Open three tabs and ask the same question: how many opportunities did last month’s LinkedIn campaign produce?
Google Ads and LinkedIn report conversions in the hundreds. GA4 shows a fraction of that. Your CRM shows 11 opportunities, 4 of which sales says came from a webinar.
Nobody is lying. The three systems are counting different things, over different windows, using different rules about who gets credit and the gap between them is where marketing budgets go to die in board meetings.
Closing that gap has nothing to do with attribution models, but rather with plumbing. Ad platforms know what they served but not what closed. Your CRM knows what closed but not what was served. Something has to sit between them, match records, respect the lag between click and revenue, and produce one number both marketing and finance will accept.
That something is a reporting stack, and there are more viable architectures than most teams realize. Series A company with a spreadsheet and disciplined UTMs can answer the question adequately. A company spending 7 figures needs a warehouse and probably an incrementality program. Buying the enterprise answer at the wrong stage is the most common and expensive mistake in this category.
These are 7 stacks that directly connect spend to pipeline, ordered from lightest to heaviest, with what each one costs you in effort and where each one breaks.
Why Your 3 Dashboards Disagree
Before choosing a stack, understand what you are reconciling. 6 structural differences explain nearly every discrepancy.
Attribution windows. Ad platforms count a conversion within a set window after a click or view. Your CRM records the opportunity when it was created, which in B2B may be months later and outside every platform window.
View-through credit. Meta and LinkedIn count impressions that preceded a conversion. Your CRM has no concept of an impression at all.
Self-reporting. Each platform grades its own homework, and each claims the same conversion. Add up the platform numbers and you will exceed your actual lead count, sometimes considerably.
Identity loss. Cookie restrictions, privacy tooling, and the shift to multi-device research break the link between an ad click and a form fill more often than dashboards imply.
Lead-to-account matching. 3people from the same company convert on 3 different campaigns. Platforms see 3 conversions; the CRM sees 1 account and eventually 1 deal.
Lag. With a 90-day sales cycle, the campaign that produced this quarter’s revenue ran last quarter. Any report comparing spend and revenue in the same period is answering a different question than the one asked.
None of these is fixable by choosing a better attribution model. They are fixed, or at least made honest, by the data pipeline underneath the model.
That is worth sitting with, because the instinct when numbers disagree is to shop for a platform. Every stack below inherits these 6 problems. The better ones make the discrepancies visible and explainable rather than pretending they resolved them, and a vendor claiming to have eliminated them is describing a product that does not exist.
Quick Comparison
| # | Stack | Typical Fit | Build Effort |
| 1 | CRM-native | Under ~$20k/month spend | Days |
| 2 | Connector and dashboard | $20k–$75k/month, several channels | Weeks |
| 3 | Closed-loop conversion | Any spend with a good CRM | Weeks, ongoing |
| 4 | B2B attribution platform | $50k+/month, dedicated RevOps | Weeks to months |
| 5 | Warehouse-native | Enterprise or data-mature teams | Months |
| 6 | Incrementality | $100k+/month, mature programs | Ongoing |
| 7 | Conversational over the stack | Any stage with clean data underneath | Days on top of an existing stack |
1. The CRM-Native Stack
Components: UTM discipline, CRM campaign objects, native CRM reporting
Typical fit: Under roughly $20,000 a month across 2 or 3 channels
Build effort: Days
What It Connects
Ad clicks to CRM records through consistent UTM parameters captured on form submission and written to hidden fields, then rolled up in native CRM campaign reporting. HubSpot and Salesforce both support this out of the box, and native attribution tooling covers first-touch, last-touch, and simple multi-touch views.
Where It Breaks
Anonymous research before the form fill is invisible, and so is anything self-serve. It also degrades the moment UTM discipline slips one campaign tagged inconsistently and the report quietly misallocates spend for a quarter.
Best For
Early-stage teams and anyone who has not yet earned the right to a more complex answer. Most companies skip this and buy a platform before their tagging is trustworthy, which produces sophisticated reports built on unreliable inputs.
2. The Connector and Dashboard Stack
Components: A data connector such as Supermetrics or Windsor.ai, a BI layer such as Looker Studio, plus CRM exports
Typical fit: $20,000–$75,000 a month across several channels
Build effort: Weeks
What It Connects
Spend and performance data from every ad platform into one dashboard, joined to CRM data on campaign identifiers. It removes the manual export step that consumes a day and a half of every month-end.
Where It Breaks
The join. Connectors are excellent at pulling platform data and indifferent to whether your campaign naming matches your CRM. Without a naming convention enforced across platforms, the join produces confident nonsense. Blended metrics also hide channel-level truth, so keep per-channel views alongside the roll-up.
Best For
Teams whose main pain is assembly time rather than attribution logic. This is the highest return per hour of work in the whole list, and it is usually the difference between a monthly report that takes a day and one that takes twenty minutes.
3. The Closed-Loop Conversion Stack
Components: CRM as source of truth, offline conversion imports, Conversions API and enhanced conversions
Typical fit: Any spend level with a functioning CRM
Build effort: Weeks to build, ongoing to maintain
What It Connects
CRM outcomes back to the ad platforms. When an opportunity is created or a deal closes, that event is pushed to Google, Meta, and LinkedIn against the original click identifier, so the platforms optimize toward qualified pipeline rather than form fills.
Where It Breaks
It depends entirely on click identifiers being captured and stored at form submission and on CRM stages being maintained consistently. Sales hygiene becomes a marketing dependency, which is an organizational problem more than a technical one.
Best For
Every B2B SaaS team running automated bidding. Strictly speaking this is an optimization stack rather than a reporting one, but it belongs here because it forces the same plumbing and it is what makes platform automation useful. Growth-onomics treats it as prerequisite work: until the conversion signal reflects pipeline, better reporting only documents the wrong outcome more precisely.
4. The B2B Attribution Platform Stack
Components: A dedicated platform such as Dreamdata, HockeyStack, or Factors.ai, connected to ad platforms, CRM, and web analytics
Typical fit: $50,000+ a month with dedicated RevOps support
Build effort: Weeks to months
What It Connects
Everything, in principle: ad platforms, CRM, marketing automation, website behavior, and often product usage, unified into account-level journeys with multi-touch models across first-touch, last-touch, linear, and position-based views. Account-level identification is the key differentiator, since B2B credit belongs to a buying committee rather than an individual.
Where It Breaks
Assumptions about your funnel shape. Some platforms assume form-fill and sales-led paths, so teams with significant self-serve signups or offline touchpoints such as events and outbound must push that data in manually. Pricing is sales-led at the mid-market and enterprise end, and implementation quality depends on CRM data hygiene you may not have yet.
Best For
Mid-market and enterprise teams with long cycles, multi-stakeholder buying, and enough budget that a 10 to 20 percent reallocation exceeds the platform cost.
5. The Warehouse-Native Stack
Components: Ingestion (Fivetran, Airbyte), warehouse (BigQuery, Snowflake), transformation (dbt), BI (Looker, Hex), reverse ETL (Hightouch, Census)
Typical fit: Data-mature organizations with engineering support
Build effort: Months
What It Connects
Raw data from every source in one place you control, modeled to your own definitions. Attribution logic becomes code you can inspect and change, rather than a vendor’s black box, and the same models feed reporting, activation, and finance.
Where It Breaks
Ownership. It requires analytics engineering to build and maintain, and marketing teams without that support end up with a stale pipeline and a dashboard nobody trusts. It is also slow to produce a first answer, which is fatal if leadership needs numbers this quarter.
Best For
Organizations that already have a warehouse and a data team, or those whose reporting requirements are genuinely bespoke. Several attribution vendors now run warehouse-native or export cleanly into one, which is often the better path than building everything from scratch.
6. The Incrementality Stack
Components: Geo holdout tests, platform conversion lift studies, marketing mix modeling tools such as Recast, Prescient, or open-source Robyn
Typical fit: $100,000+ a month with mature programs
Build effort: Ongoing, by design
What It Connects
Spend to causal effect rather than to credited touchpoints. Instead of asking which touch preceded the deal, it asks what would have happened without the spend, the only question that actually justifies a budget increase.
Where It Breaks
Statistical power. Holdouts and MMM need volume and time to produce reliable results, and below a certain spend the confidence intervals are wide enough to be useless. It also answers channel-level questions rather than campaign-level ones, so it complements attribution rather than replacing it. Expect months before the first defensible read, and organizational patience to match.
Best For
Teams whose attribution reports have stopped being believed, or who need to defend a significant budget decision with evidence stronger than platform-reported credit.
7. The Conversational Stack
Components: An AI assistant connected via MCP or an aggregation platform, sitting over the warehouse, CRM, and ad platforms
Typical fit: Any stage, provided the layer underneath is trustworthy
Build effort: Days on top of an existing stack
What It Connects
Nothing new. It is an interface over whatever you already built, letting a marketer ask “which campaigns produced opportunities over $25k last quarter, and what did they cost” and get an answer without a dashboard request or a SQL query.
Where It Breaks
Underlying data quality, amplified. An assistant will answer confidently from a broken join, and its fluency makes the answer harder to doubt than a dashboard would be. It also needs governance around who can query what, since a conversational interface removes the friction that used to limit access.
Best For
Teams with a functioning stack and reporting bottlenecks instead of data bottlenecks. The layer should be added last, not first.
The Plumbing Every Stack Needs
Whichever architecture you choose, 5 things determine whether the output is trustworthy. None of them is a purchase.
Naming conventions enforced everywhere. Campaign names and UTMs need one documented structure applied identically across every platform. Every join in every stack depends on this, and it is the single most common point of failure.
Click identifiers captured and stored. Capture GCLID, fbclid, li_fat_id and equivalents on form submission and write them to the CRM record. Without them, closed-loop conversion upload is impossible and much of your attribution is speculative.
Lead-to-account matching. B2B credit belongs to accounts. Your stack needs a reliable way to roll individual conversions up to the account and eventually to the opportunity.
Self-reported attribution. A “how did you hear about us” field on demo forms is cheap, imperfect, and frequently the closest thing to truth you will have for channels that never produce a trackable click which now includes a growing share of AI-assisted research.
Agreed lag windows. Marketing and finance must agree on how spend and revenue are compared across periods before anyone builds a report. Growth-onomics documents this in the reporting spec at the start of an engagement, because retrofitting a definition after the first board deck is where credibility gets lost.
How to Choose by Stage
Under $20k a month. CRM-native, done properly. Fix tagging, capture click identifiers, use native campaign reporting. Anything more sophisticated will out-resolve the reliability of your inputs.
$20k to $75k a month. Add a connector and dashboard layer to kill the assembly work, and build the closed-loop conversion feed. Together these solve most of the pain at this stage.
$75k to $150k a month. Evaluate a dedicated B2B attribution platform, provided you have RevOps capacity to maintain it and CRM hygiene worth modeling. If either is missing, make sure they are fixed first, otherwise the platform will not work.
Above $150k a month. Warehouse-native modeling for control, plus an incrementality program for the questions attribution cannot answer. Add the conversational layer once the numbers underneath are trusted.
At every stage the sequence is the same: fix the inputs, then automate the assembly, then add sophistication. Teams that invert this end up with expensive tooling producing numbers nobody defends.
Conclusion
The stack you need is determined by spend level and data maturity. A disciplined CRM-native setup at $15,000 a month tells the truth. An enterprise attribution platform sitting on inconsistent campaign names does not, however impressive the dashboard.
What every workable architecture shares is the plumbing: consistent naming, captured click identifiers, account-level rollup, a self-reported field, and agreed comparison windows. Those 5 decisions determine whether any of the 7 stacks produces a number your CFO will accept. None of them requires a purchase order.
The honest position on attribution is that it will remain imperfect. Buyers research anonymously, ask assistants questions nobody logs, and arrive through channels that pass no referrer. The goal is not certainty; it is a defensible, consistent view that improves budget decisions and survives scrutiny in the room where budgets are set.
If you want a reporting layer that ties paid spend to CRM pipeline, built on plumbing that holds up rather than a dashboard that looks impressive, the Growth-onomics team can audit what you have and design the stack that fits your stage.
FAQs
Why do my ad platform and CRM numbers never match?
Because they count different things over different windows. Ad platforms credit conversions within their own attribution windows, include view-through credit in some cases, and each claims the same conversion independently so summing platform numbers overstates reality. Your CRM records opportunities when they were created, often months after the click that started the journey, and rolls multiple individuals from one company into a single account. Add identity loss from privacy restrictions and multi-device research, and a gap is guaranteed. The goal is not making them match; it is choosing one system as the source of truth and understanding why the others differ.
Which system should be the source of truth?
The CRM, for anything involving revenue. It is the only system that knows what actually closed, and finance already trusts it. Ad platforms should be treated as operational data useful for optimization and diagnosis, unreliable for reporting outcomes. That means your reporting stack pulls spend and delivery data from the platforms, matches it to CRM records, and reports pipeline and revenue from the CRM side. Reversing this, and reporting platform-attributed conversions as marketing results, is how marketing ends up defending numbers sales does not recognize.
Do I need an attribution platform, or is my CRM enough?
CRM-native reporting is sufficient for most teams under $20,000 a month in spend, provided tagging is disciplined and click identifiers are captured. Consider a platform when three things are true: the budget is large enough that a 10 to 20% reallocation exceeds the platform cost, sales cycles are long enough that simple models mislead, and you have RevOps capacity to maintain the implementation. Buying one before CRM hygiene is reliable produces a more expensive version of the same unreliable answer, delivered with more confidence.
How do I account for AI-assisted research in reporting?
Imperfectly, and deliberately. A growing share of B2B research now happens in AI assistants that pass no referrer, so that influence surfaces as direct traffic, branded search, or nothing at all. 3 partial measures help: a self-reported attribution field on demo forms with an explicit AI assistant option, tracking branded search volume as a directional indicator, and monitoring AI visibility separately so you know whether you are present in those answers. None is attribution in the strict sense. Together, they prevent a real and growing channel from going invisible in your reporting.
Should marketing or RevOps own the reporting stack?
Ownership usually belongs with RevOps or data, while the definitions belong to marketing and finance jointly. The failure mode when marketing owns everything is a stack optimized to make marketing look good; the failure mode when data owns everything is technically correct reporting nobody uses. The workable split is that marketing specifies the questions and the definitions, RevOps or data builds and maintains the pipeline, and finance signs off on how spend and revenue are compared. Write those responsibilities down before building, because the argument is far more expensive after the first disputed board deck.
How long should the reporting lag window be?
Match it to your sales cycle, and agree it with finance before building anything. If median time from first touch to opportunity is 90 days, comparing this month’s spend to this month’s pipeline compares unrelated things. Most B2B teams report both a real-time operational view for optimization and a lagged cohort view for evaluating return, spend from a given month tracked forward against the pipeline it eventually produced. The cohort view is the one that belongs in a board deck, and the disagreement it prevents is worth the extra complexity.
