The usual version of this story: a screenshot of the tool stack arrives. Dozens of logos, an annual software bill that is a visible slice of ARR, and one question, which tools to cut. The honest answer is most of them.
Having run RevOps audits across more than 30 engagements over the last seven years, from pre-seed to mid-stage Series B, the pattern is the same every time. Tools get bought to solve a moment of pain. Nobody removes them when the pain passes. The stack grows, the data fractures, and the CFO eventually asks what all of it is doing for pipeline.
This post is the opposite: the smallest RevOps stack that takes a B2B team from cold outbound to closed won without leaking data, attribution or budget, what to skip, and how to evaluate the platforms underneath it on the facts their own documentation publishes rather than on a feature grid.
Pick the platform on three documented facts, not on the feature grid.
Which subscription tier unlocks deal revenue attribution, what has to be configured before the forecast will read your pipeline, and whether the native agent layer is included in the seat or metered on usage. All three are written down in the vendors' own documentation, and all three move the price and the timeline far more than any feature comparison does.
The default stack most teams end up with
Here is the typical Series A RevOps stack:
- HubSpot or Salesforce as the CRM
- A separate marketing automation platform that mostly overlaps the CRM
- Two sales engagement tools because somebody piloted both
- Three enrichment vendors with overlapping data
- A conversational intelligence tool nobody listens back on
- An ABM platform nobody logs into
- A data warehouse one engineer set up and then left
- Half a dozen point automation tools
The problem is not that any of those tools are bad. It is that nobody designed the stack. Each one got bought to plug a hole, the integrations are duct tape, and the reporting is impossible because every tool owns part of the truth.
Average waste in B2B RevOps tech budgets we audit. Two thirds of the spend goes to tools that overlap, duplicate, or sit unused.
The five jobs a RevOps stack actually has to do
Strip the marketing back and the stack has to do five things. Every tool you own should belong to one of them. If it does not, cut it.
- Record the truth. One system holds accounts, contacts, deals and revenue history. Nothing else gets to be the source.
- Find and qualify accounts. Enrichment, signals and intent that turn a list of companies into a list worth working.
- Reach the buyer. Outbound and inbound delivery: email, calls, LinkedIn, ads, forms.
- Run the work. Automation that moves data, sends messages and routes leads without humans copy-pasting.
- Show the numbers. Reporting that tells the team what to do tomorrow.
That is the whole job. Five layers, and most stacks have twelve tools.
Layer 1: the CRM
This is the heart of the stack. Get it right and the rest is easy. Get it wrong and no amount of tooling saves you.
For B2B companies up to a few hundred employees, HubSpot is almost always the right call: faster to set up, and the workflow engine is genuinely good now. Read the tier before the feature list, because the tier is the decision. HubSpot's Sales Hub pricing page lists Professional from $90 per seat per month committed annually and Enterprise from $150, and it also lists a required one-time onboarding fee of $1,500 on Professional and $3,500 on Enterprise, which most comparison posts leave out of the first-year number.
Salesforce starts to make sense above that, with complex sales motions, multi-product portfolios or heavy customisation needs. Not before. Most teams that pick it at Series A regret it, and the implementation needs a full-time admin.
Attio is the wild card: what HubSpot would be if it were rebuilt today, with a flexible data model and a fast UI, but thinner reporting and a younger workflow engine.
What to skip in this layer:
- Any CRM without a real API
- "Industry-specific" CRMs that lock you into one motion
- Anything that needs more than a week to set up for a small team
We break down the HubSpot and Salesforce choice in our CRM and RevOps work, and there is a migration playbook if you are already on the wrong one.
Layer 2: find and qualify
This is the layer most teams overspend on. They buy ZoomInfo, Apollo, Lusha, Cognism and an intent vendor on top, run all five in parallel, and the data still has gaps.
The way to do this in 2026 is waterfall enrichment: one tool, multiple data providers, fallback logic, and you pay only for what you pull. Clay is the leader. It finds and verifies contact data across providers, adds firmographic and technographic fields on demand, and watches signals such as hiring, funding and tech changes, writing them back to the CRM.
What to skip:
- Any enrichment tool that charges per seat instead of per credit
- Intent data products that send thousands of "interested" accounts a week with no signal logic
- ABM platforms at five figures a year before the motion is proven
For most teams, Clay plus a small LinkedIn Sales Navigator seat feeds the entire pipeline. The signals side is in our buyer intent guide, and the credit arithmetic is in our guide to what Clay costs across a revenue engine.
Layer 3: reach the buyer
This is where companies usually have too many tools: sales engagement platforms, email warming tools, LinkedIn automators, voicemail drop, dialers. You need one sales engagement tool, one cold outbound infrastructure, and ads if you can afford them.
For sales engagement, the choice is HubSpot Sequences if you are on HubSpot, or Outreach or Salesloft once you have a real outbound team. Below that, Sequences is good enough and the seat difference is wasted money.
For cold outbound, Smartlead or Instantly plus a separate sending domain. Cold email and the main CRM-connected domain should never share DNS. If they do, one bad campaign burns the main inbox and breaks inbound replies for a month.
What to skip:
- Two sales engagement tools in parallel
- LinkedIn automation that risks your reps' accounts
- Dialers if your reps do not actually call
The cold email deliverability piece covers how to set up the second domain.
Layer 4: run the work
This layer quietly decides whether the whole stack works. Without it, every other tool is an island and somebody is exporting CSVs at midnight.
HubSpot Workflows handles most of what lives entirely inside the CRM. The rest lives between systems: pulling enriched data from Clay back into the CRM, routing leads by territory, Slack alerts when a target account hits a signal, syncing to the warehouse. That is where n8n comes in. It is open source, self-hosted, GDPR-safe and cheap to run on a small VPS, and it connects almost anything. Zapier does similar work, but per-task pricing gets expensive at scale and the data leaves your perimeter.
The pattern we build:
- HubSpot Workflows for everything inside HubSpot
- n8n for cross-system orchestration, enrichment writeback and notifications
- One named owner, always. Automation without an owner becomes shelfware.
The n8n RevOps guide and the HubSpot workflows playbook go deeper, and our AI automation work is mostly built on top of this layer.
Layer 5: show the numbers
The mistake here is jumping to a BI tool too early. Most Series A teams do not need Looker or Tableau. They need a handful of HubSpot dashboards the leadership team actually opens: pipeline by stage, pipeline by source, deal velocity, SDR activity, win and loss by segment. That is the entire reporting layer.
A warehouse starts to earn its place with multiple products, partner channels and a real finance team. The minimal version is Fivetran plus BigQuery plus Metabase, and it buys cohort analysis and attribution the CRM cannot model.
What to skip until you genuinely need it:
- BI tools on a small team
- Multi-touch attribution platforms before you have enough closed deals to model
- Standalone forecasting tools that resell your own CRM data back to you
How do you evaluate a RevOps platform without reading another listicle?
Start from the vendor's own documentation rather than from a ranked list. Listicles compare feature names, and feature names are the part of the decision that does not hurt. What hurts is the tier a capability is locked behind, the configuration that has to exist before it works, and the meter that runs while it works. Three questions settle most shortlists, and the vendors publish all three answers:
- At which subscription tier does deal revenue attribution become available?
- What has to be configured before the forecast reads the pipeline?
- Is the agent layer included in the seat price or billed on consumption?
Where a vendor's documentation could not be read, that is said below rather than guessed. The three sections that follow answer each question in turn.
Which platforms do revenue attribution, and at what tier?
In HubSpot, the attribution report is not one product. HubSpot's knowledge base article on creating attribution reports lists it as available with Marketing Hub Professional and Enterprise and Content Hub Professional and Enterprise, and then narrows it further: contact create attribution is the broadly available one, while deal create attribution and deal revenue attribution are Marketing Hub Enterprise accounts only.
That single line reorders most shortlists. If you are buying to see which sources produced closed revenue rather than which produced contacts, you are buying Marketing Hub Enterprise, not Sales Hub, and that is usually discovered after the Sales Hub contract is signed.
The same article carries the limits, which decide whether the report describes your funnel honestly:
- The supported models are first touch, last touch, linear, time decay and empirical. Empirical replaces the older U-shaped, W-shaped, J-shaped and inverse J-shaped models and weights interaction types by how often they occur across conversion paths, giving less frequent interactions more weight.
- Reports include up to 20 million interactions each after sampling, with at least one year represented when available, and accept up to 100 event input types.
- Sampling is not uniform. High-frequency types such as email interactions and page views are sampled; form submissions, calls and meetings are not. If your motion runs on meetings, that is in your favour.
The reports were also renamed, which matters when reading an older comparison: deal attribution is now deal create attribution, contact attribution is now contact create attribution, and revenue attribution is now deal revenue attribution.
The Salesforce campaign influence and multi-touch attribution articles in the help centre could not be read, so nothing about their tiers or models is stated here. Ask for the enablement path on a live org before assuming parity.
What has to be true before a platform can forecast for you?
Forecasting is a configuration project rather than a toggle, and the length of that project is the real difference between platforms.
HubSpot's forecast tool setup guide describes the whole thing as one settings page. The tool is available with Sales Hub Professional and Enterprise and Service Hub Professional and Enterprise. Setting it up needs Super Admin permissions, and creating forecast goals needs an assigned Sales or Service seat. You choose how deal amounts are counted: weighted amount, which is the deal amount multiplied by the deal probability, total amount, or a legacy forecast amount option that only appears in accounts created before 20 March 2024. You then map each deal stage into one of five forecast categories, not forecasted, pipeline, best case, commit and closed won, and an "automate forecast categories" switch generates a workflow named after the pipeline to keep the mapping current.
Salesforce Revenue Intelligence is a dependency chain instead. The Trailhead unit on enabling and configuring Revenue Intelligence states that CRM Analytics has to be set up as a prerequisite and an accurate role hierarchy has to exist, that permissions and Pipeline Forecasts must be in place before Revenue Insights if you want forecasting data to appear there, that the Revenue Intelligence Admin and Revenue Intelligence User permission sets carry access to Pipeline Inspection, Revenue Insights and CRM Analytics, and that every user who participates in forecasting needs the "allow forecasting" checkbox set on their user record.
Pipedrive sits between the two. Its sales forecasting page describes a forecast that recalculates automatically on every deal update, a forecast versus actual comparison across reps, teams and time periods, and AI notifications when deals stall or slip. Its own list of what to look for leads with real-time pipeline visibility and revenue goal tracking rather than with model sophistication, which is the right order.
The question to put to any vendor is narrow: which object does the forecast read, and what has to exist before it reads it. In one platform the answer is a settings page. In another it is a chain of prerequisites and a role hierarchy. Neither is wrong, but one is a week and the other is a quarter, and that gap is usually larger than the licence difference.
Where do the native AI agents sit, and what do they cost to run?
Both of the large CRMs now ship an agent layer on top of the records. The question that decides the budget is whether that layer is flat or metered.
HubSpot's agent builder page splits it cleanly. Agent builder, formerly Breeze studio, is included in the Starter, Professional and Enterprise editions. Workflows are flat: simple automation in Starter, the full workflow builder with event-based triggers and advanced automation in Professional and Enterprise, at no extra cost, and they do not consume HubSpot Credits. Custom agents are metered. They consume credits on every run, the work is measured in units of agent action, and HubSpot states plainly that the actions a custom agent takes vary from run to run. Credits can be bought in packs for predictable spend or used pay as you go.
That distinction is an architecture decision, not a billing detail. Deterministic work, where the same input must always produce the same output, belongs in a workflow, where it is free and auditable. Agent work belongs where the variability is the point. Putting routing rules inside an agent is the most common way to turn a fixed cost into a variable one for no gain.
Salesforce frames Agentforce as platform infrastructure instead. Its Agentforce platform page describes agents built natively on the platform, connected to business data and metadata through Data 360, using retrieval-augmented generation over structured and unstructured data, with hybrid reasoning that combines deterministic logic and adaptive AI, MuleSoft as the automation layer, and pre-built templates, topics and actions across the Customer 360 applications. The implication is the one running through this whole post: the agent layer assumes the data layer is already in order. If the CRM is not clean, the agents inherit the mess and answer confidently from it.
Confirm two things on any agent layer before signing. Can the agent be tested before it runs live, and can its spend be capped. HubSpot's page answers both. Ask for the equivalent in writing from anyone else.
The actual stack: what I build for a Series A team
The stack below is what goes in for B2B SaaS teams between $1M and $10M ARR, at roughly $2,800 a month for fifteen users, all in.
Three deliberate omissions. There is nothing for conversational intelligence, because Gong is hard to justify until you have a bench of AEs actually being coached on the recordings. There is no separate marketing automation platform, and no warehouse: reporting lives in the CRM until roughly the $5M ARR mark. Past that you add a warehouse, multi-product reporting and regional CPQ rules, but the five layers do not change.
Where teams overspend
Three patterns come up in almost every audit I run. If your stack has any of them, there is spend to cut without losing capability.
The other pattern: teams paying for AI features their CRM licence already ships. Check what the tier covers before adding another logo, which is what the three questions above are for.
The most expensive version is the customer data platform, missing from the table only because most teams this size have not been pitched one yet. Three questions settle it before the demo: is there behavioural data at real volume outside the CRM, are there activation destinations the CRM cannot reach, and will someone own the schema and the identity rules. Three noes means you do not need one. The full test for when a B2B team actually needs a CDP is written up separately.
How to actually rebuild your stack
If your stack is the messy version, here is the order to run the cleanup in.
First, list every tool you pay for and which of the five jobs it does. A tool that fits none of them is a candidate to cut, and where two tools do the same job, one goes.
Second, check usage in the admin reports. Low monthly seat logins is the clearest cut signal there is.
Third, map the integrations. A tool that is not integrated with the CRM is a data island, and you are losing reporting on everything it touches.
Fourth, rebuild from the CRM out, then re-run the three evaluation questions above against whatever is left. This is what our GTM and RevOps work covers.
Stack feels bloated? Let us audit it.
Book a free 30-minute RevOps stack review. We will tell you which tools to cut, which to keep, and where the data is currently leaking.
Book a stack audit →FAQ
How much should a B2B company spend on RevOps tools?
For Series A teams between $1M and $10M ARR, total tooling spend sits around $2,500 to $5,000 per month for a team of ten to twenty. Past that, you are usually paying for overlap.
Do I need Salesforce if I have a small B2B team?
No. On a small team HubSpot is almost always the better choice. Salesforce makes sense with complex sales motions, multi-product portfolios or compliance requirements that demand deep customisation.
Which RevOps platform should I pick if revenue attribution is the main reason?
Check the tier before the feature list. HubSpot's attribution reporting documentation puts deal create and deal revenue attribution in Marketing Hub Enterprise accounts only, so a Sales Hub licence on its own will not produce them. Price the tier that actually carries the capability, not the one that carries the demo.
Do I need a separate RevOps platform on top of my CRM?
Usually not below the warehouse threshold. The five jobs above are all served by the CRM plus one enrichment tool plus one orchestration layer. A separate RevOps platform earns its place when you have revenue data the CRM genuinely cannot hold, not when the CRM is simply configured badly.
Can AI agents handle pipeline forecasting for me?
They can summarise and flag, and they are metered when they do. HubSpot's agent builder page keeps workflows free of credits while custom agents consume them on every run, so the forecast itself should stay in the deterministic forecast tool and the agent should be pointed at the judgement work around it, such as surfacing deals that have not moved.
When should I add a data warehouse?
Around the $5M ARR mark. Before that, CRM dashboards handle what leadership needs. Add Fivetran, BigQuery and Metabase once you need multi-product reporting or attribution the CRM cannot model.