Arthur Duboks works on go-to-market (GTM) engineering: building and repairing systems that connect sales and marketing tools, automate lead workflows, and improve conversion tracking. Their posts focus on making these systems work reliably with real customer data and existing CRMs.
They are most useful for practical GTM engineering explanations about tracking, CRM wiring, and automation with real data.
Best for
This fits a B2B SaaS company whose paid-social and pipeline reporting disagrees with what sales sees.
Not for
This does not fit consumer or DTC brands seeking high-reach sponsored video with a rate card.
Key takeaways
A model wrote 2024 into a CRM from an undated out-of-office reply.
A Salesforce task field only accepted preset values, so generated tasks said Other.
Meta attribution change can lower reported ROAS without campaigns getting worse.
Check first: Confirm whether they accept sponsorships or paid work and how to price and book it.
Main ideas
Most GTM stacks are disconnected, not under-equipped · The skill is knowing what not to automate · The prompt is the easy part; the code around it is the job
Arthur Duboks writes about go-to-market (GTM) engineering: connecting sales and marketing tools, automating lead workflows, and making conversion tracking more useful.
They show where these systems fail in practice, not just how they look in a demo.
Examples include an out-of-office date sent to a CRM as a date in the past, a Salesforce field that rejected generated task names, and Meta reports that can diverge from backend sales data.
Follow their work for practical explanations of how GTM systems behave with real data, existing tools, and human decision-making.
Duboks makes the unglamorous details of automation visible and explains why they matter.
They pair failures with concrete safeguards: validate a model's date before writing it to the CRM, test a build on the buyer's actual CRM, and send pricing questions to a person rather than letting AI invent a quote.
They also make room for exceptions.
In one post, they argue that a retry loop is unnecessary when drafts are already passing quality checks; in another, they caution that the effect of Meta's attribution change varies by account.
This mix of operational detail, human handoffs, and qualified claims gives readers useful questions to ask before trusting a workflow or its reported results.
Arthur Duboks writes about GTM engineering: building and repairing the systems that connect sales and marketing tools, and the judgment calls those systems depend on.
Read more
Across the posts, they argue that buying tools is easier than wiring them, that prompts are the copyable part of an automation while the checks around them are the work, that ad-platform numbers undercount conversions unless the brand owns its tracking, and that builds fail more often on CRM constraints and company permissions than on technology. They also predict a market shift toward repairing other people's GTM builds.
Most GTM stacks are disconnected, not under-equipped
Arthur Duboks argues that companies buy tools faster than they connect them.
Read more
Procurement is easy to budget; integration work is not, so it goes undone and the company shops again. They recommend mapping what talks to what before buying anything else: the gap is usually a couple of integrations wide, not a missing platform.
An executive told Arthur Duboks, "We have very good dots everywhere.
We have nothing that connects them."
The setup was better than most: Salesforce clean, Apollo paid for, a sequencer running, an intent tool bought in Q2.
But Apollo was used only for contact lookups, enrichment credits were burned at month-end so they would not expire, and two reps split a territory with no rule for who takes what.
Arthur Duboks argues that automating everything makes a demo look good and a system fragile.
Read more
Deciding what a person still handles is a large part of the job. The useful version of AI in a GTM system is narrow: it handles research and drafting that people do by hand, does it in minutes, and leaves anything uncertain to a human.
On one build, most replies handled themselves: an interested reply paused the sequence, updated the CRM and pinged the right rep.
A pricing question always went to a person, because an AI inventing a price is a discount nobody signed off on.
Duboks also left a retry loop out of the drafting agents because 14 of 14 drafts passed QA, with 10 to 15 percent low-confidence output as the threshold for adding one.
On a separate call, a marketing lead pushed back on an AI plan, then apologised for it.
Duboks says the scepticism was right: the lead had spent a year reading posts claiming agents replace teams.
The prompt is the easy part; the code around it is the job
Arthur Duboks argues that prompts are the most copyable part of a GTM build and the least likely cause of failure.
Read more
Failures live in the configuration around the model: a webhook sending POST while the endpoint expects GET, a CRM node reporting success after dropping a field, a date parsed without a year.
The remedy is code that checks what the model did, not a better prompt, and testing on the client's real CRM rather than a clean demo account.
That CRM carries rules nobody can redesign.
An out-of-office reply said the person was out until April 29, with no year.
The model wrote 2024 into the CRM, so the follow-up was booked for a date that had already passed.
The system gave no error or indication.
On a Salesforce build, a task subject field accepted only preset values, so every task the system created said "Other."
Deduplicating thousands of existing leads became a project run by the client's operations team.
Ad-platform numbers undercount conversions, so own the tracking
Arthur Duboks argues that ad-platform dashboards miss a real share of conversions, and that brands should own the tracking infrastructure rather than trust the browser pixel.
Read more
Safari's tracking prevention, third-party cookie blocking and ad blockers stop browser-side events; sending them from the brand's own server through Meta's Conversions API creates a first-party path.
For lead generation, they describe importing closed deals from the CRM back to the ad platform so campaigns optimise toward revenue instead of form fills.
A related post explains Meta's move to count only link clicks for click-through attribution, with likes, shares, comments and saves in a separate engaged-view bucket.
A brand spends $8K a month on Meta ads, sees 2.2 ROAS in the dashboard and about 3.8 in its backend. Arthur Duboks attributes the difference to blocked browser tracking and cites Meta's estimate that 20 to 50 percent of browser-side conversions go unreported, with some sources putting the figure at up to 70 percent on mobile.
On the attribution change, Arthur Duboks resists a blanket prediction: "honestly I'm not sure how bad it gets account to account," suggesting DTC click-to-site campaigns are probably less affected than the panic implies and that engagement-heavy placements face the real exposure. They advise pulling conversion data from before and after rollout and measuring the gap: earlier decisions about paused ads and shifted budgets relied on the inflated read.
A build stalls on people and permissions, not technology
Arthur Duboks argues that GTM projects usually stall on procurement and ownership.
Read more
Buyers should ask who can spend and what stays on their own plate.
Their list of client-side work that never appears in a proposal or timeline includes half-duplicated leads already in the CRM, deciding what a good account looks like, someone answering the phone when the system is unsure, and saying no to requests that break the logic.
To judge the build, they say to bring in the most technical person in the room before signing, not at handover.
Two things nearly delayed one build: a tool licence that needed the CTO's approval, and sending domains that needed a card, which meant asking a CFO who was away.
A separate walkthrough describes a CTO with no stake in the pitch sitting silent for twenty minutes, then interrupting to ask whether one workflow was the whole thing.
Duboks treats that uninformed read as the useful one and recommends asking for a build run on your actual CRM.
The next market is repairs and commercial maturity
Two posts look ahead to where the work is going.
Read more
Arthur Duboks predicts the best-paid GTM engineering work next year will be repairs: building became cheap, people move on, a vendor raises prices or a CRM field is renamed, and flows break with nothing written down.
They compare the work to WordPress sites and advise learning to read other people's builds rather than collecting another certificate.
A second post describes private equity shifting its post-close focus from the CFO to the CRO, with B2B software judged on pipeline velocity, CAC payback, net revenue retention and sales process repeatability rather than product or TAM slides.
The repairs post's concrete failure is a renamed CRM field that quietly breaks two flows after the original builder has left.
The private-equity post cites its own framing data: 58 percent of sponsors prioritising revenue growth, CRO demand up 41 percent year over year, exit multiples compressed and debt costs higher, leaving sponsors needing 12 percent or more EBITDA growth to hold historical returns, with 53 percent of GTM teams reporting zero lift from their tech stack against an average $535K a year on tools.
The post cites unnamed survey and executive-search data; ask for the underlying sources before relying on those figures.
One post steps outside the GTM theme to argue that showing more options can lose customers rather than win them.
Read more
Arthur Duboks describes a curve where too few choices suppress buying and too many cause people to freeze, with a working range of roughly three to ten for most products. They recommend reducing what is visible at once rather than removing products: lead with the best three and let the rest exist without making people wade through them.
Adding a third pricing tier dropped conversions by 20 percent, with bounce rate up, add-to-cart down and support tickets asking which option to choose. Arthur Duboks cites buying-behaviour research on the number of choices and describes the response as the brain wanting to feel like it chose, not to choose.
See how Arthur connects an attribution gap to specific implementation choices, and what to investigate when ad-platform results disagree with sales data.
Read more
Arthur contrasts a Meta dashboard showing 2.2 ROAS with backend results closer to 3.8, then explains how browser tracking can miss purchases. They describe sending conversion events through Meta’s Conversions API and, for lead generation, returning closed-deal data from a CRM so campaigns can optimize toward revenue rather than form fills.
For brands planning a campaign
For an attribution review, compare platform-reported results with backend sales and check whether server-side events and CRM outcomes are connected.
Read this alongside the tracking post to see how Arthur addresses measurement uncertainty: they explain a concrete reporting change and give readers a way to test its effect.
Read more
This post separates link-click attribution from conversions credited after likes, shares, saves, or comments. Arthur recommends comparing conversion data before and after the change rather than assuming every campaign has the same exposure.
For brands planning a campaign
When assessing paid social, preserve a before-and-after view and identify which campaign decisions depended on the old attribution method. This helps distinguish a reporting change from a real performance decline.
This is a useful starting point for understanding Arthur’s view of automation quality: the revealing example is a quiet failure in a real workflow, not a polished prompt demo.
Read more
Arthur describes an automation that read an out-of-office date without a year and wrote 2024 into a CRM, causing a follow-up date to fall in the past.
Their proposed safeguard is a small code check after the model’s output.
They also point to webhook defaults and CRM fields that report success despite dropping needed data.
For brands planning a campaign
For an AI-enabled lead workflow, test ambiguous dates, unexpected payloads, and whether CRM updates actually persist. Budget for validation and readable logs, not just prompt writing.
See why Arthur’s approach is not to automate every step.
Read more
The pricing example makes the human handoff—and the risk it prevents—easy to understand.
In Arthur’s example, common replies update the CRM, pause a sequence, or alert a sales rep.
A pricing question goes to a person instead of receiving an AI-drafted answer, because price depends on the account and an invented quote could create an unauthorized discount.
Arthur also explains why they did not add a retry loop when the drafts were passing QA.
For brands planning a campaign
Map which replies can be handled safely and which need human judgment. Define escalation rules and quality thresholds before adding more automation or retry logic.
This post shows how existing CRM constraints shape an automation project, and why buyers should test with their own system before signing off.
Read more
Arthur describes a Salesforce task field that accepted only preset subjects, so generated tasks were labeled “Other.”
They also recount handling a large backlog of leads without scoring fields and explain how existing status fields passed work between systems.
Their recommendation is to test a build on the client’s actual CRM rather than a clean demo account.
For brands planning a campaign
Before estimating a GTM automation project, inspect the real CRM’s field rules, existing data, and handoffs. Include data cleanup and operational ownership in the plan; a successful test on a clean account may not transfer.
Conversion tracking and attribution amid browser and platform changes2
Buying and running a GTM build inside a company2
Where the work is heading: repairs and commercial maturity2
Connect the GTM stack, don’t buy another tool1
Offer design and choice reduction1
Connect the GTM stack, don’t buy another tool
A recurring argument: most companies already own the pieces of a GTM stack, but nothing is wired together.
Read more
In one post, an executive tells Arthur Duboks, "We have very good dots everywhere.
We have nothing that connects them," and they walk through the setup: Salesforce clean, Apollo paid for, a sequencer running, an intent tool bought in Q2 — but Apollo used only for contact lookups, enrichment credits burned before they expired, and two reps splitting a territory with no rule for who takes what.
The conclusion: buying a tool is procurement, while wiring it into everything else is engineering work that never gets its own budget line.
So it never gets done, and the company goes shopping again.
The suggested first move is to map what talks to what before buying anything else: the gap is usually two integrations wide, not a missing platform.
For brands planning a campaign
A brand bringing Duboks in for automation should expect an architecture and integration audit before any product recommendation. Briefs that assume a specific new platform may be redirected toward connecting what the brand already owns, with a map of systems and handoffs more likely than a tool purchase as the deliverable.
Several posts argue that the skill in GTM automation is deciding what not to automate.
Read more
Duboks describes a build where most replies take care of themselves — an interested reply pauses the sequence, updates the CRM and pings the right rep; wrong person, out of office and meeting booked get the same treatment — but a pricing question ("How much is it?") always goes straight to a rep untouched, because an AI inventing a price on the fly is a discount nobody signed off on.
The same build deliberately left a retry loop out of the drafting agents because 14 of 14 drafts passed QA; the stated threshold for adding one was 10–15% low-confidence output.
A separate post about an AI-agent meeting describes a marketing lead who pushed back on the AI plan, then apologised for it.
Duboks says the scepticism was correct: demos work because they run once on clean data, and the useful version is narrow — research and drafting done in minutes, with a human seeing anything the system is unsure about.
Duboks suggests judging a build by its edge-case behaviour, not its prompt: ask what the system does when it receives an out-of-office reply, and what happens on lead 500.
For brands planning a campaign
This is a reliability and human-in-the-loop position, not a fully autonomous one. Brands worried about AI making unapproved commitments, quoting prices or mishandling exceptions will find the reasoning useful; a brief promising hands-off automation will meet a narrower counter-proposal.
Conversion tracking and attribution amid browser and platform changes
Two posts deal directly with measurement gaps between ad platforms and backend revenue.
Read more
One works through a case where a brand spends $8K a month on Meta ads, sees 2.2 ROAS in the dashboard and closer to 3.8 in its backend, and attributes the difference to Safari's Intelligent Tracking Prevention, third-party cookie blocking and ad blockers.
It argues that Meta estimates 20–50% of browser-side conversions go unreported (with some sources cited at up to 70% on mobile).
The proposed fix is Meta's Conversions API with server-side tracking — Google Tag Manager Server-Side, hosting on AWS or a custom pipeline — with the brand owning the server rather than routing through a third-party app.
For lead generation, the same post describes importing closed deals from the CRM back to Meta so optimisation targets closed revenue rather than form fills.
The second post covers Meta's attribution change, where only link clicks qualify for click-through attribution and likes, shares, comments and saves move to a separate engaged-view bucket.
It notes that reported ROAS drops by an amount that depends on how much conversion volume came through non-link interactions.
Pull conversion data from before and after a platform attribution rollout and measure the gap: past decisions about paused ads and shifted budgets were built on the inflated read.
For brands planning a campaign
This is the most direct match for a brief about fixing attribution. The observable work includes server-side tracking setup, first-party event routing and CRM-to-ad-platform feedback for closed deals, plus a measurement review whenever an ad platform changes its attribution rules.
Two posts argue that CRM is where GTM builds fail.
Read more
In one, a Salesforce task subject field accepted only values from a fixed list, so every task the system created said "Other"; thousands of existing leads had none of the new scoring fields, and deduplication became its own project, built by the client's operations team.
Duboks used status fields to pass work between systems — marking a lead completed moves the sequence on, while marking it out of office pushes the follow-up back — and says plainly that the build has to live in the client's CRM, even when it would have been set up differently.
The other post gives a concrete failure: a model parsed an out-of-office reply with no year as 2024, booking a follow-up for a date that had already passed, with no error and nothing visible in the workflow.
Related examples include a webhook defaulting to GET, a CRM node reporting success after dropping a field, a missing company field, a website behind a login and a job title that changed since the data was pulled.
Check the fix after the model acts — for example, use a few lines of code to validate a parsed date — rather than trying to improve the prompt. When judging a vendor, ask to see a build run on your actual CRM, not a clean test account.
For brands planning a campaign
Brands with legacy CRM configuration, inconsistent data or strict field rules should treat this as the realistic part of the work. Useful scoping questions include who owns the CRM, how much cleanup the client is expected to do, and which edge cases the system must handle without a person.
Two posts cover the non-technical failure points of buying and operating a GTM system.
Read more
One lists what stays on the client's plate and never appears in the proposal or timeline: the existing half-duplicated leads in the CRM, deciding what a good account looks like, somebody picking up the phone when the system is unsure, the picklist nobody has touched in years, and saying no to the request that breaks the whole logic.
The suggested diligence: ask the vendor what stays on your side and put a name next to it in the timeline.
The other post blames a delayed build on procurement, not technology: someone needed a licence and had to wait on the CTO, while sending domains required a card, which required a CFO who was away.
The advice is to ask who can spend and what happens when that person is out for two weeks.
A related walkthrough post adds a technical due diligence angle: get your most technical person into the room before signing, because a CTO with no stake in the pitch is the one person who will actually inspect the separate flows, the normalisation step, the error paths and the schema.
For brands planning a campaign
Duboks is a useful voice for procurement-readiness and vendor-diligence content aimed at operators and buyers. The audience includes the person signing the contract as much as the person building the system, making this a fit for content about scoping, ownership and approval bottlenecks.
Where the work is heading: repairs and commercial maturity
Two posts make forward-looking arguments about the market for this work.
Read more
One predicts that the best-paid GTM engineering work next year will be repairs: building became cheap, agencies and newly certified freelancers shipped systems that worked on day one, and problems arrive months later when a vendor raises prices, the original builder moves on or a renamed CRM field quietly breaks two flows with nothing written down.
The comparison is to WordPress sites, and the advice is to learn to read other people's builds rather than collect another certificate.
The other post describes private equity's shift from the CFO to the CRO as the defining post-close hire, citing 58% of sponsors prioritising revenue growth, CRO demand up 41% year over year, compressed exit multiples and higher debt costs requiring 12%+ EBITDA growth, and B2B software being evaluated on pipeline velocity, CAC payback, net revenue retention and sales process repeatability.
It also cites 53% of GTM teams reporting zero lift from their tech stack and an average scale-up spending $535K a year on tools.
These figures are presented as the post's own framing, sourced to unnamed executive-search and survey data, so they should be treated as the creator's argument rather than verified statistics.
For brands planning a campaign
The repair argument supports positioning Duboks for auditing, documenting and fixing systems someone else built — a fit for briefs about broken attribution or stalled automation. The PE post widens the audience to investors and PE-backed software operators weighing commercial maturity before an exit.
One post sits outside the GTM-systems theme and deals with ecommerce conversion: adding a third pricing tier dropped conversions by 20%, with bounce rate up, add-to-cart down and support tickets asking which option to choose.
Read more
Duboks cites research comparing buying behaviour across numbers of choices and describes a curve where too few options suppress buying and too many cause people to freeze, with a working range of roughly three to ten for most products. The recommendation isn’t always to remove products, but to reduce what’s visible at once — lead with the best three and let the rest exist without making people wade through them.
This is the only post about consumer/ecommerce offer design rather than B2B GTM systems, automation or attribution. It does show the same habit of grounding advice in a specific observation with a number attached.
For brands planning a campaign
Relevant to brands working on pricing pages, bundles or plan structures, this shows a wider interest in marketing and conversion beyond technical systems work. It is thinner evidence than the GTM material, so use it as supporting colour rather than the main pitch.
The posts show the tools Arthur Duboks uses in their GTM and attribution work: Meta's ad platform and conversion tracking, Salesforce as the CRM their builds must fit into, and Apollo, the Anthropic API, and PredictLeads, which they keep available for client work.
Read more
They also list Google Tag Manager Server-Side, AWS, and third-party Shopify apps as server-side tracking options available to any advertiser, not as their own relationships. Ask Duboks to clarify any sponsorship, affiliate, ownership, employment, investment, or named-client relationships before treating these platforms as commercial ties.
Meta
Mention
Duboks examines Meta's ad platform through the lens of attribution: browser-side pixel gaps caused by tracking prevention and ad blockers, the Conversions API fix, and Meta's change to how it counts click-through conversions after non-link interactions.
Read more
They explain what these changes mean for reported ROAS and what advertisers should check before and after rollout. The posts describe an advertiser spending on Meta ads and state no commercial relationship between Duboks and Meta.
One client build Duboks describes ran on Salesforce, where the task-subject field accepted only values from a fixed list and thousands of existing leads lacked Duboks's scoring fields.
Read more
They also name Salesforce as part of a prospect's existing stack during a review. The CRM belongs to the client; no employment or client relationship with Salesforce is stated.
Duboks names the Anthropic API as a tool they keep available across builds so client work isn't blocked while procurement catches up.
Read more
They reference a specific Anthropic model when describing an out-of-office date error in a client build. No commercial relationship with Anthropic is stated.
Arthur Duboks writes short, first-person posts about the systems connecting sales and marketing software: CRM constraints, automation that stops short of a person's judgment, and the gap between ad-platform dashboards and backend revenue.
Read more
Start with the posts about individual client builds: the executive who said the stack had "very good dots" but nothing connecting them, and the CTO who interrupted a walkthrough to ask whether one automated process was the whole thing.
They show how Duboks scopes a problem and where the difficulty lies.
Then read the two measurement posts on why Meta's reported ROAS and a brand's own backend numbers diverge, and what server-side tracking changes.
The forward-looking posts offer direction rather than method: one predicts the best-paid work will be repairing other people's builds; another frames the same work as a private-equity diligence concern.
Follow for GTM engineering, marketing automation, conversion tracking and CRM topics.
Both write from inside client engagements about the plumbing behind paid marketing: tracking that does not reconcile, and automation stacks whose costs and failure modes surface only at volume.
Read more
Samuel Adeitan's post on tracking audits before a holiday push walks through Meta Conversions API deduplication, new-versus-returning signals and offline order reconciliation.
A second post argues that pay-per-task tools and self-hosted automation engines are an architecture choice, not a software-licence line.
That's the same ground Arthur Duboks covers in discussions of server-side event routing and how a build performs under real load.
Samuel Adeitan's material targets performance agencies running ecommerce and holiday scaling, focusing on client media spend and agency operating costs.
Arthur Duboks writes more about B2B revenue systems—CRM field rules, lead handoffs and approval bottlenecks—and how to assess a vendor's build.
The advice therefore transfers to buyers evaluating work rather than agencies protecting client margins.
The closest subject overlap in this set: Bivas Kumar Roy writes about the same discrepancy Arthur Duboks describes—a Shopify count and a Meta Ads count for the same purchases—and recommends the same fixes: consistent Event IDs, verified deduplication in Events Manager, correct purchase values and currencies, and consent states for EU traffic.
Read more
Both distinguish a tracking problem from an attribution problem before proposing a solution, and both work on Meta CAPI and server-side measurement.
Bivas Kumar Roy presents the work as a structured diagnostic checklist, with a personal introduction and an invitation to connect.
The posts target EU businesses and agencies dealing with consent and privacy constraints.
Arthur Duboks tackles the same problem through stories about individual builds and calls, then extends beyond measurement to CRM constraints, automation judgment and procurement—topics Bivas Kumar Roy's posts do not cover.
Both argue that a platform's headline metric can mislead a buyer.
Read more
Muhammad Anas Raza's account of raising a $3 cost per lead to $16 to get leads that answer the phone, and a follow-up post listing cheap leads, weak offers and ten-minute diagnostics as the real drivers of paid performance, sit alongside Arthur Duboks's point that ads optimised for form fills rather than closed revenue teach the algorithm the wrong thing. Both ground the argument in a specific client number rather than general advice.
Muhammad Anas Raza stays inside the paid funnel: form design, qualification gates, offer quality and cost per sale.
Arthur Duboks treats the ad platform as one node in a wider system, focusing on the CRM, automation and the handoff between them.
The two are complementary rather than interchangeable reads.
Arthur Duboks has a LinkedIn following of 999, captured on 9 October 2026—the only measured audience figure available.
Read more
That suggests a small, practitioner-level placement rather than a reach buy.
The profile includes 13 LinkedIn posts dated 29 January to 7 October 2026, all aimed at people who build, buy or repair go-to-market systems.
For a B2B SaaS campaign focused on marketing attribution fixes and GTM automation, that concentration supports relevance, not distribution.
With a low four-figure following, any single sponsored post is likely to reach a small audience; the value depends on who sees it, not how many.
The posts do not identify who follows or engages with them.
There is no breakdown by geography, age, gender, seniority, employer type or company size, and no impressions, reactions, comments or reposts for individual posts.
Post topics alone cannot establish the audience’s professional makeup.
Ask for a LinkedIn follower demographics export covering industry, seniority, company size and geography, plus engagement numbers for two or three recent posts on topics and formats close to your campaign.
This will help assess whether the audience matches the marketers, RevOps leads and founders you need.
If employer, occupation or job seniority matters, request survey results or another documented source; a platform demographics export does not verify those attributes.
LinkedIn followers
999
Follower count shown on the LinkedIn profile · 2026-10-09
No partnership, sponsorship, or brand-collaboration formats are described in Arthur Duboks's posts.
Read more
The 13 posts are practical write-ups about go-to-market engineering — attribution, CRM wiring, and automation failure modes — with no media kit, rate card, sponsored-post offering, or way for brands to inquire about working together.
The only contact surface shown is the LinkedIn profile itself.
Before planning a paid engagement, confirm directly whether they accept sponsorships or paid work and what terms apply.
Good fit: a B2B SaaS company whose paid-social and pipeline reporting disagrees with what its sales team sees, and which wants a realistic plan for connecting tracking, CRM, and automation rather than another tool purchase.
Read more
Arthur Duboks's posts speak to nearly every part of that need.
The strongest match is measurement.
One post works through an $8K-a-month Meta spend where the dashboard shows 2.2 ROAS and the backend shows closer to 3.8, traces the difference to browser tracking loss, and lays out a server-side route: Conversions API, a server the brand owns, and CRM closed-deal data fed back to the ad platform so optimisation targets revenue rather than form fills.
A companion post covers Meta's shift to link-click-only attribution, explains why reported ROAS can drop without campaigns getting worse, and gives readers a concrete test: pull conversion data before and after rollout and look at the size of the gap.
For a company deciding whether its numbers are broken, misread, or both, this is unusually on-point practitioner content.
The second match is automation realism.
Posts cover a model writing 2024 into a CRM from an undated out-of-office reply and booking a follow-up in the past, a Salesforce task field that only accepted preset values so every generated task said "Other", and a large backlog of leads with none of the new scoring fields.
They describe validating the model's output with a few lines of code and accepting that much of the work is deciding what happens on messy payloads.
A pricing question always routes to a person because an invented quote is an unauthorised discount.
A retry loop was deliberately left out until low-confidence drafts rose above 10–15 percent.
For a company seeking automation, this is the counterweight to a vendor promising hands-off agents.
A third match is the pre-purchase view: map what talks to what before buying another platform, because the gap is usually two integrations wide; ask the vendor what stays on the client's side and put a name and date next to it; get the most technical person into the room before signing, because that person will inspect the flows, the normalisation step, the error paths, and the schema.
Two posts extend this to the commercial context: PE sponsors shifting value creation toward revenue and commercial maturity, and a prediction that the best-paid work next year is repairing systems someone else built badly.
That framing suits a campaign aimed at operators and buyers, not just builders.
The measured audience is a LinkedIn following of 999, observed in October 2026, with no demographics, impressions, or per-post engagement data.
The posts are short text posts on a single channel, making this a fair fit for a technical content collaboration or an advisory-style engagement, but a weak fit for a distribution or reach buy.
The posts do not mention a media kit, rate card, disclosed paid partnership, or inquiry route; confirm terms directly before planning paid work.
The strongest pitch is a creator whose credibility comes from close technical observation, not scale.
When to choose a specialist
Poor fit: a consumer or DTC brand that wants a high-reach creator for an awareness campaign with sponsored video, a published rate card, and fast turnaround on a fixed campaign plan.
Read more
Arthur Duboks's posts do not support that shape of engagement.
The only measured audience figure is a LinkedIn following of 999, captured 9 October 2026.
There is no geography, age, gender, occupation, seniority, or company-size breakdown, and no impressions, reactions, comments, or reposts for individual posts.
A low four-figure following on one professional platform is not a reach argument.
Across 13 posts, there is no mention of a media kit, rate card, sponsored-post format, sponsor policy, disclosed paid partnership, or inquiry route; the only contact surface shown is the LinkedIn profile.
Confirm directly whether sponsorship is available and how to price and book it before planning a standard creator placement.
The format points the same way.
The posts are short-form technical write-ups about attribution, CRM wiring, automation failure modes, and buying GTM systems.
One post sits outside that theme and covers ecommerce offer design—a third pricing tier dropping conversions by 20 percent—but is explicitly an outlier and still reads as operator advice with a research reference, not consumer lifestyle content.
The posts show no video, long-form articles, podcasts, talks, or other channel a visual awareness campaign would need.
The bad fit extends to two B2B-shaped needs as well.
A brand that wants fully autonomous AI handling of inbound replies, including pricing, will meet a narrow counter-proposal: pricing questions go to a person, and retries are only added above a low-confidence threshold.
A brand that has already chosen a platform and wants it implemented without an integration audit will be pushed toward mapping existing systems first.
Both positions appear in the posts, so this section should not promise a vendor-style implementation plan.
A worked campaign example
A worked angle for the good-fit need: a B2B SaaS company wants to find out whether its paid-social numbers are wrong and connect tracking, CRM, and automation without buying another platform.
Campaign angle — "Attribution reality check": a three-part practitioner series plus a short working checklist, published from Arthur Duboks's own account and co-designed with the client's demand-gen and RevOps leads.
Part one, what the gap looks like.
A post in the shape of the $8K-a-month example: the dashboard says 2.2, the backend says 3.8, and the difference is browser tracking loss rather than a reporting glitch.
The post names what to inspect: which browsers and blockers buyers use, whether events fire from a server the brand controls or through a third-party app, and whether closed deals from the CRM are fed back to the ad platform.
The client contributes the anonymised shape of its own gap.
Part two, is it a reporting change or a performance drop?
A post built on the Meta attribution shift: only link clicks now qualify for click-through attribution, while likes, shares, comments, and saves move to a separate bucket.
The test is a before-and-after pull of conversion data, plus a list of decisions made last year on the old read—paused ads, shifted budgets—that may now deserve a second look.
This is useful for the client's team as well as the audience.
Part three, the parts that break quietly.
A post using the out-of-office reply parsed as 2024, the task field that only accepted preset values, and the CRM node that reported success after dropping a field.
The closing test is the one described in the posts: ask what a build does on an out-of-office, and what happens on lead 500.
Deliverables beyond the posts: a one-page checklist the client's team can run without Duboks present, covering server-side event routing, CRM-to-ad-platform feedback for closed deals, exception handling for ambiguous dates and payloads, and a named owner for any manual approval step.
Plus a 45-minute scoping session with the client's most technical person in the room before any contract is signed.
The posts argue that this person, with no stake in the pitch, inspects the flows, the normalisation step, the error paths, and the schema—and that two questions in ten minutes beat three weeks of procurement.
Success test for the campaign: a reader from the client's target accounts can state whether their own gap is tracking loss, an attribution rule change, or a process failure, and name the next thing they would inspect.
Confirm terms, format, and any disclosure requirements directly before adding sponsored distribution.
The angle is built for a practitioner audience the size of Duboks's, not for reach.
A better choice when the hard part is data plumbing at agency or store scale, rather than CRM pipeline work.
Read more
Samuel Adeitan's two posts work through server-side event routing for a holiday peak: making browser and server events share the same event IDs so purchases are not counted twice, passing new-versus-returning customer tags into the bidding payload, and matching settled orders against ad-platform click data to see actual cash collected.
They then address the economics of that work: pay-per-task platforms burning six figures of tasks on a bulk import versus a self-hosted workflow engine, small scripts for string and deduplication work, and a queue to keep spikes from crashing downstream limits.
For a team whose pipeline cannot scale past prototyping, or planning a pre-holiday audit of tracking layers, that is closer to the job than the CRM and automation failure modes Arthur Duboks writes about.
The framing is agencies and ecommerce stores.
A B2B SaaS company gets little on CRM fields, lead handoffs, and sales follow-up—the ground Arthur Duboks covers with a Salesforce field that accepted only preset values and a lead routed to a person on pricing.
The range is narrow: two posts, both focused on infrastructure.
A better choice when the symptom is a numbers disagreement and the question is whether the cause is tracking, attribution, or both.
Read more
Bivas Kumar Roy sets out a diagnostic order for exactly that case—a report showing 82 purchases in the store and 103 in the ad platform—and walks through comparing attribution windows, checking whether Pixel and Conversions API sent the same purchase with matching event IDs, auditing whether Purchase fires more than once, testing payment redirects, and reviewing consent states for EU traffic.
The distinction is useful to buyers: fixing duplicate events will not solve an attribution difference, and changing attribution settings will not fix broken purchase tracking.
This approach helps identify the fault before paying anyone to fix it.
The scope stops at measurement.
There is nothing on automation, CRM builds, or lead routing—the second half of a B2B SaaS company's needs.
The examples are ecommerce- and Shopify-flavoured, and the two posts cover the same comparison from different angles rather than a wider body of work.
A better choice when the real question is lead quality and paid-funnel economics rather than measurement.
Read more
Muhammad Anas Raza's central case is a campaign producing 150 leads at $3 each that the sales team cannot use.
They deliberately pushed cost per lead to $16 by switching off the one-click auto-fill form, adding mandatory qualification gates for budget and timeline, and requiring manual phone entry.
Lead volume fell by around 70 percent, and the calls improved.
The follow-up post states the position plainly: cheap leads ruin businesses faster than expensive ones, and a bad offer breaks a good ad.
For a brand whose dashboards look fine but whose pipeline does not convert, that diagnosis is more direct than tracking repair.
There is little mechanism-level detail on the tracking infrastructure or automation build.
Arthur Duboks explains server-side event routing and what a model does with an undated out-of-office reply; this material stays at funnel economics and offer quality.
It is also aimed at founders and agency clients, with a diagnostic audit offered in exchange for contact, so a B2B SaaS team gets commercial framing more than an implementation path.
Arthur Duboks writes 13 short-form LinkedIn posts, each presented in full rather than as titles or excerpts.
The posts are first-person accounts of client builds, calls and walkthroughs, and most deal with GTM systems, CRM and automation constraints, paid-social attribution, or buying and repairing GTM systems.
One post covers ecommerce offer design and is the only post outside that theme.
The posts do not include publication dates, so the order and development of ideas within them cannot be established; comparison material from similar creators dates from September and October 2026, and the LinkedIn account was observed on 9 October 2026.
Every post is text; no video, audio, long-form article, talk or podcast is available, and there is no transcript.
Coverage is limited to a single platform and a small following, so the posts show what Arthur Duboks writes about and how, not reach or audience composition.
No media kit, rate card or disclosed paid partnership appears in the posts here, and they do not verify a partnership route.
The posts do not mention a signature book, show or earlier body of work, so this profile does not claim one.
