity rate, then by your show rate for meetings booked. The show rate division is the one that surprises people, because at a 50% show rate the required booking number doubles. Run it in the calculator on this site rather than by hand.
Should the outbound budget include the tooling?
Yes, and the data, the infrastructure and the management time. A cost per meeting that excludes tooling is comparing an agency's all in figure against an in house figure that is missing half its costs, which is how in house looks cheaper right up until somebody adds up the annual contracts. Group everything the channel would not need if the channel did not exist.
Do you publish pricing?
No, not on this site and not in ranges. Cost depends on market, volume, channel mix and how much infrastructure already exists, and a range quoted before those are known is anchoring, not information. Pricing conversations happen on a scoping call where the inputs are on the table.
Last updated: 2026-08-06