Stp Marketing

Concept of the skill

What it is: STP marketing is the strategy sequence of segmenting a market, targeting one or more attractive and reachable segments, and positioning the offer for the chosen target relative to alternatives.

Mental model: Treat the market as heterogeneous. First find meaningful buyer/user groups. Then choose which group or groups the organization can serve better than alternatives. Then define the position the offer should occupy for that target and translate it into product, price, channel, sales, and communications implications.

Why it exists: Agents often jump straight to messaging, personas, or a broad "everyone" target. This skill forces the choice logic that makes a marketing strategy coherent: who is meaningfully different, which group is worth focus, and what differentiated value should that group remember.

What it is NOT: It is not standalone product positioning, macro-environment scanning, industry-structure analysis, Blue Ocean value innovation, copywriting, or marketing-mix execution.

Adjacent concepts: market segmentation, target market selection, target audience, customer profile, ideal customer profile, buyer persona, positioning statement, perceptual mapping, go-to-market strategy, differentiated marketing, niche marketing, account-based targeting, TAM/SAM/SOM, mental availability.

One-line analogy: STP picks the people to serve before it writes the message to them.

Common misconception: A persona is not a segment, and a segment name is not a strategy. The segment must be evidence-backed, reachable, and actionable; the target choice must be justified; and the position must be tailored to that target.

STP Marketing

Domain Context

Use STP marketing for go-to-market strategy, campaign strategy, product strategy, nonprofit program planning, B2B service positioning, market research synthesis, brand strategy, sales focus, and review of marketing plans that need a coherent customer-choice logic. Use public, aggregate, or synthetic examples only. Do not include personal data, customer-level records, payment data, secrets, confidential deal details, or private business facts in examples or evals.

STP is strongest when a team has a broad possible market and needs to decide which customer groups matter most and how the offer should be framed for them. It is weaker when the target and competitive frame are already fixed and the task is only to sharpen a product's category/value framing; that belongs to positioning. It is also weaker when the user asks for macro trends, industry attractiveness, or tactical campaign copy.

For B2B contexts, do not force consumer-only variables. Use firmographic, account, buying-center, use-case, maturity, budget, technology-stack (technographic), urgency, decision-process, intent/signal, or job-to-be-done bases when they better explain demand and reachability. For nonprofit or public-sector contexts, replace profit-only language with mission fit, beneficiary need, funding model, reachability, delivery capacity, and measurable outcomes.

For B2B and account-based work, separate the account segment from the buying-center roles. A target account pattern can be "regulated mid-market manufacturers modernizing quality reporting", but the positioning may need different proof for the end user, economic buyer, technical evaluator, procurement reviewer, security/legal approver, and executive sponsor. Do not collapse the account into one persona. If the task is ABM, the named account list is downstream activation; STP still needs the common account pattern, buying trigger, role map, target rationale, and differentiated position.

For product-led contexts, treat usage evidence as one input, not the whole strategy. Segment by privacy-safe in-product behavior, activation path, use-case, team size, feature adoption, collaboration pattern, free-to-paid conversion trigger, expansion signal, or self-serve versus sales-assisted motion when those variables explain different needs or economics. Do not turn raw telemetry clusters into target segments until they have a named job, reachable channel, value hypothesis, and validation plan.

Prefer consented data. Where the organization holds zero-party data (information customers voluntarily share, such as preference-centre selections or stated intent) or first-party data, it is usually the most privacy-safe and most durable segmentation input — though it is not automatically the strongest evidence for every claim (see the evidence ladder in Step 2: observed behavior is the strongest signal for actual behavior, retention, and adoption). Choose the evidence by claim type, and state the data source and consent basis for every segmentation variable.

Coverage

This skill teaches agents to:

1. Define the market, offer, decision, geography, buyer/user boundary, and evidence available before segmenting; size the boundary with TAM/SAM/SOM when the decision turns on whether the market is large enough.
2. Choose segmentation bases that plausibly explain different needs, behaviors, intent, buying processes, willingness to pay, access channels, or outcome requirements.
3. Build segment profiles with evidence, not invented personas, and label each claim by evidence strength and provenance.
4. Test whether segments are measurable, substantial, accessible, differentiable or responsive, actionable, and ethically/privacy-safe.
5. Evaluate target attractiveness using segment size, growth, economics, urgency, fit, competitive intensity, reachability, channel access, mission/objective fit, and organizational capability.
6. Choose a primary target and optional secondary targets, assign each surviving segment a portfolio role, and name resource tradeoffs.
7. Select an appropriate targeting pattern: undifferentiated, differentiated, concentrated/niche, micromarket, or account-based — without over-narrowing past the brand's need for broad availability.
8. Write a positioning statement tied to the chosen target, category or competitive frame, key need, differentiated value, proof, and primary alternative.
9. Translate the position into implications for product, pricing, channels, sales motion, content, campaign briefs, customer success, and research.
10. State evidence gaps, validation tests, and monitoring triggers, and name the fastest responsible test for each weak link.
11. Treat AI clusters, analytics cohorts, ad-platform audiences, and synthetic personas as hypotheses to validate, not finished segments.
12. Route adjacent tasks to the right framework instead of stretching STP beyond its mechanism.

Philosophy of the skill

STP is useful because marketing strategy is a focus choice. A broad market usually contains groups with different needs, buying triggers, adoption barriers, budgets, channels, and decision criteria. If the agent skips segmentation, it tends to write "for everyone" messages that fit nobody well. If it skips targeting, it treats all segments as equally worth serving. If it skips positioning, it leaves the chosen target without a memorable reason to choose the offer.

The method is sequential but iterative. Bad targeting can reveal weak segmentation. A position that cannot be made credible can reveal a poor target choice. A channel constraint can make an otherwise attractive segment unreachable. The agent should move through segmentation, targeting, and positioning in order, but it should loop back when the evidence shows that an earlier choice does not hold.

STP also has an ethics and privacy boundary. Segmentation can become discriminatory, manipulative, or privacy-invasive when it relies on sensitive attributes, hidden inference, or customer-level data without a legitimate and safe use. Use aggregate or synthetic examples, prioritize explicit zero-party or first-party data, avoid protected or sensitive targeting unless the purpose is clearly lawful and beneficial, and state when a segmentation basis should be replaced by need, behavior, or context instead.

STP should create useful focus, not strategic tunnel vision. Narrow segments can improve relevance, but hyper-segmentation becomes harmful when it fragments reach, ignores category buyers, or makes the brand hard to remember and hard to buy. Large-scale brand-growth evidence (Ehrenberg-Bass) shows that brands grow mainly by reaching more category buyers and being easy to remember and easy to buy — that is, by building mental availability (easy to think of in relevant buying situations) and physical availability (easy to find, evaluate, and buy through the channels the target uses) — not by ever-tighter targeting. A narrow beachhead is acceptable when resources are constrained or learning is the goal, but it still needs buying situations, reach channels, distinctive proof, and an expansion logic. If a proposed target is too small, too hidden, or too operational to build mental and physical availability, downgrade it to an activation cohort or validation hypothesis rather than making it the whole STP strategy. This is a tension to balance, not a rule that overrides targeting: differentiated and concentrated strategies remain valid — the test is whether narrowing is earned.

Working with AI, analytics, and platform tooling

AI research tools, analytics clustering, customer-data platforms, CRM scoring, ad-platform audiences, and persona generators are evidence and execution surfaces for STP, not substitutes for it. They can collect data, cluster audiences, draft personas, draft positioning lines, suggest audiences, optimize bids, personalize creative, and activate target lists — but they do not perform the strategy judgment STP teaches: defining the market boundary, choosing bases that explain real differences, testing segment quality, choosing a target under resource constraints, tying positioning to the chosen target and competitive alternative, and stating what is still unproven.

Treat every machine-produced cluster, lookalike, predicted score, ad audience, or persona as a hypothesis until it has four things:

1. A human-readable segment definition tied to a need, job, trigger, behavior, buying process, or value driver.
2. Evidence provenance: which data or research produced it, how current it is, and what it excludes.
3. Reachability and actionability: how the organization can reach, serve, sell to, or learn from it without unsafe data use.
4. A positioning implication: what should change in product, price, channel, sales, proof, or message because this segment exists.

Two specific traps to guard against:

• AI-generated / synthetic personas are unvalidated by default. A persona produced by an LLM or a "synthetic respondent" tool approximates patterns already present in its training and inputs; it is not field evidence. LLM-derived personas systematically under-represent marginalized, older, and accessibility-affected populations, and "bias laundering" — presenting an inferred persona in the language of empathy and research — makes the missing validation hard to spot. If synthetic personas are used, disclose the data sources, mark what was inferred versus observed, label them synthetic, and pair them with at least one real-user or market test before they drive a target choice. Do not present simulated quotes, needs, objections, or behaviors as if they came from real people. The defensible mode is hybrid: fast synthetic exploration plus real-world validation, with a methods note.
• Reachability is not what the old ad stack assumed. Do not infer that a segment is accessible from third-party-cookie audiences or interest taxonomies that no longer exist. Google retired the advertising-related Privacy Sandbox APIs (Topics, Protected Audience, Attribution Reporting, and others such as Private Aggregation) in October 2025 — while non-advertising Privacy Sandbox technologies (CHIPS, FedCM, Private State Tokens) continue — and reversed its third-party-cookie deprecation plan in April 2025, leaving the durable industry shift toward consented first-party data, contextual reach, and direct/owned channels. A segment that was reachable through a detailed interest, third-party-cookie audience, lookalike, or custom list last year may now require first-party data, consented lists, contextual targeting, broader automated optimization, or a different channel. Check reachability against the channels and consented data the organization can actually use now, not against a deprecated ad-targeting capability.

Do not let platform language replace strategy language. A Google Performance Max audience signal, AI Max query expansion, Meta Advantage+ audience suggestion, CRM propensity score, or CDP cohort can help discover or activate demand, but it is not automatically a strategic segment. Use platform reports to update the evidence, then rerun the segment-quality and target-choice tests.

AI search, answer-engine optimization (AEO), and generative-engine optimization (GEO) are downstream visibility and channel implications of a chosen position, not a replacement for segmentation or target choice, and not an STP failure mode to design around. If AI-mediated discovery is a real reach channel for the target, you may check whether the target-specific position, proof, and product/business data are available through crawlable, people-first, source-backed content — but route GEO/AEO/SEO execution to the relevant SEO/content skill. Do not use GEO/AEO hacks, inauthentic mentions, or special machine-readable files as a substitute for a credible position or real evidence.

Workflow

1. Frame the STP decision

Start by naming what the analysis must decide. Do not segment before the market and offer are clear.

Offer, product, service, or program:
Market or category being considered:
Geography or channel boundary:
Buyer/user/customer boundary:
Decision this STP must inform:
Evidence available:
Constraints: budget, channel, regulation, capacity, ethics/privacy:
Current hypothesis, if any:

If the user gives no evidence, produce a sparse STP scaffold and mark assumptions. Do not invent segment sizes, needs, willingness to pay, or channel behavior.

Size the market boundary before segmenting (TAM/SAM/SOM). When the decision turns on whether the market is large enough to justify the effort, bound it with three nested estimates: TAM (Total Addressable Market — everyone who could conceivably need the offer), SAM (Serviceable Addressable Market — the portion reachable with your model, geography, channels, and right to operate), and SOM (Serviceable Obtainable Market — the share you can realistically win given competition, differentiation, and capacity). Use honest proxies and label every assumption; do not manufacture precise figures. The number a target choice should be tested against is SOM, not the headline TAM — an attractive TAM with a tiny reachable SOM is a common trap.

2. Segment the market

Choose segmentation bases that explain real differences in need, behavior, reachability, or value. Do not use every base. Use the few that matter for the decision.

Needs/jobs and JTBD, used rigorously. A "needs" basis is strongest when it names the job the buyer is trying to get done and the outcomes by which they judge success, not a demographic stand-in. Jobs-to-be-done (JTBD) and needs-based segmentation are complementary, not competing: JTBD qualitatively articulates what the customer is trying to accomplish and under what circumstances; needs-based segmentation then quantifies, clusters, and operationalizes those jobs into addressable segments. Outcome-Driven Innovation (ODI) is the most rigorous form — it segments by under-served desired outcomes rather than by who the buyer is. Reach for a job/outcome basis when demographic or firmographic groups keep showing the same buying behavior for different reasons, or different behavior for the same reason; that mismatch is the signal that the real segmenting variable is the job, not the buyer's traits.

For B2B, use a nested approach when appropriate — start with broad demographics/firmographics, then layer operating variables (technology, usage rate), purchasing approach (centralized vs decentralized), situational factors (urgency, application), and personal characteristics (risk tolerance, values). This multilayer approach, derived from Shapiro and Bonoma's nested-model research, prevents single-variable oversimplification.

For each segment, create a profile:

Segment name:
Human-readable definition:
Defining criteria (bases):
Primary difference: need / job / trigger / behavior / buying process / value driver:
Evidence:
Evidence provenance (which data/research produced it; how current; what it excludes):
Evidence strength: verified / source-supported / inferred / synthetic hypothesis / unverified
Size or proxy:
Growth or urgency:
Stability: durable segment / event-triggered / lifecycle stage / dynamic operational cohort:
Reach channels:
Buying/adoption barriers:
Economic or mission value:
B2B roles, if relevant: end user / economic buyer / technical evaluator / procurement / influencer / executive sponsor:
Competitive alternatives:
Positioning implication:
Mental/physical availability implication:
Ethics/privacy concerns:
Confidence:

Name segments by the difference that matters, not by a convenient descriptor. "Mid-market finance teams with manual reconciliation risk" is stronger than "SMB women 25-44" because it names a job, context, and decision implication. Demographic, geographic, firmographic, and account attributes are acceptable when they explain need, access, regulation, buying process, or economics; otherwise they are only descriptors and should not carry the strategy.

Keep strategic segments stable enough to guide positioning. Real-time propensity scores, churn-risk flags, campaign-engagement cohorts, and AI-discovered microclusters can be useful operational subsegments, but they are too fluid to carry the whole positioning unless the underlying need or buying context is stable.

Label evidence by strength — not all evidence is equal. When you fill in Evidence:, Evidence strength:, and Confidence:, name where each segment claim sits on this ladder, strongest first:

Stamp each segment claim with one evidence label — verified, source-supported, inferred, synthetic hypothesis, or unverified. A target choice resting on rungs 5–6 alone must be marked provisional and paired with a validation test before it drives investment. State the rung; do not let a rung-6 inference read as a rung-1 fact. Note also that rung-1 observed behavior can over-represent current customers and current channels — it does not automatically capture future or noncustomer demand.

3. Test segment quality

Before choosing a target, test each segment against quality criteria.

Merge segments that do not behave differently. Split segments that hide different jobs, barriers, or channels. Drop segments that cannot support an action.

Run a privacy, fairness, and current-channel check before target selection:

Personal data used:
Consent / lawful basis / permission state:
Sensitive or protected attributes involved:
Is the attribute necessary for a lawful, beneficial purpose?
Can a need, behavior, context, or channel proxy replace a sensitive label?
Could the segment exclude, exploit, or unfairly burden a protected group?
Current platform/channel availability:
Measurement limitation or attribution caveat:
Required safeguard:

If the segment relies on protected traits, inferred sensitive attributes, customer-level records, or platform-restricted categories, do not proceed as if it is ordinary targeting. Either replace the basis with a safer need/behavior/context variable, narrow the use to a clearly beneficial and compliant purpose, or route the compliance question outside STP before acting. STP can flag the risk; it is not legal advice.

4. Choose targets

Evaluate each segment before choosing. Do not let a favorite persona win without evidence.

A compact restatement many teams use for a first cut is Size, Profitability, Accessibility: is the segment large enough to matter, will lifetime value and ROI justify the acquisition cost, and can you legally and technically reach it. Treat that triad as a screen, not a substitute for the fuller table above — fit, timing, strategic value, and risk routinely flip a target choice that the three-criterion screen would have passed.

As a quantitative economics check, the LTV:CAC ratio is a common heuristic — a segment whose customer lifetime value to acquisition cost is roughly 3:1 or better is generally easier to sustain, and a much lower ratio is a warning sign. Treat it as a guideline, not a universal gate: it does not fit nonprofit, public-sector, or mission-value contexts, where you substitute mission-value-per-beneficiary for LTV and delivery-cost-per-beneficiary for CAC, and even in commercial contexts a strategic beachhead may justify weaker near-term economics.

Do not over-narrow past the brand's need for broad availability. The focus logic above has a counterweight from large-scale brand-growth evidence (Ehrenberg-Bass): brands grow mainly by reaching more category buyers and being easy to remember and easy to buy, not by ever-tighter targeting. A target so narrow that the brand becomes hard for the wider category to recall, find, evaluate, or buy trades away mental and physical availability — the very assets that drive growth — for a precision the business may not need. Concentrated, niche, or account-based targeting is right when resources, fit, a beachhead, or a genuinely distinct need demand it; it is wrong when it is just an excuse to talk to fewer people or to dress narrow ambition up as strategy. When you recommend a focused target, confirm the focus is a deliberate resource/fit/beachhead choice and that the chosen target still leaves the brand easy to remember and easy to buy among the buyers who matter.

Assign a portfolio role, not just a rank. Real STP output is usually a portfolio of segments with different jobs, not a single pick. Give each surviving segment one explicit role:

Do not choose multiple primary targets unless the organization can fund distinct offers, channels, proof, and positioning for each. A differentiated targeting strategy is a resource commitment, not a bigger table.

Name the fastest responsible validation for every weak claim. For each segment resting on evidence rungs 5–6, an unproven reachability assumption, or an untested willingness to pay, state the fastest responsible test that would confirm or kill it — a handful of customer interviews, a smoke-test landing page, a small paid-reach test on a channel you actually control, or a concierge/manual delivery to one account — and the result that would change the target choice. "Fastest responsible" means the cheapest test that still produces real-world signal; do not let an attractive-sounding segment advance to positioning on unvalidated rungs without naming its next test.

Then state:

Primary target:
Why this segment wins:
LTV:CAC estimate or mission-value proxy:
Beachhead, if any (and expansion path):
Secondary target, if any:
Why it is secondary:
Watchlist segments (and promotion trigger):
Segments rejected and why:
Targeting pattern: undifferentiated / differentiated / concentrated / micromarket / account-based
Resource tradeoff:
Weakest claim + fastest responsible validation:

Use expected-value when the user has quantified outcomes and probabilities. Use porters-five-forces when the main uncertainty is industry profit pressure. Use pestel when the target choice depends on macro forces such as regulation, social change, or economic conditions — and reference the relevant macro factors explicitly when they materially affect segment attractiveness (e.g., regulatory tailwinds, inflation-driven budget shifts, sustainability mandate changes).

5. Position for the chosen target

Positioning comes after target choice. Write the position for the primary target first; secondary targets may need separate positioning.

Use a perceptual map when the target's choice criteria are unclear or when competitors cluster around the same claims. Choose two attributes the target actually values, not the axes the team wishes mattered. Plot the main competitive alternatives, mark the current or intended offer, and look for positions that are open, valuable to the target, credible for the organization, and defensible against alternatives. Do not use perceptual maps as decoration: if the axes do not come from customer evidence or the map does not change the positioning statement, skip it.

Consider which positioning strategy type fits the target and competitive context:

Use this statement shape:

For [target segment]
who [need/job/problem],
[offer] is a [category / frame]
that [differentiated value]
because [proof / reason to believe],
unlike [primary alternative].

Then test it:

If the task is only to choose a category, competitive alternative, value themes, or product narrative after the target is already fixed, route to positioning.

5b. Validate the STP logic

Before turning the strategy into execution, name the fastest responsible validation for each weak link.

When evidence is thin, present the STP as a learning plan with hypotheses and tests. Do not turn an unvalidated segment into a confident campaign brief.

6. Translate into implications

STP should change decisions. End with implications and next evidence.

Where relevant to the context (service, nonprofit, people-intensive B2B), extend implications beyond the classic 4Ps to include People (staff, skills, training), Process (ordering, onboarding, support workflows), and Physical evidence (tangible cues, facilities, credentials).

Do not finish with only a table. Finish with the few choices the STP analysis supports and the evidence still missing.

Output Template

STP Marketing

Decision:
Market boundary:
Offer:
Geography/channel boundary:
TAM / SAM / SOM (if sized):
Evidence used:
Evidence missing:
Tooling/audience inputs used, if any:

Segmentation
1. Segment:
   Definition:
   Criteria:
   Primary difference:
   Need/job/trigger:
   Evidence provenance:
   Evidence strength: verified / source-supported / inferred / synthetic hypothesis / unverified
   Size/value proxy:
   Stability:
   Reachability:
   B2B roles, if relevant:
   Privacy/fairness risks:
   Positioning implication:
   Mental/physical availability implication:

Targeting
Primary target:
Why this segment wins:
LTV:CAC estimate or mission-value proxy:
Target role: primary / beachhead / secondary / watchlist / rejected
Rejected/secondary segments:
Targeting pattern:
Resource tradeoff:
Current reachability caveat:
Weakest claim + fastest responsible validation:

Positioning
For:
Who:
Offer/category:
Differentiated value:
Proof:
Alternative:
Positioning strategy type:
Role-specific proof, if B2B:

Implications
Product/service:
Price/funding:
Channel:
Promotion/message:
Sales/partnership:
Customer success/retention:
People/process/physical evidence (if relevant):
Research/validation:
AI search / agentic discovery, if relevant:
Monitoring triggers:

Boundary Rules

…(truncated)