Essays, teardowns, and frameworks from the Checkpoint team: RevOps, GTM, HubSpot, AI, and the unglamorous plumbing that decides whether a quarter lands.
Noah CharakA CRM go-live project plan confirms the tasks are done. It does not confirm the flow works. The cheapest insurance before cutover is a timed, end-to-end dry run of the whole customer journey, with fix-time booked in before launch.
Noah CharakEvery tool now sells you buying signals. A feed of signals is not a strategy. The teams that turn signals into pipeline define the critical event first: the specific moment a buyer actually needs what you sell.
Carolina DecastriMost CRM implementations nail the build. The automation works in demos, the pipeline renders correctly. Then the team starts using it, and the first field that populates with a value nobody can explain costs you a quarter of the goodwill you just built.
Noah CharakFinance is forecasting one number. Sales is forecasting another. The board is squinting at a third. Here is why they almost never line up, and the operator change that fixes it.
Carolina DecastriEvery lifecycle workshop hits the same fork: does ICP fit gate entry to the next stage, or does it score within it? Most teams pick one without realizing the choice reshapes their entire funnel.
Noah CharakThe standard advice is to wait until a lead is qualified before you create the deal. The result is a drawer of fifty hot SQLs the forecast never sees. The pre-pipeline stage fixes the visibility gap without breaking the forecast.
Noah CharakYour ad platform optimizes for whatever conversion event you send it. If the only event you send is a form fill, the algorithm will buy you more form fills. The fix is to send the SQL-stage transition from your CRM back to the platform as an offline conversion. Two workflows, one webhook, six weeks. Here is how it works.
Noah CharakWhen half of your open deals are past their close date, your forecast is fiction. Here is the diagnostic we run on every inherited HubSpot or Salesforce pipeline before any forecast conversation is allowed to start.
Noah CharakA brownfield migration is the audit, not the re-platform. Here is the framework we run on every inherited HubSpot and Salesforce instance before any data moves.
Noah CharakA Salesforce-to-HubSpot migration is a field-level triage exercise, not a 1:1 mapping. Run keep / edit / delete on every Salesforce field, force the picklist consolidation, unwind the formula fields into HubSpot calculations or workflows, and translate lookups into association labels. The signed triage spreadsheet is the cutover artifact; the import is just what runs after.
Noah CharakAI-assisted RevOps automation now ships net-new HubSpot workflows from a natural-language brief, but the failure modes are small, accumulating, and invisible in the chat log. The fix is a verification protocol that reads system state instead of trusting the assistant's report: and a recovery loop that rolls back at the workflow level, not the database level.
Noah CharakMost mid-stage B2B SaaS teams reach for Salesforce CPQ when the actual problem is product-library debt and a quote document that does not reflect what was sold. The CPQ-lite stack: HubSpot line items as the source of truth, PandaDoc as the rendered quote, a disciplined product library, and three approval rules, covers most of the use case at a fraction of the cost. Below a certain rep count and SKU count it outperforms a CPQ rollout; above it, the same stack becomes the bridge to a real CPQ migration.
Noah CharakMid-stage SaaS and financial-services teams default to Sales Hub seats for relationship managers and discover after rollout that the workflow they actually need: ticket SLAs, ownership escalation, conversation routing, lives in Service Hub. The right setup is Service Hub for the RM workflow with a single Sales seat at the team-lead level for upsell deals; pick the seat before the contract, not after.
Noah CharakB2B sales motions with a parallel offer or onboarding workflow do not belong in a single HubSpot deal pipeline. The right architecture is a deal pipeline owned by the AE running in parallel with a ticket pipeline owned by RevOps, with the deal gated from Closed Won until the ticket is closed. Modeling the operational work as deal stages turns the AE into a coordinator and slows pipeline.
Carolina DecastriSelling B2B SaaS into Italy is not a translated version of the DACH or UK motion. The buying conversation is relationship-anchored, the pricing frame arrives by call two, and the HubSpot architecture needs preferred-language and routing-region as first-class properties on contact create, not afterthoughts.
Noah CharakMost teams buy programmatic outbound by category: sequencing, enrichment, signals, and stitch the tools together with email exports. The right architecture is signal-first, a single named trigger fires a scrape, the scrape writes into an enrichment spine, the spine routes matched contacts into the CRM with a buying-signal property, and the CRM owns the workflow that schedules the LinkedIn send. Any tool is replaceable; the signal-to-action contract is not.
Noah CharakThe five-phase Checkpoint methodology: Discovery, Design, Build, Launch, Optimize, is the load-bearing skeleton of any HubSpot implementation. The predictable failure mode is compressing Discovery, a week of conversations gets misread as alignment, Design lands on an unstable schema, and Build re-litigates decisions that should have been signed off in week two.
Carolina DecastriUTM attribution in HubSpot breaks the moment an anonymous session becomes a known contact: cookies and session tokens stop telling the same story as the contact properties. The fix is a session-token property on the contact, populated on first form fill, that maps backward into the Looker session table, turning attribution from a guess into a join.
Noah CharakMost stalled HubSpot implementations are not technical failures, they are ownership failures. The rubric Checkpoint runs assigns one controlling owner across the project, named domain owners across marketing, CRM, external data, service, and reporting, and a short approver list with a defined response window. With it in place, the third-week disagreement over a property is a fifteen-minute decision, not a meeting series.
Carolina DecastriMost B2B SaaS teams treat the sales-to-CS handover as a kickoff meeting, which is why CS opens every onboarding cold. The fix is a sales-room artifact captured before close that names the buyer's outcome, the champion, the blocker the seller solved, and the success criteria. CS reads the artifact in week one and runs the kickoff as a confirmation, not a discovery.
Noah CharakThe HubSpot MCP integration with Claude turns a meaningful slice of CRM administration into a conversation: bulk property renames, property-group reorganizations, and workflow pauses now ship end-to-end against a sandbox without a developer in the loop. The pattern works today because those operations are declarative and idempotent, which is the class of change a language model can scope cleanly. The operating question is not whether to use it, it is which three guardrails you wrap around it before any change touches production.
Carolina DecastriMost HubSpot instances ship with a single lead source property and three teams disagreeing about what it means. The fix is property architecture, not a better dashboard: a first-touch field (write-once on contact create), a last-touch field (updatable by integration), and a converting-touch field (locked at MQL or form submit). Each report picks one property; none of the reports average them.
Harmeet ObhraiWhen an embedded RevOps consultant rotates out, the relationship doesn't transfer, the schema does. The artifact that holds 90 days post-rotation is a written record of every object, property, association label, and workflow trigger, with one sentence of rationale per non-obvious decision. A Loom is not a handover, and the next consultant reads the rationale, not the recording.
Noah CharakPlatform teams at venture funds inherit messy HubSpot instances across a dozen portfolio companies, then try to fix them one at a time. The support model that compounds is a single 90-minute audit template: pipeline definitions, lead source attribution, marketing-contact accounting, renewal hygiene, run identically across the portfolio, with interventions matched to one of four problem categories rather than to the company.
Carolina DecastriThe default booth-CSV import marks every conference attendee as a marketing contact, doubles the HubSpot bill, and triggers the wrong workflows. Import as non-marketing contacts, tag with a campaign property, run a single-step subscription confirmation, and only flip to marketing-contact once opt-in is recorded. Sales follow-up runs on a parallel event-lead property, not on subscription status.
Noah CharakBefore any data moves, the binary question is whether your CRM project is a clean rebuild (greenfield) or an inheritance (brownfield). Three diagnostic questions decide it: history value, deadline pressure, and schema-debt scope. Two of three pointing the same direction is enough to commit; one of three is the conversation worth slowing down for.
Carolina DecastriMost HubSpot renewal forecasts break because reps treat the deal amount as a single typed-in number while finance reconciles at the line-item level. The fix is to anchor renewal deals to the originating subscription's line items, lock GAV to the sum of active line items, and split upsell and downgrade into associated sub-deals. The dashboard then matches the invoicing system without changing what the rep does day-to-day.
Noah CharakMost B2B SaaS pipelines collapse to three stages: qualified, proposal, closed, and lose deals in the gap between problem confirmed and contract signed. The five-stage model that survives contact with real pipelines is Educate, Discover, Value, Setup, Closed Won, where each stage maps to a distinct buyer behaviour and a distinct seller action. The Value stage is the one most teams skip, and skipping it is why deals die in legal.
Harmeet ObhraiWhen a HubSpot portfolio custom object lives one hop away from the deal, through the primary contact, the deal's native association table will not surface it. The three options are a native association table, a report-table embed, and a custom UI extension; the workable answer is a workflow that auto-associates every portfolio on the primary contact onto the deal at creation, even if it surfaces four or five records instead of one.
Noah CharakSales methodology is not a religion; it is a forcing function for deal reviews. Pick by deal complexity and seller experience: BANT for high-velocity inbound, MEDDPICC for committee deals where missing roles slip quarters, and SPICED as the defensible default for everything in between.
Carolina DecastriMost HubSpot instances run a single lead grade that averages company fit and behavioral engagement, and neither team trusts it. Split the score into two contact-level properties: a fit score for cold-list filtering and an engagement score for inbound queue prioritization. Each team weights the one that drives its actual next action.
Noah CharakMost mid-stage SaaS teams overbuy Salesforce: Service licenses, integration platform tiers, and community seats accumulate quietly between renewals. The cost-control move is a 90-day pre-renewal license audit that converts the conversation from price to scope. The audit is the leverage; the negotiation falls out of it.
Carolina DecastriMost HubSpot pipelines have stages whose definitions have drifted across reps, which is why the forecast no longer means anything. The fix is a one-page entry/exit-criteria sheet: definition, entry, exit, owner, that the team agrees on in a workshop before any reporting work happens. Anything that can't be defined in one sentence is two stages, not one; anything where entry and exit criteria match is a flag, not a stage.
Noah CharakBefore a pre-PMF founder hires a RevOps consultant, runs paid ads, or invests in outbound, the cheapest falsifiable next step is ten ICP conversations. The ten meetings rule hardens that advice into a checklist with a single, defensible exit condition: at least seven yes-yes-yes outcomes, or the ICP is not real yet.
Occasional, useful, and never noisy. RevOps and GTM essays from the Checkpoint team.