Building Your Growth Hypothesis Tree

Most growth planning at Series A and growth-stage B2B companies produces a tactical list. The team sits in a room, generates ideas, prioritises them against each other, and ends up with a backlog of initiatives for the next quarter. The backlog looks like a plan. It isn't. What it's missing is the structure that connects the tactical initiatives to what the business actually needs to be true for growth to work.

A growth hypothesis tree is that structure. It's a decomposition of your top-level growth goal into the specific hypotheses about the world that would have to be true for the goal to be achievable. Done well, it makes visible which hypotheses you already have evidence for, which you're assuming without evidence, and which tactical initiatives would actually test the risky assumptions versus just producing motion. This article is about building one, reading one, and using it to change how growth planning works.

What's Wrong With The Tactical List

The tactical list has three structural problems that only become visible when you try to use it for actual growth operations.

It doesn't explain why any initiative matters. A list of 15 initiatives, launch LinkedIn, rebuild landing pages, hire an SDR, write more content, add retargeting, is a list of things to do. It doesn't explain what each thing is supposed to produce, what would happen if you did it and it worked, or why this particular set of things was chosen over another set. Without the connection to goals, prioritisation becomes arbitrary and post-hoc rationalisation.

It can't distinguish between testing and executing. Some initiatives are testing something, running a new channel to see if it works for this ICP. Other initiatives are executing a known pattern, scaling a channel that's already proven. These require different operating disciplines, different success criteria, and different investment levels. A flat list treats them identically.

It doesn't surface what the plan is actually betting on. Behind every growth plan are specific assumptions about the world: this ICP exists in reach, this channel works for it, this message resonates, this product is good enough, this price is right. If any assumption is wrong, the dependent initiatives fail. A tactical list doesn't make the assumptions visible, so the team can't see which bets they're actually making.

What The Hypothesis Tree Produces Instead

A hypothesis tree decomposes the top-level goal into a cascade of hypotheses. Each hypothesis is a specific, testable claim about the world that would have to be true for the level above to work. The tree makes three things visible that the tactical list hides.

Which claims are load-bearing. Some hypotheses in the tree support many others. If they're wrong, large parts of the plan collapse. Other hypotheses are relatively isolated, if they're wrong, only specific initiatives fail. The tree shows which hypotheses matter most, which directs where testing attention should go.

Which claims have evidence and which don't. For each hypothesis, you can mark whether you have strong evidence, weak evidence, or no evidence. The tree makes visible where your plan is resting on unsupported assumptions. These are the places to test first, because they're where the plan is most likely to fail.

Which initiatives actually test which hypotheses. Every initiative in the plan maps to one or more hypotheses it tests or executes against. Initiatives that don't map to any hypothesis in the tree shouldn't be running, they're activity without strategic connection. This mapping enforces discipline on prioritisation.

How To Build A Hypothesis Tree

The structural steps to produce a working tree. Takes 2-3 hours with the right people in the room.

Step 1: Name the top-level goal

One sentence. Not "grow faster" but "reach £5M ARR by end of next year." The goal has to be specific enough that you can tell whether you hit it, and the hypothesis tree has to reach it.

Step 2: Decompose into the main levers

What has to be true at a high level for the goal to work? For most B2B growth goals, this decomposes into roughly four levers: we can reach our target audience with acceptable economics; we can convert the reached audience at acceptable rates; we can retain customers once acquired; our ACV supports the economics of the whole system.

Each of these is a hypothesis. Each one either has evidence for it or doesn't. Each one further decomposes.

Step 3: Decompose each lever into testable sub-hypotheses

"We can reach our target audience with acceptable economics" is a big claim. Break it into specific, testable sub-hypotheses. Something like:

• LinkedIn ads produce MQLs at CPL ≤ £250 for our ICP
• Google Search produces MQLs at CPL ≤ £200 for our ICP
• Content marketing produces organic MQLs that close at better rates than paid
• Our ICP can be reached at the required volume given our total addressable market

Each of these is individually testable. Each has an evidence status (proven, weakly evidenced, assumed). Each maps to specific potential initiatives.

Step 4: Mark evidence status for each hypothesis

For every hypothesis in the tree, honestly mark: strong evidence (we've proven this), weak evidence (we have directional signal), or assumption (we don't really know this is true, we're assuming it).

This is the step that most reveals where your growth plan is actually risky. Teams consistently discover that 20-40% of the hypotheses their plan depends on are marked "assumption." Those are the places where the plan is most likely to fail, and the places that should be tested first.

Step 5: Map initiatives to hypotheses

For each planned initiative, name which hypothesis in the tree it tests or executes against. Initiatives that don't map to any hypothesis should either be removed or you need to add the missing hypothesis they imply.

This step produces the cleanup effect. Tactical initiatives that seemed important often turn out to have no hypothesis connection, they were included because they felt productive, not because they'd test or execute against anything that matters.

Reading The Tree To Set Priorities

Once the tree is built, three specific questions tell you how to prioritise.

Which hypotheses are load-bearing and unevidenced? These are the highest-priority tests. If you don't know whether LinkedIn works for your ICP and your whole paid media plan depends on it, testing LinkedIn comes before testing anything downstream of it. Running downstream initiatives before validating the load-bearing assumption is building on sand.

Which hypotheses are evidenced and ready to execute? These are the places to scale rather than test. If content marketing has demonstrably worked for you for three quarters, the initiative isn't "should we do more content", it's "how do we produce more at the quality that's been working."

Which hypotheses are weakly evidenced and worth strengthening? These are the places for smaller confirming tests. Not large new investments, but small, cheap tests that either strengthen the evidence or surface the hypothesis as false.

The quarterly planning conversation changes. Instead of "which tactics should we run," it becomes "which hypotheses are we testing, which are we executing against, and what's the evidence status we want to reach by quarter end." This is a more strategically productive conversation.

A Worked Example

A simplified hypothesis tree for a Series A B2B SaaS company targeting £3M ARR in the next year.

Top goal: £3M ARR by end of next year

Main levers (with evidence status in brackets):

• We can reach mid-market B2B CMOs at scale [weak evidence]
• Our positioning resonates with this segment [strong evidence]
• Our close rate on reached prospects is ≥20% [weak evidence]
• Retention at 12 months is ≥85% [strong evidence]
• ACV sustains positive unit economics [strong evidence]

Decomposing "we can reach mid-market B2B CMOs at scale":

• LinkedIn Ads produce qualified MQLs at CPL ≤£250 [assumption]
• Google Search produces qualified MQLs at CPL ≤£200 [strong evidence, current]
• Content-led inbound produces qualified MQLs at scale within 18 months [assumption]
• Outbound produces qualified meetings at acceptable cost [weak evidence]

Looking at this: close rate, content-led inbound, and LinkedIn Ads are all unevidenced or weakly evidenced assumptions that the plan depends on. These are the places to test first. Google Search is proven, the initiative here is scale, not test. Retention, positioning, and unit economics are solid, the team doesn't need to worry about them in the next quarter.

The tactical initiatives fall out of this cleanly. The highest-priority tests are validating the assumptions. The secondary initiatives scale what's proven. Initiatives that don't test or scale anything in the tree, generic "improve our brand" work, speculative new channels with no thesis, get deprioritised.

Where Hypothesis Trees Sit in the NBT Growth Model

Hypothesis tree work is a Stage 2 Formulation activity, part of defining the growth route the business will pursue. It's more rigorous than a standard planning exercise because it forces the logical structure underneath the plan into the open.

The relationship to the strategy-execution gap is direct. The gap between strategy and execution often exists because the strategy is expressed as abstract goals and the execution is expressed as tactical initiatives, with no connecting structure between them. The hypothesis tree is the connecting structure. Each hypothesis is specific enough to link to initiatives; each group of hypotheses rolls up to the strategic goals.

The relationship to test-and-learn that compounds is also direct. The tree tells you which hypotheses to test; the test-and-learn discipline tells you how to run tests that compound. Together they produce a growth operation that knows what it's betting on and systematically reduces the unproven bets.

If your growth plan currently lives as a tactical list and you're not sure which initiatives actually matter, or you want help decomposing your goal into a testable structure, get in touch for a strategy call. We'll walk through the tree-building exercise with you and identify where the assumptions are.