A field study of implementation work
The work behind a
HubSpot implementation.
What public partner reviews and job descriptions can tell us about the work around a CRM—and where the evidence stops.
By Jetstack AI
Original publication
Findings & methodology
Study design / not a results chart
From public text
to an interpretation.
- Source material
Partner reviews · freelance briefs · in-house roles
- Analysis
Keyword matching or model-assigned work tags
- Permitted inference
What the sampled text mentions or describes
Each source describes a different kind of work. They are not interchangeable measures of market demand.
Method & scope
Different texts.
Different questions.
The original study combined public descriptions of partner engagements with advertised freelance work and in-house roles. The collection date recorded in the original analysis is 10 October 2026.
The partner-review date range was 8 January 2015 to 9 October 2026; the final year was incomplete. Freelance posts covered July to October 2026. The in-house job-ad collection covered October 2026. These windows should not be compared as if they were equivalent annual samples.
Jetstack AI recorded a keyword recount on 13 October 2026. The figures here are attributed to that original analysis, not independently recounted. This page does not reproduce reviewer names or quote individual reviews.
What the method cannot establish
Realised consulting spend, delivery duration, implementation quality, causation, customer outcomes or time saved by using any particular tool.
Reading partner reviews
Count the language.
Do not overstate its meaning.
The original pass used explicit keyword groups. It did not infer that every occurrence was positive or that a single word described the whole engagement.
In the original analysis, onboarding or implementation terms appeared in 53.3% of 2026 reviews, compared with 37.9% in 2023. Speed or responsiveness terms appeared in 35.2%, compared with 19.6%. These are shares of each year’s reviews—not shares of all 41,026 reviews.
| Year | Reviews | Onboarding / implementation | Speed / responsiveness |
|---|---|---|---|
| 2019 | 3,025 | 14.4% | 20.8% |
| 2020 | 3,985 | 18.8% | 20.6% |
| 2021 | 3,494 | 28.8% | 20.6% |
| 2022 | 4,542 | 32.3% | 18.8% |
| 2023 | 5,524 | 37.9% | 19.6% |
| 2024 | 6,422 | 44.8% | 21.0% |
| 2025 | 6,733 | 50.7% | 30.8% |
| 2026 (partial) | 4,681 | 53.3% | 35.2% |
A mention is not a project brief.
The original keyword pass searched review titles and text for onboard, implement, set up and setup. These terms can describe a partner engagement, but do not establish its scope, duration or commercial value.
A keyword is not a sentiment score.
The speed group matched quick, fast, on time, ahead of schedule, turnaround and responsive. A negative sentence can still contain a matching word. The method measures mentions—not praise, delivery speed or productivity.
Topics overlap.
Workflow and report were keyword stems. Migration matched migrate, moved from and switched from; training matched train, teach and educate. One review can mention several themes, so categories are not mutually exclusive.
| Topic | 2023 | 2026 (partial) |
|---|---|---|
| Workflows | 5.4% | 12.4% |
| Training | 10.5% | 14.8% |
| Migration | 8.0% | 11.3% |
| Reporting | 3.4% | 6.8% |
A review saying “not fast enough” still matches a speed keyword. Calling every match “praise” would exceed the method.
Classifying the work
A second pass asks
what was done.
The original classification used Claude Haiku to tag 5,000 reviews of at least 120 characters, including 4,000 written from 2023 onwards. Of those, 4,397 described the work and form the denominator below. Implementation and setup appeared in 45.3%; training and documentation in 31.5%.
Bar scale: 0–100%. Multiple tags per review; shares do not sum to 100%.
How to interpret model-assigned labels
A work label is an interpretation of text, not a validated invoice line. A single review may describe implementation, training and marketing services together. Shares across overlapping tags should not be summed into a market split.
Freelance briefs
A posted budget is
not a paid engagement.
The original analysis covered 9,842 unique Upwork posts from searches for 14 CRM and automation tools, July–October 2026. After excluding non-CRM work, 2,335 posts remained, including 399 HubSpot jobs. Workflow automation appeared in 52% of the 2,335 CRM jobs.
The median advertised fixed-price budget was $150 for both HubSpot jobs (97 stating a fixed price) and CRM jobs overall (716 stating a fixed price). For hourly HubSpot posts, the original analysis reported a median advertised upper rate of $38 per hour.
| Work type | HubSpot | All CRM |
|---|---|---|
| Reporting | 19% | 13% |
| Sales process | 18% | 7% |
| Email and campaign assets | 16% | 10% |
| Data clean-up | 14% | 7% |
| Migration | 11% | 6% |
Budgets and rates were amounts stated by posters—not verified awards, final payments or project values. This search-based sample is not a price benchmark for partner implementations.
In-house roles
A hiring signal.
Not an adoption measure.
The October 2026 collection covered 1,078 distinct full-time ads on Greenhouse, Lever, Ashby and SmartRecruiters, including 581 in-house CRM roles. Workflows and automation appeared in 46.1% of those 581 roles; reporting appeared in 40.6%.
Bar scale: 0–100%. Categories overlap; shares do not sum to 100%.
The sample leans toward venture-backed technology companies, with Salesforce more prominent than HubSpot. Role descriptions are hiring signals, not measures of actual work completed or technology adoption.
Source register
Where the analysis points.
These are the dataset and analysis references recorded by the original study. File references describe provenance; they are not download links or a claim that the raw material is publicly available.
| Source | Method | Boundary |
|---|---|---|
| HubSpot Solutions Partner reviews ↗ | Keyword matching in review title and text; grouped by review year. | Public reviews are self-selected. Mentions do not measure sentiment, hours, spend or the full market. |
| Review work classification | Claude Haiku assigned multiple work tags to 5,000 reviews of at least 120 characters; 4,397 described work. | Model-assigned tags are interpretive, not invoice lines. Multiple tags can describe one review. |
| Freelance CRM job posts ↗ | Deduplicated posts from CRM/automation searches; filtered for CRM work and classified by work type. | Advertised budgets are not paid invoices. Search coverage is not a representative sample of all consulting demand. |
| In-house CRM job ads | Distinct ads from Greenhouse, Lever, Ashby and SmartRecruiters; separated into in-house CRM roles. | The sample leans toward venture-backed technology companies. Job-ad mentions do not establish adoption or implementation spend. |
For implementation teams
Keep diagnosis, delivery
and verification distinct.
The practical distinction does not need a market-size claim. Reviewing an inherited portal, implementing a configuration and testing its behaviour are different tasks.
- Portal Audit
Read-only evidence and a branded report.
- Projects
Reviewable plans, separate approvals and durable runs.
- Portal reuse
Destination matching and reference remapping, one portal at a time.