The Assumptions That Could Kill This Company
Every startup dies from an assumption it didn't test. Here are the 10 highest-risk beliefs underpinning TELEGENT — each with a falsifiable experiment, quantitative success threshold, and a mitigation strategy if the assumption proves false. This is a living document.
Where TELEGENT Could Break
Categorized by domain. Each assumption is ranked by two axes: how likely it is to be false (uncertainty) and how much damage it would cause if false (impact).
Is there a real, urgent problem people will pay to solve?
#1, #2, #3Can we actually build something that delivers measurable value?
#4, #5Will the technical approach scale and remain defensible?
#6, #7Can we capture enough value to sustain and grow?
#8, #9Can the team execute against an ambitious timeline?
#10Impact × Uncertainty Matrix
Does the Market Actually Need This?
The three assumptions that matter most. If any of these is wrong, nothing else matters.
Mid-market and enterprise organizations experience quantifiable "revenue leakage" (15–35% of potential revenue) that executives recognize as a problem they need to solve.
If revenue leakage isn't a recognized, urgent problem — if executives see it as 'just how business works' or don't believe AI can fix it — there is no market. Every other assumption becomes irrelevant.
Run a "Problem Validation Sprint". Conduct 40 structured discovery interviews with CEOs/COOs of target organizations ($5M–$100M revenue). Do not pitch the product. Only ask: "Walk me through how you track missed revenue opportunities today. What happens to a lead that comes in after hours? How do you know what you are losing?" Score responses on a Problem Awareness Rubric (1–5).
SUCCESS CRITERIA≥28 of 40 (70%) score 4 or 5 on Problem Awareness. ≥20 spontaneously mention revenue leakage before being prompted. ≥15 can quantify rough dollar amounts they believe they're losing.
Pivot to a narrower, more acute pain point: emergency response time compliance, missed call rate reduction, or a single-vertical focus where the problem is undeniable (e.g., behavioral health admissions where every missed call is a lost patient). Build a point solution for one metric before expanding to the full platform.
Organizations will pay $4,997+/month for a platform that quantifies and recovers revenue leakage — the value of recovered revenue clearly exceeds the platform cost in the buyer's mind.
This is the single most dangerous assumption. Many tools that 'pay for themselves' fail to close because the buyer can't mentally connect the price to the value before purchasing. If willingness-to-pay is actually $997–1,997/mo, the unit economics break at current CAC assumptions.
Run a Van Westendorp Price Sensitivity Meter with 50 qualified prospects who've completed the Business DNA™ Assessment. Ask four questions: (1) At what price would this be so expensive you'd never consider it? (2) At what price would you think it's expensive but still consider it? (3) At what price would you think it's a bargain? (4) At what price would you question its quality? Plot indifference price point and optimal price point. Follow with 10 structured 'buy/no-buy' interviews at the optimal price.
SUCCESS CRITERIAOptimal price point (OPP) ≥ $3,500/mo. Indifference price point (IDP) ≥ $4,200/mo. ≥6 of 10 buy/no-buy interviews say 'yes' at OPP.
If OPP is $1,500–3,500/mo: restructure to a usage-based model (per-location pricing, per-lead pricing) that scales with value. If OPP <$1,500/mo: the business is a high-volume SaaS play requiring different CAC and sales motion. Consider a product-led growth model with a free tier converting to $199–499/mo per seat.
The target buyer (CEO/COO of $5M–$100M organizations) understands the "revenue intelligence" category well enough to evaluate and purchase without excessive education. The category does not require invention.
Creating a category costs $5–20M+ and 3–5 years. If TELEGENT has to simultaneously invent the 'revenue intelligence' or 'business impact intelligence' category AND build the product, the capital requirements are 3–5× current assumptions. Most category-creating startups run out of money before the category exists.
Run 20 'category comprehension' tests. Show target buyers the TELEGENT homepage and value proposition for 90 seconds. Then ask: 'In your own words, what does this product do?' Score responses on a Comprehension Rubric (1=completely confused, 5=perfectly articulated the value). Also test 3 alternative positioning frames: (A) 'Revenue Recovery Platform', (B) 'AI Operations Command Center', (C) 'Business Impact Intelligence'. Measure which frame produces highest comprehension.
SUCCESS CRITERIA≥14 of 20 (70%) score 4 or 5 on comprehension within 90 seconds. One positioning frame clearly outperforms others (≥25% higher comprehension). ≤3 buyers say 'I don't understand what this is.'
If comprehension is low: do NOT create a new category. Anchor to an existing, understood category: 'Autonomous Revenue Operations' or 'Automated Lead-to-Revenue Platform.' Use the existing category as the hook, then differentiate within it. If one positioning frame dominates, adopt it immediately across all marketing. If no frame works, the product description itself may need simplification before market launch.
Can We Actually Build Something That Works?
Technical feasibility meets measurable value delivery. The scoring systems must actually mean something.
The proprietary scoring systems (ODS™, TVS™, RDS™, IAS™, OMS™, OPS™) actually correlate with real business outcomes. A high Revenue Recovery Score means the organization genuinely recovers more revenue — not just that our algorithm produces a high number.
If our scores don't correlate with real outcomes, we're selling an expensive dashboard that looks sophisticated but doesn't actually drive decisions. Customers will figure this out within 90 days. Churn will be catastrophic. Worse: if scores are inversely correlated (high score → bad outcomes), the product actively harms customers.
Run a Prospective Validation Study. Deploy the full scoring system with 10 beta customers. At deployment (T0), measure all 6 scores. At T30, T60, T90, measure ACTUAL revenue recovery, cost savings, and capacity created — using the customer's own financial data, not our platform's numbers. Calculate Pearson correlation coefficient between each score and its corresponding real-world outcome. Also track: do customer decisions change based on scores? If BIIQ™ goes from 42 to 68, does anything different happen in the business?
SUCCESS CRITERIACorrelation coefficient r ≥ 0.65 for at least 4 of 6 scores with their corresponding real-world outcomes (p < 0.05). ≥7 of 10 customers show directional improvement in BIIQ™ that matches directional improvement in actual business metrics. Zero inverse correlations. At least 5 of 10 customers can point to a specific operational change they made based on a score insight.
If correlations are weak (r < 0.40): simplify the scoring models to fewer, more directly measurable inputs. Replace algorithmic scoring with direct measurement (e.g., instead of 'predicted revenue leakage,' show 'calls missed / calls received × average deal size'). If one or two scores have strong correlation but others don't: spin those scores down and focus on the ones that work. If correlations are zero or inverse: kill the scoring system entirely and pivot to a pure measurement/reporting platform.
We can build deep integrations with CRM, phone, scheduling, and ERP systems fast enough to deliver value within 30 days of customer onboarding. Integration breadth and depth can be achieved without a 20-person integrations team.
Enterprise integrations are the graveyard of B2B SaaS startups. Every customer's stack is different. Salesforce alone has 5+ major editions with different APIs. Phone systems have inconsistent webhook formats. If integration takes 60–90 days per customer, the sales motion breaks (no one signs a $5K/mo deal that takes 3 months to go live), and services costs eat all margin.
Run a 'Time-to-First-Value' sprint with 8 beta customers across 4 different CRM/phone stack combinations. Track: (1) total engineering hours to achieve 'minimum viable integration' (data flowing bidirectionally for one core metric), (2) calendar days from contract-sign to first dashboard with real data, (3) number of engineering escalations per integration, (4) customer-reported 'integration friction' on a 1–5 scale. Also build a 'connector complexity matrix' classifying each integration by API maturity, documentation quality, and edge cases.
SUCCESS CRITERIAMedian time-to-first-value ≤ 18 calendar days. ≥6 of 8 customers achieve first value in ≤25 days. Mean engineering hours per integration ≤ 40 hours. ≤2 engineering escalations per integration requiring custom code. Connector complexity matrix identifies at least 4 'low-complexity' connectors suitable for self-serve onboarding.
If integration takes too long: prioritize a 'Bring Your Own API Key' model where customers self-connect supported systems via OAuth. Build 3 deep, bulletproof integrations (Salesforce, HubSpot, RingCentral) and publish API specs for everything else. Offer a 'concierge integration' add-on at $2,500 one-time for non-standard stacks. If even deep integrations take too long: consider an integration middleware partner (Workato, Tray.io, Paragon) instead of building natively — trades margin for speed.
Will the Technology Actually Scale?
Architecture decisions made now compound in year 2. The Knowledge Graph and ML pipeline are the highest technical risks.
A graph database (Neo4j/ArangoDB) can power the Knowledge Graph at production scale — handling millions of entities, billions of relationships, and sub-second query performance — without becoming a cost or operational nightmare.
Graph databases in production are notoriously difficult to operate. Query performance degrades non-linearly as the graph grows. Graph partitioning is still an unsolved problem in distributed systems. If the Knowledge Graph requires a dedicated 3-person team to operate, or if query latency exceeds 2 seconds for common traversals, the scoring and intelligence layers built on top of it become unusable. Most startups that bet on graph DBs eventually migrate to relational + application-level graph logic.
Build a scale-test harness that simulates 36 months of growth in 2 weeks. Start with a seed graph of 100K entities/1M relationships (representing 50 customer deployments). Scale to 10M entities/500M relationships (representing 1,000 deployments). At each order-of-magnitude checkpoint, run the 10 most common query patterns (entity resolution, relationship traversal, subgraph extraction for scoring). Measure P50, P95, P99 latency, memory usage, and query cost. Test with both Neo4j AuraDB Professional and a self-hosted setup.
SUCCESS CRITERIAP95 query latency ≤ 500ms for top 10 query patterns at 10M entity scale. P99 latency ≤ 2s. Monthly infrastructure cost at 10M entity scale ≤ $2,500. Zero query patterns that timeout (>30s). Backup/restore completes in ≤4 hours. At least 3 of top 10 query patterns are index-covered without full scans.
If graph DB doesn't scale: implement a hybrid architecture — relational DB (PostgreSQL) for entity storage + application-level adjacency lists for relationship queries + Redis for caching frequent traversals. Pre-compute scoring subgraphs as materialized views updated hourly. If even hybrid doesn't work: reduce the Knowledge Graph's scope to a simpler entity-linkage model (contact → company → interaction chain) rather than a full property graph, and defer complex graph analytics to Year 3.
ML models for opportunity detection, lead scoring, and revenue leakage prediction can achieve sufficient accuracy with the training data available from 50–100 initial customer deployments — without needing hundreds of thousands of labeled examples.
ML models trained on small datasets produce fragile, biased, or inaccurate predictions. If the Opportunity Discovery Score™ is a random number generator, user trust evaporates immediately. However: this is ranked lower-impact because (a) the platform can deliver value with rules-based scoring initially, and (b) the scoring models improve as data accumulates — early inaccuracy is tolerable if the trend is toward accuracy.
Train initial models on synthetic data generated from the Business DNA™ assessment patterns (already collected from 200+ assessments). Validate against manual expert scoring: have 3 domain experts independently score 100 real lead scenarios. Compare model predictions to expert consensus. Calculate Cohen's Kappa for inter-rater reliability between model and experts. Also measure: does model accuracy improve with each additional 10 training examples? Plot the learning curve to estimate data requirements for target accuracy.
SUCCESS CRITERIACohen's Kappa ≥ 0.6 between model and expert consensus for at least 4 of 6 scoring dimensions. Learning curve shows monotonic improvement with additional data (no plateaus below 0.65 Kappa). Expert inter-rater reliability (IRR) ≥ 0.7 (if experts can't agree, model can't be expected to). Model bias audit shows no disparate impact across industry verticals or company sizes.
If ML underperforms: launch with a transparent rules-based scoring engine (customers can see and adjust the rules). Collect labeled data passively for 6–12 months. Introduce ML as a 'beta enhancement' once sufficient training data exists. If specific scoring dimensions resist ML (e.g., Organizational Maturity Score™): use structured assessments (surveys, audits) for those dimensions rather than ML prediction. Accept that 1–2 dimensions may never be ML-driven — that's acceptable if the core dimensions (ODS, OPS) work.
Can We Actually Make Money?
Unit economics determine whether the company is a business or a project. These two assumptions govern survival.
Customer Acquisition Cost (CAC) payback period will be under 12 months at scale, with a CAC:LTV ratio of at least 1:3. The sales motion (Business DNA™ Assessment → Proof Center™ → Sales Call → Close) converts at predictable rates.
B2B SaaS companies with CAC payback >18 months die — they run out of cash before the compounding engine kicks in. If the current 'high-touch consultative' sales motion requires 6–8 touchpoints and 45+ days to close a $5K/mo deal, the fully-loaded CAC is likely $15–25K, producing a 12–18 month payback. That's survivable but painful. If CAC is $30K+ (18+ month payback), the company needs venture-scale capital or dies.
Track a cohort of 30 prospects through the full funnel: Business DNA™ Assessment → Proof Center™ engagement → Sales Qualified → Opportunity Created → Closed Won. Measure conversion rates at each stage. Track fully-loaded CAC: (total sales & marketing spend for the cohort / closed deals). Include SDR salaries, tool costs, demo infrastructure, executive time. Also track: (1) average sales cycle length, (2) average touchpoints to close, (3) win rate by industry vertical, (4) win rate by company size. Interview 5 won deals and 5 lost deals to understand decision drivers.
SUCCESS CRITERIACAC payback ≤ 12 months at $4,997/mo ACV. Lead-to-close conversion rate ≥ 8% (from qualified lead). Average sales cycle ≤ 35 days. At least one clear ICP segment with win rate ≥ 15% (the 'sweet spot'). Lost-deal interviews reveal ≤2 deals lost to 'we don't understand the value' (pricing/competitor/timing losses are acceptable).
If CAC payback is 12–18 months: shift to a product-led motion where the Business DNA™ Assessment converts directly to a self-serve Starter tier at $997/mo with in-product upgrade prompts. Use sales only for Professional ($4,997+) deals. If CAC payback >18 months: the current model doesn't work — pivot to a transaction-fee model (% of recovered revenue) or a lower-touch, higher-volume play at $497/mo with a 10× larger TAM. If one ICP segment has dramatically better economics: focus exclusively on that segment and drop everything else.
Monthly net revenue churn will stay under 5% (ideally under 3%), driven by high switching costs from deep integrations, executive habit formation around the Daily Briefing™, and measurable ROI that makes cancellation irrational.
If monthly churn is 5%+, annual churn exceeds 46% — the business is a leaky bucket. At 3% monthly churn, you lose 30% of revenue annually before expansion revenue. This is the #1 silent killer of B2B SaaS. The moats described in the platform blueprint (integration depth, executive workflow lock-in) are theories until proven with retention data.
Track a cohort of first 30 paying customers for 6 months (or as long as available). Measure: (1) gross monthly revenue churn, (2) net monthly revenue churn (including expansion), (3) logo churn vs revenue churn (are you losing small accounts or big ones?), (4) churn by customer segment (industry, size, deployment duration). Conduct exit interviews with every churned customer. Also track leading indicators: (a) weekly active users on dashboards, (b) Daily Briefing™ open rate, (c) number of integrations connected, (d) NPS at 30/60/90 days. Identify which leading indicator best predicts churn.
SUCCESS CRITERIAGross monthly revenue churn ≤ 3%. Net monthly revenue churn ≤ 1% (expansion offsets losses). At least 3 of 30 customers show expansion revenue (upgrading tier or adding seats). Exit interviews reveal ≤20% of churn is 'product doesn't work' — pricing/budget/fit losses are more acceptable. Leading indicator identified with predictive power (≥70% of churned accounts showed warning signal ≥30 days before churning).
If gross churn is 5–8%: implement a customer success 'time-to-value' sprint at onboarding (goal: measurable ROI within 14 days). Add executive business reviews at 30/60/90 days. If churn is concentrated in a specific segment: stop selling to that segment. If Daily Briefing™ open rate is the leading churn predictor: invest heavily in making it indispensable (personalization, mobile push notifications, Slack/Teams delivery). If churn >8%: the product has a fundamental value-delivery problem. Halt new sales. Fix the product. This is an existential metric.
Can the Team Ship It?
A lean team (≤12 people) can simultaneously build the core platform, maintain 6 scoring models, integrate with 6+ third-party systems, support 50+ customers, run sales, and execute the product roadmap — without burning out or accumulating fatal technical debt.
This is a 'boiling the frog' risk. It doesn't kill the company in month 3 — it kills it in month 18 when the team is exhausted, the codebase is duct tape, customers are complaining about bugs, and competitors who raised $50M are shipping faster. It's ranked lower impact because there are straightforward mitigations (hire more, slow down, cut scope). But the assumption is almost certainly false if the full roadmap is attempted simultaneously.
Run a 'Capacity Stress Test.' Map all committed deliverables for the next 6 months (product features, integrations, customer commitments, sales targets). Assign estimated person-weeks to each. Compare to available person-weeks (team size × weeks). Calculate the 'honest capacity ratio' (available / required). Also run a 'focus audit': for each team member, track their time across categories (building new features, fixing bugs, supporting customers, internal meetings) for 2 weeks. Identify what percentage of time goes to the top-3 priorities.
SUCCESS CRITERIAHonest capacity ratio ≥ 0.85 (we have 85%+ of the people we need for what we've committed). Top-3 priorities receive ≥ 60% of total team time. ≤20% of time is 'unplanned work' (fires, escalations, bugs). No single team member at >110% allocation. At least 2 identified deliverables can be deferred without customer impact. Technical debt backlog is tracked and has a monthly reduction target.
If capacity ratio is 0.60–0.85: defer 2–3 roadmap items by one quarter. Institute a 'no new work on Friday' rule for tech debt reduction. If capacity ratio is <0.60: the plan is impossible. Choose: (a) raise capital to hire to 20+ people, (b) cut scope to the 3 things that matter most, or (c) extend the timeline 6–9 months. The wrong answer is 'try harder.' If unplanned work exceeds 30%: invest in quality and automation (better test coverage, better monitoring, better documentation) to reduce firefighting before starting new features. If a single person is a single point of failure for any critical system: that is the #1 priority to fix — document, pair program, cross-train.
When to Go, Pivot, or Kill
The decision tree that emerges from these 10 assumptions. Each validation experiment produces a clear signal.
| Assumption | If TRUE → | If FALSE → | Run By |
|---|---|---|---|
| #1Revenue leakage is real | ✓ Full steam ahead on market | ✗ Pivot to single acute metric or single vertical | Week 4 |
| #2WTP ≥ $3,500/mo | ✓ Current pricing model viable | ✗ Restructure to usage/seat pricing or high-volume SaaS | Week 7 |
| #3Category comprehension | ✓ Scale marketing spend | ✗ Anchor to existing category; simplify positioning | Week 9 |
| #4Scores correlate with outcomes | ✓ Scoring IP is real competitive moat | ✗ Replace algorithmic scores with direct measurement | Month 3 |
| #5Time-to-value ≤ 18 days | ✓ Self-serve onboarding viable | ✗ Build 3 deep integrations; partner for rest | Month 2 |
| #6Graph DB scales | ✓ Architecture validated for 3 years | ✗ Hybrid relational + cache architecture | Month 2 |
| #7ML models accurate | ✓ ML becomes core differentiator | ✗ Rules-based engine; ML as beta enhancement | Month 3 |
| #8CAC payback ≤ 12 months | ✓ Current GTM motion scales | ✗ Product-led growth or transaction-fee model | Month 4 |
| #9Monthly churn ≤ 3% | ✓ Retention engine compounding | ✗ Halt sales; fix product; invest in CS | Month 6 |
| #10Team capacity adequate | ✓ Roadmap on schedule | ✗ Cut scope, hire, or extend timeline | Week 3 |
The First 90 Days Matter Most
Assumptions #1, #2, and #3 (Market Risk) should be validated before significant engineering investment in assumptions #4–#7. If the market doesn't exist, the product quality is irrelevant. The recommended sequence:
Assumptions Get Tested, Updated, Replaced
This risk register should be revisited monthly. As experiments complete, assumptions graduate from 'hypothesis' to 'validated' or 'refuted.' New assumptions will emerge. The ones that survive become the foundation of the business.