All In
FR EN
Back to blog
allinmedia 17 min read

Last-Click Attribution Is Hiding Most of What LinkedIn Actually Built

| By Patrick de Carvalho

Contents

I have run LinkedIn campaigns for clients whose sales cycle outlasts the ad campaign that started it by six months. If you report on those campaigns with last-click, you will conclude the ads did nothing, right before the rep closes a deal that started with an impression the reporting tool never counted. This is not a rare edge case. It is how most complex B2B purchases actually happen, and it is why LinkedIn built three separate tools, the Conversions API, the Revenue Attribution Report, and a model called LiDDA, to fix a measurement problem the last-click model was never built to solve.

In short: Last-click attribution credits the final touch before a form fill or a sale, which in B2B is almost never the touch that mattered. LinkedIn's Conversions API (CAPI) recovers conversions lost to cookie restrictions, adding a 31% increase in attributed conversions on average. The Revenue Attribution Report connects your CRM to Business Manager, now including HubSpot alongside Salesforce and Dynamics 365. LiDDA, LinkedIn's transformer-based attribution model, found 150 times more credit for upper and mid-funnel campaigns than last-click ever showed. None of this replaces judgment. It gives judgment better numbers to work with.

What last-click gets wrong in a B2B deal

Last-click attribution gives 100% of the credit to whatever a buyer clicked right before converting. On a same-day e-commerce purchase, that shortcut barely distorts the picture. On a B2B purchase involving six to ten stakeholders and dozens of research sessions spread over months, it distorts almost everything.

Gartner's 2024 research puts the self-directed share of a B2B buying journey at roughly 80%, up from 57% in 2015. Buyers now do most of their evaluating before a salesperson ever hears from them. The same research counts 6 to 10 stakeholders per purchase decision and an average of 27 touchpoints across channels before a deal closes. A LinkedIn video ad seen in month one, a Thought Leader Ad commented on in month two, a Sales Navigator message opened in month three: none of that shows up in a last-click report if the final action was a direct visit to your pricing page in month five.

Three consequences follow, and they compound each other.

Upper-funnel campaigns look worthless. Awareness content, video, and Thought Leader Ads almost never win the last click. A finance team reading only last-click numbers will cut the budget line that started the relationship, because the report cannot see it.

Budget drifts toward the bottom of the funnel. Money moves to the channels that already look productive: branded search, direct traffic, retargeting. Pipeline dries up eighteen months later, and nobody can explain why, because the cause was a decision made on incomplete data long before the effect showed up.

Data gets lost before it even reaches the report. Safari's Intelligent Tracking Prevention, Firefox's Enhanced Tracking Protection, and ad blockers strip a growing share of client-side pixel events before they are ever recorded. What last-click measures is shrinking on top of already being the wrong metric.

Last-click Rule-based multi-touch LiDDA (data-driven)
Credit to upper funnel Close to none Split by a fixed rule (linear, U-shape) Weighted by measured influence
Counts view-through impressions No Partially Yes, modeled
Built for long B2B cycles No Workable with a wide window Purpose-built
Setup effort None Moderate High (needs CAPI and CRM data)

CAPI: recovering the conversions the pixel never sees

The Conversions API, CAPI for short, is LinkedIn's server-side tracking system. Instead of relying on a JavaScript pixel firing in the visitor's browser, your server sends conversion events straight to LinkedIn over an API, bypassing everything that can block a browser-based tag.

Client-side tracking through the Insight Tag alone loses signal at every step: blocked cookies, refused consent, ad blockers, private browsing. CAPI works around that by sending first-party identifiers, most often a hashed email address, directly from your server.

According to LinkedIn's own Conversions API playbook, advertisers who deploy CAPI see a 31% increase in attributed conversions and a 20% drop in cost per action on average. A beta study run in November and December of 2024 measured a 39% decrease in cost per qualified lead among participating advertisers. MarketerHire, a talent marketplace, connected CAPI to its CRM through Zapier and reported a 30% decrease in cost per qualified lead along with a 35% improvement in the conversion rate from booked appointment to qualified buyer, according to a case study published on Zapier's blog.

Three ways to deploy it, in order of technical lift:

  1. Direct API integration. Your engineering team sends events through HTTP POST calls. The most flexible route, and the right one if you have a developer on staff or on retainer.
  2. A partner integration. Dreamdata, Factors.ai, CustomerLabs, and Stape all offer pre-built connectors that cut setup time to a few hours.
  3. Server-side Google Tag Manager. If you already run an sGTM container, a LinkedIn CAPI tag can push conversions server-side in parallel with your existing client-side pixel.

One detail trips up almost every first deployment: deduplication. When the browser pixel and CAPI both fire for the same event, LinkedIn can count it twice unless both flows share the same conversion ID. Set that up before you trust a single number the report gives you.

The Revenue Attribution Report: your CRM, inside Business Manager

The Revenue Attribution Report, RAR in LinkedIn's own shorthand, is a report inside Business Manager that connects your CRM directly to your advertising data. It answers the question a sales director actually asks: how much pipeline, and how much closed revenue, did our LinkedIn campaigns actually touch?

Here is what changed since the last time most agencies wrote about this: as of mid-2026, the RAR connects to Salesforce, Dynamics 365, and now HubSpot, according to LinkedIn's own Marketing Solutions help documentation. HubSpot support closes a gap that forced smaller B2B teams into third-party workarounds for years. If your last briefing on this tool said HubSpot was unsupported, that briefing is out of date.

The report groups its numbers into three categories.

Category What it shows
Topline Revenue won, return on ad spend, pipeline amount
Funnel Leads, open opportunities, closed-won opportunities
Conversion Lead conversion rate, opportunity win rate, average days to close

RAR marks a CRM deal as "LinkedIn-influenced" using an any-touch model: if any contact tied to that deal engaged with your LinkedIn marketing, through an impression or an engagement such as a click, a comment, or a landing page visit, inside the chosen lookback window, the deal counts. The default lookback window is 180 days, and LinkedIn's help documentation confirms it can be extended, a range built for cycles that stretch well past a single quarter.

Setting it up takes three things: an active Business Manager account, a connected CRM with the native LinkedIn connector, and matching naming conventions between your ad campaigns and your CRM lead sources. Skip the third one and the first two are decoration.

LiDDA: LinkedIn's own model for crediting the touches that aren't the last one

LiDDA, LinkedIn Data Driven Attribution, is the company's proprietary attribution model, published as a research paper on arXiv in May 2025 by a team of LinkedIn engineers led by John Bencina. Where rule-based models like linear or U-shaped attribution split credit according to a fixed formula decided in advance, LiDDA uses a transformer architecture, the same family of neural network behind most modern language models, to learn which touchpoints actually moved a buyer toward a decision.

The model combines two approaches that used to live in separate teams: bottom-up attribution at the individual member level, and top-down media mix modeling that accounts for macro factors like seasonality and overall spend. It ingests sequence, member profile, company size, industry, campaign type, and ad format together, instead of scoring each touchpoint in isolation.

LinkedIn's own engineering blog reports the result of running LiDDA against its internal marketing data:

"Initial results show a 150x increase in credit found in Modeled Attribution which paces well with Marketing's spending increase during this time frame." LinkedIn Engineering Blog, "Buyer Journey Insights with Data-Driven Attribution"

Video ads, display, and social content, the channels last-click had flattened to near zero, picked up most of that credit. LinkedIn's internal team estimates the shift contributed a 5% lift in marketing-driven revenue for fiscal year 2025 from better in-quarter budget decisions alone.

LiDDA is still rolling out account by account, not universally available yet. It also needs volume: a model trained on your sequence data performs best above roughly $5,000 in monthly spend. Below that threshold, a manually configured multi-touch model remains the more realistic option, and there is no shame in that. A data-driven model starved of data is not more accurate than a rule, it is just harder to explain when it is wrong.

Hashing is not anonymizing: the privacy layer, briefly

CAPI sends identifiers, usually an email address, hashed with SHA-256 before it leaves your server. Worth saying plainly, because the phrasing gets abused in vendor decks: hashing is not the same as anonymizing.

Under UK GDPR, the Information Commissioner's Office confirmed in its May 2025 guidance that hashed data is pseudonymized, not anonymous, and pseudonymized personal data stays inside the scope of data protection law. The EU GDPR treats it the same way. On the other side of the Atlantic, LinkedIn's advertising agreement commits both parties to comply with applicable privacy law, the EU GDPR and the California Consumer Privacy Act by name, for any data processed through CAPI.

The practical takeaway for a B2B marketing team in any of the three jurisdictions: hashing an email before sending it does not remove your obligation to have a lawful basis for processing it, a clear privacy notice, and a data processing agreement with LinkedIn. It reduces the exposure surface. It does not close the file.

The All In Attribution Stack: five layers, in order

Sequence matters here more than any single tool. Skip a layer and the ones above it report numbers you cannot defend in a board meeting.

Layer Tool Rough setup time What it needs
1. Server-side collection LinkedIn CAPI (direct, sGTM, or a partner) 1 to 3 days Server access or an sGTM container, plus an existing Insight Tag
2. Extended attribution windows Campaign Manager settings 30 minutes Admin access to Campaign Manager
3. CRM connection Business Manager plus Salesforce, Dynamics 365, or HubSpot 1 to 2 days A configured CRM and an active Business Manager account
4. Third-party multi-touch Dreamdata, Factors.ai, or a comparable tool 2 to 5 days Ongoing subscription budget
5. Decision-ready reporting A dashboard (Looker Studio, Power BI) 1 to 3 days Connected data sources from layers 1 through 4

Layer 1, collection. Deploy CAPI alongside the Insight Tag. Send server-side events for form submissions, MQLs (marketing qualified leads, meaning a lead that hit a scoring threshold set by marketing), SQLs (sales qualified leads, meaning a lead a rep has personally validated), and created opportunities. Share one conversion ID across both flows to avoid double counting. Success looks like a gap under 5% between server-recorded conversions and what your CRM shows.

Layer 2, windows. Set click-through to at least 90 days and view-through to at least 30 for standard B2B cycles. For the RAR specifically, the 180-day default is a reasonable floor; extend it further if your cycle runs past eight months. Success looks like 20% to 40% more attributed conversions than the default settings produced, without every campaign suddenly claiming credit for everything.

Layer 3, CRM. Connect Salesforce, Dynamics 365, or HubSpot to Business Manager. Map campaign fields consistently. Confirm lead statuses, MQL, SQL, opportunity, closed-won, are populated correctly before you trust anything downstream. Success looks like a pipeline figure in the RAR that lands within 15% of what your sales director already reports.

Layer 4, third-party attribution. For a cross-channel view beyond LinkedIn, or if HubSpot support still doesn't cover your exact use case, add Dreamdata (twice recognized at LinkedIn's own Partner Marketing Awards, most recently in 2026 for B2B Thought Leadership) or Factors.ai, which integrated with LinkedIn's Company Intelligence API in November 2025. Success looks like identifying the first three touchpoints on at least 70% of your closed-won opportunities.

Layer 5, reporting. Consolidate into one dashboard built around four numbers: pipeline ROAS (return on ad spend, meaning influenced pipeline divided by LinkedIn budget), revenue ROAS (closed revenue divided by the same budget), cost per influenced opportunity, and the average time from first touch to closed-won. Success looks like presenting these numbers to a leadership team without translating LinkedIn's vocabulary into business vocabulary first.

Full deployment across all five layers runs one to two weeks for a team with some technical support. Layers 1 and 2 alone can be live in a few days and already correct the worst distortions.

Where this breaks: five mistakes worth naming

Deploying CAPI and stopping there. Better collection with a 30-day attribution window still throws away everything that happens on day 31. The two fixes are not substitutes for each other.

Treating "influenced" as "caused." The RAR tells you a contact engaged with LinkedIn before becoming a lead. It does not prove LinkedIn caused the deal. A $500,000 influenced pipeline figure means $500,000 in opportunities touched your campaigns somewhere along the way, not that LinkedIn single-handedly generated $500,000. Say it that way in front of a finance team, or expect a hard question you cannot answer.

Letting CRM and LinkedIn use different definitions. If your CRM marks a lead qualified after a behavioral score crosses a threshold, and LinkedIn counts a form submission as the conversion, the two systems will never agree on a number, and every meeting will restart the same argument.

Dropping last-click too fast. Keep it running in parallel for three to six months during any transition. It is the number your finance team already trusts, and you will need it to explain the gap when the new model reports something different.

Underestimating CRM data quality. Attribution is only as good as what goes into it. Duplicate contacts, opportunities with no dollar amount attached, and inconsistent pipeline stages will corrupt every report built on top of them. Clean the CRM before you connect it, not after.

What I don't know

I don't have a controlled, third-party study measuring LiDDA against last-click outside LinkedIn's own reporting. The 150x figure and the 5% revenue lift both come from LinkedIn's engineering blog, describing LinkedIn's own internal marketing spend. That is a credible primary source for what happened inside one company. It is not independent replication, and nobody outside LinkedIn has published one yet that I could find.

I don't know your specific numbers. The 31% and 20% CAPI figures are LinkedIn-wide averages; the MarketerHire result is one company's result. Every deployment I have seen produces a different number once you account for industry, deal size, and how clean the CRM was going in.

I don't know exactly which lookback window range the RAR now supports beyond the confirmed 180-day default; LinkedIn's help documentation states the window is extendable without publishing the full list of options as of this writing. If you need an exact figure for a specific account, check Business Manager directly rather than trust a number circulating on a blog, including this one.

FAQ

What is LinkedIn's Conversions API (CAPI) and why does it matter for B2B?

CAPI is a server-side tracking system that sends conversion events from your server directly to LinkedIn, bypassing the browser limitations that block a growing share of pixel-based tracking. LinkedIn reports a 31% average increase in attributed conversions and a 20% drop in cost per action for advertisers who deploy it, which matters most for B2B teams whose long cycles already lose data to attribution windows that are too short.

Does the Revenue Attribution Report work with HubSpot?

Yes, as of 2026. The Revenue Attribution Report now connects to Salesforce, Dynamics 365, and HubSpot through LinkedIn's Business Manager, according to LinkedIn's own Marketing Solutions documentation. Earlier guidance describing HubSpot as unsupported is outdated.

What attribution window should a B2B company use on LinkedIn?

For sales cycles running three to nine months, set click-through to at least 90 days and view-through to at least 30 days in Campaign Manager. For the Revenue Attribution Report, the 180-day default lookback window is a reasonable starting point, and LinkedIn confirms it can be extended for longer enterprise cycles.

What is LiDDA and can any advertiser access it?

LiDDA, LinkedIn Data Driven Attribution, is LinkedIn's transformer-based model that assigns credit to touchpoints based on their measured influence rather than a fixed rule. It found a 150 times increase in credit for upper and mid-funnel campaigns compared to last-click, according to LinkedIn's engineering blog. It is rolling out progressively and performs best above roughly $5,000 in monthly ad spend.

Is hashed data in LinkedIn CAPI compliant with GDPR and CCPA?

Hashing reduces risk but does not remove the data from the scope of privacy law. The UK's Information Commissioner's Office confirmed in May 2025 that hashed data is pseudonymized, not anonymous, and stays subject to GDPR. LinkedIn's advertising agreement commits both parties to GDPR and CCPA compliance for CAPI processing, but the advertiser still needs a lawful basis and a clear privacy notice.

How long does it take to set up a full LinkedIn attribution stack?

Plan one to two weeks for all five layers: server-side collection, extended attribution windows, CRM connection, third-party multi-touch tooling, and consolidated reporting. Server-side collection and attribution windows alone, the two layers with the fastest payoff, can be live within a few days.

Sources

  1. Gartner, B2B buying journey research (2024), self-directed journey share, stakeholder count, and touchpoint data, as compiled in Brixon Group, "The Modern B2B Buying Journey", 2026.
  2. LinkedIn Marketing Solutions, "LinkedIn Conversions API Playbook for Marketers", 2025: 31% conversion increase, 20% CPA reduction, 39% cost-per-qualified-lead decrease in beta.
  3. Zapier, "LinkedIn Conversions API: Scale lead gen with Zapier", 2025: MarketerHire case study.
  4. LinkedIn Marketing Solutions Help, "The Revenue Attribution Report in Business Manager", 2026: HubSpot, Salesforce, and Dynamics 365 support, attribution model description.
  5. LinkedIn Marketing Solutions Help, "Revenue Attribution Report metric definitions from CRM data", 2026: 180-day default lookback window.
  6. Bencina, J. et al., "LiDDA: Data Driven Attribution at LinkedIn", arXiv, May 2025.
  7. LinkedIn Engineering Blog, "Buyer Journey Insights with Data-Driven Attribution", 2025: 150x credit increase, 5% FY25 revenue lift.
  8. Information Commissioner's Office (UK), guidance on anonymisation and pseudonymisation, May 2025, cited in Taylor Wessing, "New ICO guidance on anonymisation and pseudonymisation".
  9. Dreamdata, "Dreamdata Wins LinkedIn Partner Award for B2B Thought Leadership", 2026.
  10. BusinessWire, "Factors.ai Integrates With LinkedIn's Company Intelligence API for Full-Funnel B2B Attribution", November 2025.

All In: measure what LinkedIn built, not what happened to get clicked last

An attribution stack does not replace a sales director's judgment. It gives that judgment numbers built on 27 touchpoints instead of one.

All In is the B2B media that decodes LinkedIn, expert blog, weekly podcast and newsletter for SME leaders and sales directors who want to turn LinkedIn into measurable growth. An original creation by Patrick de Carvalho, on LinkedIn since 2004. Motto: "I Never Lose."

Discover All In and get what matters on LinkedIn, every week.