Most mid-market teams already own Service Hub. Far fewer have the architecture underneath it — a ticket data model that makes SLA reporting trustworthy, routing that reads customer tier and health, a knowledge base wired into the ticket flow, and service signals that reach sales before a renewal goes wrong. We implement Service Hub, rebuild the ones that aren't working, and build the custom layer HubSpot doesn't ship.
500+ client deployments across HubSpot and Salesforce — including Service Hub builds, rescues and helpdesk migrations for SaaS, legal, healthcare and fintech teams in the US, Canada, UK, UAE, Australia and New Zealand.
What We Work On In Service Hub
These are the six findings that come up again and again in HubSpot portal audits. None of them is a missing feature. Every one of them is a design decision nobody made.
A shared inbox, one pipeline, and three workflows nobody remembers creating. Tickets sit in "New" for two days because the routing logic doesn't account for priority, customer tier or issue type. By the time someone picks it up, the customer has escalated — or left.
Targets were set in month one. Eighteen months later nobody opens the SLA dashboard, because stages get moved by hand, timestamps are wrong and breach escalations never fire. Leadership asks about response-time trends and nobody has a credible answer.
Thirty articles were written in the first quarter and then it stopped. The structure doesn't match how customers search, nothing has been updated in a year, and the portal never surfaces the right article at the point in the ticket flow where it would have deflected the ticket.
Sales closes the deal and hands over. Service has no view of what was promised, what the account is worth, or when it renews — so it solves problems without context, and sales walks into the renewal without knowing what went wrong. Churn surprises everyone.
The customer agent went live because it came with the licence. It is pulling answers from stale knowledge base articles, escalating inconsistently, and frustrating customers who expected a faster resolution. AI on a weak knowledge and ticket model doesn't reduce tickets — it creates new ones.
NPS, CSAT and CES are all collecting and the dashboards update. But the detractor workflow never triggers a human, promoter scores never reach the advocacy program, and CS still works from instinct about which accounts are at risk. Survey data with no response architecture is noise with a score attached.
HubSpot Service Hub is HubSpot's customer service and success product: a help desk workspace with ticketing and omnichannel conversations, SLA management, a knowledge base and customer portal, NPS, CSAT and CES surveys, customer health and success tooling, and Breeze AI — including the Breeze Customer Agent, which resolves conversations autonomously. Its defining property is not any single tool. It is that every ticket sits on the same CRM record as the deal, the marketing history and the lifecycle stage, so a support interaction and a renewal conversation are looking at one customer instead of two. HubSpot's Service Hub documentation
That property is also the trap. Because the data is shared by default, a badly modelled ticket object quietly degrades everything downstream of it — SLA reporting, health scores, churn signals, the accuracy of the AI agent. Service Hub does not fail loudly. It reports confidently on numbers that are wrong.
So it is worth being precise about who does what. HubSpot ships the platform. Twopir designs the architecture — the ticket data model, the routing and escalation logic, the knowledge taxonomy, the health scoring inputs and the cross-hub workflows — and builds the custom objects, integrations and UI extensions the standard product does not cover. What the client gets is a service system whose reporting their CFO will accept and whose churn warnings their CS team acts on, rather than a portal with every feature enabled.
These are three different pieces of work with three different price shapes, and most teams only need one of them. The fastest way to know which is the free audit — it exists precisely so nobody buys a rebuild they didn't need.
For teams moving off a shared inbox, a spreadsheet or a standalone helpdesk. We architect the whole system before we configure any of it.
Where this ends: everything here is standard Service Hub configuration. No code, no private app, nothing that needs a developer to maintain after handover.
For a live portal that isn't producing the retention outcomes it should. We keep what works and rebuild only the components causing the gap.
Where this ends: we do not start over. If the audit says your model is sound and the problem is three workflows, that is the engagement — and it is a small one.
For teams whose service process has outgrown what configuration can express. This is engineering work, scoped and priced as engineering work.
Where configuration ends: when the behaviour you need cannot be expressed as a property, a workflow, a routing rule or a report. Anything reachable through those four stays configuration — and we will tell you when that is the honest answer.
Service Hub ships a long feature list. This is the shorter list of things that decide whether the system works: the model, the enforcement, the deflection, the signal, the AI and the reporting.
The ticket data model comes first — properties, pipelines, stages, types and tier mapping — because everything downstream reads from it.
Response and resolution targets by priority and customer tier, with escalation that fires before the breach rather than reporting it afterwards.
Structure before content. We design the taxonomy around your highest-volume ticket categories, then wire article surfacing into the ticket and chat flow.
Surveys connected to workflows, and health scores built on signals that are actually maintained — not on checkbox properties nobody updates.
The Customer Agent is only as good as the knowledge base and ticket model beneath it. We build those first, then configure the escalation boundary.
Dashboards built around the questions leadership actually asks — which accounts need a conversation this week, and what support patterns precede churn.
Service Hub shares the CRM data model with Sales Hub and Marketing Hub, so a lot of this is native. The work is deciding which direction each signal travels and what it triggers when it lands — and building the bridge where the platform boundary is real.
| System | Direction | What moves | Why it matters |
|---|---|---|---|
| HubSpot Sales Hub | Service ↔ Sales | Ticket volume and sentiment, health score, open escalations; deal value, renewal date and owner in return. | The rep walking into a renewal sees the support history, and CS sees what the account is worth before it escalates. |
| HubSpot Marketing Hub | Service → Marketing | Churn-risk and detractor flags out; promoter identification into advocacy and referral programs. | A declining health score can start a win-back sequence instead of waiting for a cancellation email. |
| Salesforce | Two-way sync | Contacts, accounts and tickets or cases, with a defined system of record per object and per field. | Teams running both platforms stop reconciling two customer records by hand. This is where most dual-stack builds go wrong. |
| Product & usage data | Product → Service | Login frequency, feature adoption and error events pushed onto the contact or company record. | Health scoring built on real usage is a leading indicator. Health scoring built on ticket count is a lagging one. |
| Billing & subscriptions | Billing → Service | Plan tier, MRR, contract dates and payment status onto the company record. | Tier-based routing and conditional SLAs need a plan field that is right today, not one someone typed in last year. |
| Zendesk · Freshdesk · Intercom | One-time migration | Ticket history, conversation threads, contact records and knowledge base content, mapped to the new model first. | Properties are mapped before any data moves, so reporting is continuous across the cutover instead of restarting at zero. |
| Slack & Microsoft Teams | Service → Ops | Breach-imminent alerts, escalation notifications and at-risk account digests. | An escalation that only appears in a dashboard nobody has open is not an escalation. |
Scroll the table sideways on a narrow screen. Dual-stack teams should start with the Salesforce Service Cloud page as well — the choice of system of record is the first decision, not the last.
Five phases, one engagement. Scope, milestones and go-live dates are fixed before any build begins — the durations below are typical mid-market ranges, not a quote.
How customers onboard, where they get stuck, what precedes non-renewal, and what your CS team does when it spots a risk. A working session with service and CS leadership — not a feature demo.
Properties, stages, types, priority logic and tier mapping, designed before a single workflow exists. This is the phase that decides whether your SLA reporting will be trusted in a year.
Pipelines, routing, escalation, knowledge base structure and portal go in, with data migrated and property mapping validated on a parallel run before cutover.
SLA enforcement, survey response workflows, and health scoring last — because a health score built on inputs that aren't yet reliable is worse than no health score at all.
Only once the model and knowledge base hold do we configure Breeze. Then monthly reviews: volume trends, routing as the team scales, and AI accuracy as the content changes.
Four engagements, described by the people who ran them. Each figure is scoped to the client it was measured on — none of them is a benchmark, and none of them is a promise.
We had a Service Hub that technically worked — tickets came in, reps responded — but our NPS was flat and churn crept up quarter after quarter. Twopir audited the architecture, rebuilt the health scoring model on actual churn signals, and connected the detractor workflow to a human CS touchpoint within four hours. Net Revenue Retention improved 14 points in two quarters.
From a shared Gmail inbox to a fully automated service system in eight weeks.
Our knowledge base had 200 articles and a 12% deflection rate. Twopir restructured the taxonomy, rewired the search logic and connected it to our ticket flow. Deflection reached 38% in 60 days. That is a material reduction in support cost, and our agents now handle the complex issues instead.
Customer health scoring that flagged churn six weeks before renewal.
Service Hub work is worth doing when churn, SLA accountability or support cost has become a board-level number. These are the situations we are usually called into.
Acquisition is solved and churn is now the growth problem. You need support health connected to net revenue retention, not another dashboard.
Regulated environments where the ticket model has to account for care-team routing and provider–patient interaction tracking, not generic B2B support flows.
Client intake, matter communication and service accountability running through one system with SLA-level commitments, across US and UK practices.
Service interactions that carry compliance weight — audit trails, enforced SLAs and CS oversight without building custom tooling to get them.
Consolidating onto HubSpot and needing a migration that keeps ticket history, keeps reporting continuous, and lands faster than an internal build.
Support running across US, UK, UAE and APAC hours, where routing, business calendars and SLA clocks have to reflect more than one time zone.
Service Hub implementations rarely fail on features. They fail because the data model, the enforcement logic and the reporting were never designed to hold up under real volume.
Every workflow, health score and feedback loop we build has a named retention or expansion outcome attached to it. If a feature has no outcome behind it, we do not turn it on for you.
Three of the four things most teams ask us to "build" are configuration, and we will say so. The custom development we do quote for is the work that genuinely cannot be expressed as a property, a workflow or a report.
When your stack spans HubSpot and Salesforce, one team designs the system of record for each object and each field. That decision is where dual-stack builds usually break, and it is not a decision two agencies can share.
A SaaS support motion, a law firm's client communication and a healthcare provider's care-team routing are three different data models. We have built all three, and we do not start any of them from the same template.
Routing that fits a team of eight breaks at twenty-five, and a Breeze agent degrades as the knowledge base drifts. Monthly reviews and ongoing optimisation are part of how we scope, not an upsell afterwards.
The platform-level view: what we build across HubSpot, and how to decide whether you need one hub or the whole stack.
ExploreThe other half of the renewal conversation — expansion signals from service reaching the pipeline the rep actually works.
ExploreBreeze across the whole platform, not just service — where the agents earn their credits and where they don't.
ExploreWhere the churn signal goes next: re-engagement built from health data rather than from a list someone exported.
ExploreThe whole motion — marketing, sales and customer success designed as one system rather than three that hand off.
ExploreThe other answer to the same question. If Salesforce is your system of record, start here — then come back for the bridge.
ExploreHubSpot Service Hub is HubSpot's customer service and success product. It covers a help desk workspace with ticketing and omnichannel conversations, SLA management, a knowledge base and customer portal, NPS, CSAT and CES surveys, customer health and success tooling, and Breeze AI including the Breeze Customer Agent. What separates it from a standalone helpdesk is that every ticket sits on the same CRM record as the deal, the marketing history and the lifecycle stage — so a support conversation and a renewal conversation are looking at one customer rather than two.
Most of our mid-market clients land on Professional, listed at $90 per seat per month billed annually. That tier is where real service automation begins: SLA management, knowledge base, customer portal, surveys and advanced routing are all available there. Enterprise, listed at $150 per seat per month, is the right call for high volumes, complex routing hierarchies, strict SLA policies and deeper permission and reporting control. Two things worth knowing before you buy: seats on both tiers split into Core Seats and Service Seats, and only the people who need the deeper service tooling need a Service Seat — and HubSpot's current seats model no longer imposes a minimum seat count on Service Hub, so older advice about a ten-seat Enterprise floor is out of date. Check the live figures with HubSpot before signing; we recommend the tier that matches the use case, not the highest licence fee.
A focused build — help desk setup, ticket pipelines, SLA configuration and core routing automation — typically runs 4 to 8 weeks. A full customer service system, adding knowledge base architecture, the customer portal, Breeze configuration, NPS, CSAT and CES workflows, customer health scoring and Sales Hub or Salesforce alignment, runs 10 to 16 weeks. A migration from Zendesk, Freshdesk, Intercom or Salesforce Service Cloud sits inside those ranges rather than on top of them, because the property mapping happens during the data model phase either way. Scope, milestones and go-live dates are fixed before any build begins.
The practical test is whether the behaviour you need can be expressed as a property, a workflow, a routing rule or a report. If it can, it is configuration — pipelines, SLA policies, escalation paths, survey logic, health score weightings, dashboards and the whole knowledge base structure all sit on that side of the line, and none of them needs a developer to maintain. Custom development starts where that vocabulary runs out: custom objects for service processes a ticket cannot model, CRM cards and UI extensions that put your own data inside the help desk, private apps, custom-coded workflow actions, API integrations with product, billing or provisioning systems, and a two-way Salesforce data architecture. We scope and price those separately because they are engineering work, and we will tell you when the honest answer is that you do not need any.
Yes, and it is the engagement we take on most often. The pattern is consistent: pipelines that multiplied without a design, SLA targets that display but never escalate, a knowledge base nobody uses, an AI agent answering from stale content, and health scores that do not match which accounts are actually at risk. We do not recommend starting over. Our free audit takes five business days and returns a written document covering the ticket data model, SLA enforcement gaps, automation and routing health, knowledge base structure and deflection rate, feedback setup and response gaps, and Breeze readiness — with a prioritised fix list and the effort behind each item. Often the answer is a handful of targeted rebuilds rather than a new portal, and the audit is what tells you which.
Inside HubSpot the connection is native rather than an integration: Service Hub, Sales Hub and Marketing Hub share one CRM data model, so a ticket, a conversation and a health signal are already visible to the rep who owns the account and the marketing team running re-engagement. The design work is deciding what each signal triggers — churn risk starting a CS intervention and a marketing win-back, expansion signals from a CS conversation creating a pipeline opportunity. Salesforce is a genuine integration and needs a real decision: which platform is the system of record for contacts, accounts and cases, field by field. We define that first and build the two-way sync against it, which is where most dual-stack service builds go wrong.
Since 14 April 2026 HubSpot has priced Breeze Customer Agent on outcomes: 50 credits per resolved conversation, around $0.50 at the standard credit rate, and conversations it does not resolve cost nothing. That changes the calculation — the risk is no longer wasted licence spend, it is a bad answer reaching a customer. It is worth switching on when your knowledge base is consistently tagged and current, your ticket properties for priority, type and tier are reliable, and lifecycle stage and health data on the contact record are maintained. Without those three, the agent answers from stale content, escalates inconsistently and creates tickets rather than closing them. We build the foundation first and configure the escalation boundary explicitly, so it hands off at a point you chose.
The audit covers ticket architecture, SLA enforcement, routing health, knowledge base and deflection, feedback response gaps and Breeze readiness. Written findings in five business days, with a prioritised fix list. No cost, and no commitment to proceed.
US EST · UK GMT · AEST overlap hours — response within 24 hours
