Pricing & Packaging

Aimee's way of separating a true forward deployed engineer, as Palantir defined the role, from rebranded professional services. First, is it a paid engagement with a billable metric? Second, is the customer effectively renting an engineer for work that may or may not involve a product?

Two versions of what gets called forward deployed work. The Palantir model goes into the customer's environment and builds whatever the customer needs, fully bespoke. Solution architecture helps a customer use the vendor's software largely out of the box, perhaps with scripts or API extensions, without custom development.

Stand up professional services where customers are consistently asking for help. Avoid starting from the premise that services are a lucrative revenue stream to be packaged and sold.

The order for building professional services. Commit to the capability and hire people who can learn to do it. Build a repeatable playbook. Package it with deliverables, timelines, outcomes and a price. Only then run it as a P&L with utilization targets and margins.

Sourcegraph's three packaging models. A fixed-fee package, such as the mandatory implementation attach at $10K base or $25K advanced. A bucket of hours burned down over the contract, used for the resident architect package. An FDE-style engagement priced against delivered outcomes.

A way to validate services pricing. Take the fully loaded cost of an engineer and how many engagements that engineer can run at once. Work out the price point needed to break even, then what it would take to hit a services revenue goal.

Keep at least part of delivery in-house to preserve the feedback loop into the product and the customer relationship. Use partners for coverage, such as regional support, while keeping internal teams on complex or strategic accounts.

Sourcegraph's framing for its delivery competency as a core differentiator: years of playbooks that let it deliver adoption consistently and repeatedly.

Scenario selection, product configuration, commercial terms, and a locked review step — designed around what the front-end user must do rather than how the plumbing works.

Treat guided selling design the way you would treat replicating any top performer: go into conversation intelligence, break down how the best sellers actually sell, and encode that rather than an idealised process.

The binding constraint on out-of-the-box quoting is not what it can produce but what it fails to prevent. Restrict the options a scenario allows rather than correcting errors after the quote goes out.

Before recommending anything, map CPQ, contract lifecycle management and billing/subscription — plus ERP or accounting where relevant — and confirm they can communicate flawlessly.

Automate granular alerts for the things you expect to go wrong — an invoice schedule that has not appeared, a billing date that has slipped — instead of reconciling on a quarterly cadence.

The large stack decisions are determined by company characteristics rather than chosen — headcount decides the CRM, the pricing model decides whether metering exists.

Enterprise Salesforce suite; HubSpot growth stack (the all-in-one default under 200 people); modern AI-native lean stack (deliberately thin, Clay and a warehouse doing the work); dual CRM in transition (Salesforce and HubSpot in parallel, mid-migration or permanently split after an acquisition); and the consumption stack (usage billing wired to real metering).

Tier one — CRM, marketing automation, enrichment — table stakes at any stage. Tier two — sales engagement, CPQ, warehouse, routing — added when the motion demands it. Tier three — next-gen CRM, AI agents, usage metering — the frontier.

Prioritise the unglamorous systems between revenue and the invoice — metering, quote-to-cash — over more visible tooling, because those are the ones nobody funds until they break.

Apply first-principles reasoning to a backlog of requests to find the repeating root cause, collapsing fifty ideas into roughly three themes, then choose against the company's goal for the current or next quarter.

Require evidence of a first small milestone before funding scale, rather than skipping the messy manual trial-and-error phase that reveals what is actually scalable.

State the belief, the reasoning and the measurement in advance, then treat the outcome as a signal rather than the defining verdict, measuring the process separately.

Forward-deployed engineers and deployment strategists own post-sale activation and consumption, so account executives are compensated on bookings rather than on realised usage.

An application layer on a headless Salesforce holding the three things reps touch daily — a Kanban forecast board, a hackathon and AI-day calendar, and transcript-prefilled deal updates — with everything routing back to the CRM as source of truth.

An environment where non-engineers can ship internally built tools, with authorisation controlled at the integration level — read-only on some connections, read-write but never delete on others — rather than inheriting the builder's god-mode access.

A product-led company can layer an enterprise motion on top of an existing consumer-grade acquisition engine relatively easily; a sales-led company attempting the reverse faces cultural, financial and capability barriers that compound.

Holding the point of data capture justifies expanding into every downstream step that capture enables — enrichment, routing, segmentation, payments.

Reserve for enterprise the things only an enterprise genuinely needs — data centre location, security and privacy, scale, administration and SSO — rather than the things that are simply most valuable.

Fence off a defined budget, state up front what success looks like and how it will be measured — usually recognition rather than new business — run it, then hold an honest review including the option to stop.

Hiring a credentialed practitioner from the buyer's own world — a discipline PhD, or someone who held the exact job at a target account — to act as a consultative partner across pre-sale and post-sale.

Invest against demonstrated signal in increments rather than scaling ahead of proof — the opposite of hiring fifty AEs as a growth strategy.

When output metrics — forecast, ACV, deal velocity, quota attainment — fall short, the cause is almost always upstream, so the reporting must reach pipeline generation at rep and campaign level.

A government-wide set of security and compliance controls a technology company must meet before U.S. federal agencies are allowed to put government data into its system. FedRAMP High carries ~425 controls (versus ~95 for SOC 2) and requires a separate enclave, encryption in transit and at rest, FedRAMP-only subprocessors, and assessment by an accredited third-party assessor.

After a 3PAO assessment, a company must find a federal sponsor — a CISO or CIO within an agency (or the DoD/DoW) willing to underwrite its cyber risk and grant an Authorization to Operate. This is the step where a senior official stakes their reputation and job on the vendor.

Every part of FedRAMP can be solved with enough money and time except getting the sponsor — paying for that is bribing the government and is illegal. Sponsors are won through funded mission owners, program budget holders, networking, and combined top-down (political appointees, agency secretaries) and bottom-up pressure.

FedRAMP is annually re-audited and requires monthly continuous monitoring: a check-in with the government across CVEs, misconfigurations, and overall security posture, with strict remediation SLAs (30 days for high-criticality findings, 90 days for medium).

Doing FedRAMP alone is like buying land in the most exclusive zip code and building your own house (permits, architects, supplies, inspection, forever). Knox instead runs the 'luxury condo building' on Main Street: customers move into a single-tenant floor, bring their own furniture (CI/CD, APM, hyperscaler), and inherit ~80% of the 425 controls plus Knox's agency sponsors.

In an authorization-gated market, the fastest (if not cheapest) route to FedRAMP can be to acquire a company that already holds it, then build on that boundary — rather than pursue a multi-year organic authorization.

Because so few vendors clear FedRAMP, entire federal software categories run on one or two authorized options — ITSM has only ServiceNow and Salesforce; the OMB HRIS RFP came down to Oracle and Workday. Monopoly/duopoly conditions push prices up and product quality years behind commercial equivalents.

The federal government rarely buys anything for under $1M — the contracting overhead makes smaller deals uneconomic — and it purchases for 10,000–100,000 users at a time, favoring vendors already proven at commercial scale.

Before spending a dollar on federal, ask whether you've proven yourself at enterprise scale commercially — the one gate that can't be spun up. Commercial-first companies typically qualify around ~200 people / ~$50M revenue with a few enterprise logos; defense-tech, government-first companies can pursue it as early as 50–75 people.

FedRAMP acts as a pre-diligenced, CYA-grade trust signal in commercial regulated markets. After a certification announcement the first inbound is often a financial-services or healthcare buyer, not a government agency — so FedRAMP behaves like a 'super SOC 2' that differentiates and closes commercial deals.

Design monetization in sequence: packaging first, then pricing structure, then the pricing metric (what you charge on, and whether forward- or backward-looking), and the price point dead last.

Bundle features around the outcomes a customer is trying to achieve, not around a product-team ranking of which features get used most.

The amount of pricing/packaging complexity a business can carry is capped by its ACV and the type of customer it sells to — high ACV can absorb complex enterprise pricing; low-ACV startup sales demand simplicity.

Classic SaaS had near-zero marginal cost and 80–95% margins; AI reintroduces a real, usage-driven cost to serve that most SaaS-native companies can't even measure, so pricing that doesn't follow cost loses money.

Outcome-based is what you price on; usage-based is how you meter it — and they can combine. Outcome pricing fits vertical products with a uniform outcome and is a trap for horizontal products where the same usage means different things to different users.

You're a candidate for outcome-based pricing only if: (1) you can clearly define one outcome across your customer base; (2) customers agree to and accept that exact definition; and (3) the value of that outcome is similar across all customers.

Make usage-based pricing palatable by combining a recurring base fee with usage on top, wrapped in real-time visibility, per-team spend controls, caps/notifications, and token-level billing traceability.

De-risk migrating your most important customer by working backward: test new pricing on the least-important segment or market, then new business, then run validation interviews with your 2nd–5th customers, plus internal validation with the sales team, before migrating the anchor account.

Treat pricing and packaging as a living product surface, revisited every product-release cycle or sales cycle (whichever is longer) — not set once and left for five to seven years.

A transformation framing in which agentic AI augments the revenue team rather than replacing it — the 'plus' signals that humans stay in the motion, owning relationships and accountability, while agents handle preparation and scale.

The higher the deal value and the more up-market the customer, the less AI belongs in the customer-facing interaction — and the further down the tail (SMB), the more the agent can own the motion with a human reviewing the output.

A two-variable decision model for how much AI to put into any motion: account value (ACV) and which product surface the customer is touching (and how mature that surface is).

Split customers into enterprise (large advertisers), mid-market (D2C brands, performance agencies), and SMB, and assign a different agentic role to each based on that segment's customer-service needs and risk tolerance.

The newer and less proven a product, the more human-in-the-loop the motion should be — because the fastest way to learn from customers experiencing something new is to talk to them, not to automate the interaction.

AI enablement swings between a centralized owning group and fully decentralized team-by-team ownership; the healthy resting point is in the middle — cost-and-tool guardrails set centrally, process redesign owned by the teams.

Whether you can decentralize AI at all is gated by the strength of your underlying data — a clean CRM and a healthy data stack are the precondition for letting functions own their own AI.

Judge AI initiatives by revenue and efficiency outcomes — speed to market, meeting volume, pipeline-stage conversion, revenue per head, ARPU — rather than by AI spend.

The operator's path to senior leadership: deliberately pursue new lines of business, international expansion, and reorgs so you see and understand the entire business, not just one function.

The point of 'carrying a bag' isn't the title — it's going through a period where you genuinely feel the pressure of contributing to the top line, a career experience you have to go collect.

Run your weeks against roughly four equal priorities, each assigned a color, and audit your calendar so it's about a quarter of each — a mechanism to keep strategic time allocation honest.

Spend the opening months of a new role absorbing information from two sources — the people on the ground doing the selling and implementing, and the available data — then marry those perspectives into a working theory of what's actually happening before setting priorities.

When you miss a quarter, an intellectually honest post-mortem usually ties the miss to a planning or strategy assumption you weren't honest about — not to near-term deal execution.

Charge only when the AI agent completes the entire job correctly (e.g., reads and infers every field on a document 100% right), so price tracks roughly one-to-one with the value delivered.

Report usage-based revenue to the board by stacking two clearly labeled layers: contracted ARR ('take it to the bank') plus a conservative fraction of the forecasted amount booked as 'estimated ACV' (EACV).

Start with a small, high-confidence use case you know you can nail, prove value fast, and use that win to earn the right to the next project — becoming the customer's primary consideration for what's next.

Qualify and sell the deal first, then run the POC only to confirm the solution works and the teams click — never as a desperate Hail Mary to generate intent that isn't there.

Before offering a POC, square away the MSA, legal and security, and budget, and get both IT and operations (the business side) at the table and excited. The POC then only validates the solution and the working relationship.

Have the implementation / agent-PM team scope the work during the sales cycle, so the buyer meets who they'll work with, gains confidence, and sellers can't over-promise.

Build the revenue plan so every channel has its own win rate and ASP and new-logo ('NUCO') plugs the gap to the number — robust enough that a missed quarter can be traced to a specific assumption that broke, with leading indicators warning you a quarter or two out.

Leadership sets a non-negotiable number; what's up for debate is only the resources required to deliver it. The exercise is iterative and cross-functional, and pairs with a proactive 'what would it take to 10x my org' model run before the CEO asks.

The fear that native-AI companies will displace SaaS incumbents that bolt AI onto legacy architecture — and the buyer's inability to tell a truly native-AI product from a SaaS wrapper.

Meet the customer in the middle by proving value in their own environment through a proof of concept, sharing risk, then expanding — making the POC the default go-to-market move rather than a concession.

AI deals require selling horizontally across the functional buyer (HR generalists, HR ops, HR leadership), IT, an AI committee, and security — any of whom can veto or delay the deal.

The economic buyer — the person with discretionary authority to say yes and move budget — has gone horizontal in AI deals; IT is often the new economic buyer even on an HR purchase.

Embedding technical forward-deployed engineers into the implementation team to manage LLM change, build guardrails against hallucination, and hand-hold customers through early adoption — modeled as an accounts-per-FDE/CS gearing ratio in headcount planning.

An implementation philosophy (borrowed from Indiana football coach Curt Cignetti's 'stacking days, stacking wins') of engineering a continuous drumbeat of provable metrics and success stories the champion can tell internally.

Extending the go-to-market engineer role beyond top-of-funnel prospecting into mid-funnel deal execution — automated SOWs from call transcripts, company-specific deal coaching, and CRM-plugged GTM diagnostics.

The enterprise qualification methodology Scott helped develop; in AI's chaos the two most decisive elements he stresses are Champion ('no champion, no deal') and Decision Criteria — the capability shopping list a buyer uses to evaluate vendors.

Rather than extracting the buyer's decision criteria, supply it: an editable, weighted, unbranded capability scorecard that lets the customer objectively compare vendors for the problem they're solving.

Auto-fill MEDDPICC in the CRM from Gong call transcripts while also keeping rep-entered MEDDPICC, then compare and contrast the two to triangulate what's actually happening in accounts.

When a company hires a go-to-market leader, RevOps and enablement are the first two hires; the ecosystem is built before AEs so reps ramp fast into a well-oiled machine.

Each new media/computing paradigm — radio, TV, web, social — builds a wholly new advertising infrastructure around it, and advertising never disappears. AI is the next paradigm and will get its own reinvented ad stack.

Instead of one agency-created, brand-approved ad shown to everyone, every person is served an ad generated specifically for them — the long-sought 'holy grail' that AI can finally deliver at scale.

A ladder from a fixed banner to fully generated advertising: start with dynamic text generated to match the conversation, move up to generated assets, then to dynamic UI — the ad format and interface generated per user and per brand.

The publisher's surface area is treated as sacred, and the ad platform optimizes for the end user first, the publisher second, and the advertiser third — because end-user value drives engagement, which in turn makes advertisers happy.

Click-through rate is a misleading success measure because users can click without converting; the right target is the advertiser's actual objective (conversion/value), which should hold steady rather than decay if the ads are genuinely relevant.

The early assumption that 'user asks about shoes → show a shoe ad → they buy' is wrong; buyers research with an LLM but still purchase elsewhere to price-shop and earn rewards. Closing that behavior gap takes time, not just better technology.

The more expensive or complex a purchase, the more a buyer researches first — so AI ads convert best for high-consideration, financial categories (taxes, student loans, credit cards), amplified by seasonality like tax day.

A north star of returning advertising to craft — the era when a Got Milk campaign or a full-page New York Times ad was a genuine piece of art — using generative tools to produce brand-true experiences people admire rather than block.

In a world of unlimited information and opportunity, the scarce edge is clarity — the ability to focus on the few highest-priority problems and not over-index on the feeling of stress. Individual clarity and organizational clarity move together.

A concept coined by a16z's Martin Casado: unlike product-market fit (fitting a product to existing demand), market annealing means shaping the market itself — educating buyers, defining the demand, and shaping the product in parallel.

The platform is a world-class kitchen that can prepare any 'meal' (extract value from any image or video data). Rather than acting as waiters serving every hungry customer a different dish, you build one focused 'hamburger stand' — a single killer app sold 100% outbound — to prove a restaurant can be built on top of the kitchen.

A powerful platform doesn't create instant value on its own; it needs a killer application that lets customers get value immediately — the way Databricks needed notebooks before people could realize its value.

Instead of marking up hyperscaler infrastructure and competing on its margin, pass that cost through at parity and charge on the usage or value delivered on top of it.

A weekly leadership operating cadence with no fixed agenda or set times: the team gathers to solve the hardest problems and works until the set of things is finished, ordering pizza along the way.

A progression of accountability: an IC empowers themselves; a head of sales empowers through people and owns a team; a CRO takes accountability for the whole company — vision, strategy, fundraising, and product influence.

Just as engineering accrues technical debt, an organization accrues partner debt, customer debt, and employee debt by taking on too many things at once and failing to fulfill the promises made.

A three-stage progression: prove genuine product-market fit once (a real problem, consumer-first), then achieve go-to-market fit by making the motion repeatable, then scale it into a 'dynasty' rather than a single lucky win.

Build around a real consumer problem and behavior, deliberately not leading with the novel technology (crypto/blockchain) or a get-rich-quick token.

A trading mechanic where a player offers unwanted items for a desired one; the system real-time-buys the target from one seller and real-time-sells the offered items to multiple buyers worldwide, netting a cashless swap.

Layering successive, lower-friction on-ramps into the economy — crypto, then credit card, then cashless quick trade, then an AI NPC negotiator — so each new rung pulls a broader persona into trading.

A staged handover of asset control from the studio to the player: start closed (assets exist but access is tightly controlled), move to shared responsibility ('you have a key, I have a key'), then to full player control (take it and go, even revoke the game's access).

When AI is integrated properly into how code is deployed and how engineers work, each person effectively gains a code reviewer, a junior programmer, and a security analyst — roughly tripling their output.

Using different AI models to challenge each other's work — e.g., writing a piece of code with Claude and having ChatGPT analyze it in the role of a security compliance officer.

Mythical's build sequence: (1) economic tech — changing the economies inside games; (2) social interaction — new ways for players to compete and play together (Pulse Arena tournaments); (3) open platform — letting outside studios build on Mythical's stack.

A reframe of investor rejection: a no usually reflects the VC's own vision or their LP-mandated 'deal box,' not a flaw in the idea — 'it's not a bad idea, it's just not the idea they have.'

Because every public SaaS company answers to the same SEC rules, quote-to-cash should be a standardized, out-of-the-box process — not a uniquely engineered snowflake per company. A 'unique' process is a problem to fix, not a competitive advantage.

Instead of a separate CPQ, billing system, and revenue-recognition system integrated between CRM and ERP, run a single platform that handles CPQ, AR/billing, and ASC 606 rev rec — sitting between the CRM and the GL with no reconciliation and one product catalog.

An AI agent that lets any seller generate a compliant quote by typing a plain-English request into Slack (or mobile, email, or the CRM). The agent parses the request, asks for any missing policy-required inputs, applies product rules, and returns a quote PDF.

A business-focused Q&A layer that asks a seller simple questions (where is the customer located, what segment) and converts the answers into the right products, compliance, and discounting — instead of making the rep understand how the CPQ is configured.

Usage/consumption billing comes in distinct models: pure pay-as-you-go (no commitment, invoice on actual use), pre-committed plus overage (commit to a volume like 200,000 API calls/month, pay extra above it), and credit pools (buy a $100k pool and draw down across products, AWS/GCP-style).

A policy-driven method for turning variable consumption into a defensible ARR: for pay-as-you-go, take average consumption over a trailing 3-6 months and recognize a set percentage (e.g., 80%); for committed-plus-overage, the commitment is fixed ARR and overage recognition depends on how straight-line it is and what the auditor will accept (from ~95% down to ~20%).

When a customer adds licenses and renews early, you cancel the current term (crediting the unused period, like dropping a car lease) and restructure into a new term. Done right it's one opportunity, one order form, with credits and proration auto-calculated and reporting that shows it as upsell — not churn plus a new deal.

The design principle that the primary consumer of a CPQ should be the seller, not deal desk or RevOps. Reps should be able to run even complex deals (multi-year ramps, partner margins, special payment clauses) and the entire post-signature lifecycle themselves.

The seller creates an 'order' (draft during the sale cycle, confirmed once closed) and that same object generates the invoice and feeds finance. There's no separate quote-to-invoice re-keying, so numbers can't diverge between what sales sold and what finance bills.

The consistent core set of capabilities the market keeps asking to have in one place: sales engagement (cadences and sequences), conversation intelligence, data and enrichment, and predictable forecasting.

Bernardo's three equally-likely scenarios for these platforms: (1) consolidation and rebranding succeed into specialized all-in-one platforms; (2) vendors can't escape their legacy branding and stay boxed into what they were known for; (3) a platform becomes the ecosystem — builds a CRM and takes on Salesforce and HubSpot directly.

A re-evaluation discipline: dust off a structured scorecard and grade every vendor across its full, current feature set — not just its original category — and refresh it far more often than quarterly or annually.

The core buyer decision: assemble best-in-class point solutions for each category, or align the whole go-to-market operation on a single consolidated platform. As tools commoditize, the pull is toward picking one 'pony,' driven by cost, bundling economics, and current negotiating leverage.

The mistake of misperceiving whether your highest value is functional or emotional and then stacking it incorrectly across messaging, content, and sales training — leaving real value unclaimed in the buyer's mind.

A shift from measuring value in price (the only conventional unit) to a broader unit Miller calls 'worthiness,' since price answers none of the buyer's real questions about effectiveness or how the offer will make them feel.

A metaphor treating value exchange between vendor and customer as gravitational attraction — heavier objects (more mass) pull lighter ones toward them — used to deconstruct why some offers pull customers and others don't.

The first source of gravity: how valuable an offer is innately, before the market perceives it at all — a breakthrough technology is extremely valuable sitting in the lab before anyone knows it exists.

The second source of gravity: getting an offer close enough to the customer that they know about it and can be pulled toward it — high-touch for enterprise, low-touch volume for SMB.

The third and most-overlooked source of gravity: the opposing pull of competing options that holds the customer in place and prevents a value exchange — the gravity of the next best alternative.

An occasional, outsized, nonlinear payoff that occurs when timing or the sudden significance of a job-to-be-done causes value to compound rather than add.

The single element of value that is most strongly connected to your brand — unique and special to you. Tom also calls it your superpower, and cites data that it can represent as much as 70% of perceived value.

A triangle whose three sides are the ways humans subconsciously perceive value: functional, emotional, and economic. In any value exchange the brain stacks one element as primary — up to ~70% of the perception.

Products are tools that help customers get jobs done. Whether a customer acquires a tool depends on the job they're trying to do and how functional or emotional that job is.

A maxim Tom credits to Kellogg's MBA program: the primary job of a marketer is to lower the cost of customer thinking. Adding feature on feature or benefit on benefit dilutes rather than compounds perceived value.

Tom's metaphor for RevOps: the 'plumbing' is instrumenting, maintaining, running, extracting, and visualizing the data; the 'poetry' is interpreting that data into a performance narrative. The combination is the value.

An informed estimate of what a usage-based deal will be worth over its first 12 months (or a chosen period), assigned even when zero dollars are contractually committed, so the deal can be reported, forecast, and managed.

A closed-loop discipline of tracking each deal's real consumption against its assigned expected value — daily, monthly, or otherwise, but at least through the first year — to see where estimates over- or under-called.

A method for estimating expected value that starts from a data baseline — usage trends of similar companies and of a customer's first three, six, and nine months — then layers in rep discovery, safeguards, and discounts to land a defensible number.

The anti-pattern of forcing usage into a committed contract by discounting the per-unit price — e.g., committing 25% of expected volume for a 10% price cut — to buy reporting predictability.