LinkedIn Ad Signals That Correlate With Deal Open Rate (and the Influence Thresholds to Track)

Ok, I had to front-load the "trust" signals here a bit, because trust is running very low lately. In LinkedIn ads revenue attribution, at least.

I've read about a dozen posts lately about how influenced pipeline doesn't make any sense. Because of the inflated claims that someone opened $47 million in pipeline from $2k in ad spend. And because "it's you who picks" the threshold, and there can be dozens of other reasons why the deal opened.

Some folks, like Kristof Bardos, felt pretty strongly about it 😅

...there was literally a sleuth of posts like this in my feed lately (well, for a while).

But while we're all dunking on influenced pipeline, the irony is that...

...all pipeline is influenced pipeline 🤷‍♀️

Yes, even "last touch" or "last click" attribution. Some folks still hail it as "the reason" why someone converted, because it's the last touchpoint before the conversion (which is like saying the reason why someone came home was the door). It is just a threshold that we assume correlates with deal open rate.

And it happens to be a very, very low threshold (1 click).

So last-click attribution is influenced pipeline too 🙃

TLDR: what LinkedIn ad engagement signals correlate most with deal open rate?

• Influenced pipeline is not a bad way to measure the impact of LinkedIn ads on revenue per se. Poorly set thresholds that are not data driven are.
• The real question is: does your threshold actually correlate with a higher deal open rate? (ZenABM's revenue attribution dashboards will now test this for you, every week.)
• But also: what other signals correlate with higher deal open rates, so you can use them to prioritise sales outreach, feed your outreach agent, and sync them directly to LinkedIn Campaign Manager (for instance through ZenABM, as a matched third-party audience for retargeting with warm campaigns)?

I asked Claude to find out. 979 LinkedIn ad signals, 118 companies, 1.17M target accounts, 7,466 deals in the last 180 days. And because the answer should come from your own data, not ours, we are now adding it to ZenABM's revenue attribution dashboards as the new Influence Coefficient (more on it below).

👉 Try ZenABM free for 37 days and see which signals correlate with YOUR deal open rate.

What we found

• 🟢 Influenced deals are bigger. Median $20.3k vs $7.0k for non-influenced. About 3x.
• 🟢 The default threshold doesn't hold up. 10 impressions in 180 days doesn't predict deals. Light reach (1 to 49 impressions) actually went with fewer deals (0.4x to 0.6x).
• 🟢 Engagement volume matters more than impressions. 10+ engagements: 3.3x more deals. 50+ engagements: 6.8x.
• 🟢 Format matters. Surprise, surprise: feed ads correlate with deal open rate a lot more than right rail ads. Carousel click: 7x. Image click + document click: 10.6x. Text and spotlight ads went with fewer deals.
• 🟢 No signal explains everything. 55% of deals in the study came from accounts that never saw an ad 🤷‍♀️ The correlations are real and statistically significant at such high account numbers, but weak (nothing reached phi 0.1; I explain the exact maths below).
• 🟢 The most meaningful signals at ZenABM: our default 10 impression threshold gave no lift (1.2x, phi 0.004, not significant). 5+ engagements gave 12.4x lift, phi 0.040 (significant at 2,351 accounts), and covered 15% of our deals.

Top signals to use as your influence threshold (by phi)

Top signals for sales triggers (by lift)

How to use LinkedIn ad engagement signals for sales outreach and retargeting

• 🟢 Prioritize calls. Your SDRs can't call 10,000 accounts. Give them the accounts that hit a top by lift signal this week. Few accounts, much better odds.
• 🟢 Automate outreach. Push a top by phi signal to your CRM and trigger a sequence the moment an account crosses it. More accounts, steady lift, enough volume to automate.
• 🟢 Retarget. Feed accounts with a signal into a warm LinkedIn audience and spend more where companies are already leaning in.

Does your influence threshold actually correlate with a higher deal open rate?

The question we should be asking is NOT whether influenced pipeline is good or bad. It's "does your threshold actually correlate with a higher deal open rate?" and "what LinkedIn ad engagement signals correlate most?"

Good news: it's totally measurable. And fairly easy to measure with Claude, actually.

The thing is, almost nobody measures it.

Because it used to be extremely difficult to measure B.C. (Before Claude, hehe), and old habits (or lack thereof) die hard.

Enter "Influence Coefficient": our own metric testing the correlations between different signals & YOUR deal open rate, on YOUR own data

Everything below in this post is a study I ran once, across all our customers. But your threshold should come from your own accounts and your own deals, not from an average of 118 other companies. And it should be checked again every time your campaigns change.

So we built it into ZenABM, and we are now adding it to ZenABM's revenue attribution dashboards. Your influenced pipeline will no longer rest on a threshold someone picked once, but on the signals that actually correlate with your deal open rate. The new Deals dashboard has three tabs: Influenced deals, Deal influence signals and Influence thresholds management. (It's rolling out to beta teams first. The screenshots below use a made-up example company, "Northbeam Software", with 41,212 target accounts and 271 first deals in the last 180 days.)

How it works

Every Friday, ZenABM tests your LinkedIn ad signals (impressions, engagements, clicks, ad formats, thought leader ads, ABM stages, and combinations of them) against your own first deals, looking at the 180 days before each deal. It also tests multi-channel signals (HubSpot visits by source and page views), but visits in the last 14 days before a deal don't count, because they are mostly the visit that created the deal.

Only signals that are statistically significant at your number of target accounts make the list. Combinations must also pass a split-half check: found on one half of your companies, confirmed on the other half.

Then every company gets one number from 0 to 100, its influence coefficient:

1. Score each signal. Signal coefficient = phi × log2(lift). Phi says how strong the link is. Lift says how big the effect is. log2 makes 2x worth 1, 4x worth 2 and 8x worth 3.
2. Add up the company's signals, best first. Raw score = s₁ + s₂ ÷ 2 + s₃ ÷ 3 + … Each extra signal adds less, because signals overlap (a company with 5+ engagements also has 1+ engagement).
3. Scale to 0 to 100. Coefficient = 100 × raw ÷ the highest raw score in your team. Your most influenced company is 100. A company with no signal is 0.

Example: "Harbor Logistics" had 4 signals before its deal.

Raw score 0.112. The top company in the team has 0.253, so Harbor Logistics' coefficient = 100 × 0.112 ÷ 0.253 = 44. Its $50,000 deal counts as $50,000 × 44% = $22,000 of influenced pipeline.

Influenced deals: two pipeline numbers, side by side

The first tab shows influenced pipeline two ways. From influence threshold: every deal whose company had 1+ significant signal counts in full ($2,402,500, 93 of 271 deals, 44% of $5,522,000). From influence coefficient: each deal counts at its company's coefficient, so a deal at 40 counts 40% of its value ($1,422,960, 26%).

The deals table is sorted by influence coefficient, with each deal's signals (strongest first), the number of signals, best phi, best lift, and impressions, clicks and engagements before the deal.

Deal influence signals: what correlates on YOUR data

The second tab lists every signal and combination that is significant on your accounts, with lift (and its 95% range), signal coefficient, coverage, phi and "significant at". You can sort by lift or by signal coefficient.

Every row has a Track signal button. One click turns it into a ZenABM Signal that is tracked every week, synced to your CRM (or Slack), and pushed as a retargeting audience to LinkedIn Campaign Manager.

Influence thresholds management: a data-driven threshold

The third tab suggests a threshold from your own data and compares it with the one you use today. In the example, "1+ influence signal" goes with 1.9x the deal rate (95% range 1.5 to 2.5, phi 0.028, significant at 5,037 companies). The current rule, 10+ impressions, gives 1.5x (phi 0.018) and needs 12,143 companies to be significant. One click switches your deal flag and CRM sync to the new rule, and the same rule then drives the influenced pipeline in your ZenABM revenue attribution dashboards.

It also answers "do more signals mean more deals?". In the example: 1 signal 1.3x (not significant), 2 to 5 signals 1.6x, 6 to 9 signals 2.3x, 10+ signals 4.1x.

So the question "does my threshold actually correlate with a higher deal open rate?" gets answered for you, on your own data, every week. Not once, by whoever set up the account.

Why I ran this study (and nearly had an identity crisis)

Considering the public backlash against "influenced attribution", and that ZenABM uses influenced attribution as the attribution method in its ABM revenue attribution dashboards, I had to test whether the attribution thresholds set by our users a) actually correlated with the deals they were opening, and b) had an impact on deal value (ACV).

So I spent a good chunk of last weekend looking for signals that correlate with a higher deal open rate...

...and I was actually prepared to have an identity crisis 😬

Now, imagine working on a tool (ZenABM 👋) whose main use case is ABM revenue attribution, and finding out that its attribution method makes no sense...

That would have been tough. But it turned out I don't need to have an identity crisis.

Claude found some interesting correlations. Not very strong overall, but significant at this sample size.

Here are the top 10 signals by lift and phi (I explain what this means below!) that can be used as "influence thresholds":

But here are also other LinkedIn ad engagement signals with meaningful correlations to deal open rate, which you can use to prioritize accounts for sales:

ZenABM makes it super easy to capture these signals with its new Signals feature, and syncs them back to the CRM (or anywhere else you choose).

You can create them easily in the interface or through ZenABM's MCP server (which has read and write functionality), and sync them as retargeting audiences directly to your LinkedIn Campaign Manager!

Try them free for 37 days here!

How I found these LinkedIn ad engagement signals

• I asked Claude to look at correlations between certain signals (and "influence thresholds") and deal open rates...
• ...in a cohort of 118 companies and 1.17M (million!) target accounts that were served LinkedIn ads in the last 180 days and had impressions (essentially, our customers with a connected CRM).
• Deals opened in the last 180 days: I asked Claude to look only at companies with open deals (first deal only, each deal counted once) in the last 180 days. 7,466 companies fell into this bucket. They were compared with all the target accounts reached with LinkedIn ads with no deal in the last 180 days.

Methodology: how to calculate the correlations (and find your significant "influence threshold")

To measure how strongly a LinkedIn signal relates to pipeline, use the phi coefficient, the standard correlation for two yes/no variables (it's Pearson's correlation applied to binary data).

For every signal, each of the 1.17M target accounts lands in one cell of a 2×2 table:

• had the signal and opened a deal (a),
• had the signal but no deal (b),
• no signal but opened a deal (c),
• had neither signal nor deal (d).

Phi = (ad − bc) ÷ √((a+b)(c+d)(a+c)(b+d)). It runs from −1 to 1, where 0 means no relationship, 1 means the signal perfectly predicts deals, and −1 means a negative correlation.

Unlike lift, phi only gets high when a signal is both predictive and common enough to cover a real share of your deals. That's why it's the right number to look at for defining "influenced pipeline".

Here are the metrics to consider when you pick your "influence threshold" and other signals to track:

• Lift: deal rate with the signal ÷ deal rate without it. Shows how much better the group converts. For example, for 50+ engagements: 4.3% vs 0.63% = 6.8x.
• Coverage: share of all deals that had the signal, a ÷ (a+c). Shows how much pipeline the signal explains. 50+ engagements explains only 1.2% of deals in our cohort.
• Phi: lift and coverage in one number. 0.02 here. Best in the study: 0.031. Nothing reached 0.1 ("strong").
• 95% range on lift: from the standard error of log lift. If it reaches down to 1x, there's no correlation. 5.5x to 8.3x is good.
• Significant at: (1.96 ÷ phi)², the number of companies you need before that phi is statistically significant. For 50+ impressions, that was ~10,100.
• Split-half check: with 979 rules tested, some win by luck. Find signals on a random half, then keep only those that hold on the other half.

Influence thresholds and signals tested

First, I tested the defaults in our product. Then I asked Claude to search for other signals to test. It ended up testing 979 distinct signal combinations, like different levels of engagement with different ad formats!

Default (customizable, of course!) influence threshold and ABM stages

These are fully customizable for ZenABM users. But when someone signs up, we get them started with some defaults, and some users never adjust them. So I used them as the starting point:

• Default influence threshold = 10+ impressions in 180 days.
• Default ABM stages:
• Identified: 1+ impressions.
• Aware: 50+ impressions.
• Interested: 10+ engagements OR 5+ clicks.
• Considering: 1+ conversion event.
• Selecting: 1+ deal.

As expected, the defaults didn't perform very well. Only the default "Interested" stage showed a significant lift:

Other influence thresholds and LinkedIn ad engagement signals tested

Every signal went through the same test, scored on lift (with a 95% range), coverage, phi and "significant at". In total, Claude tested 979 LinkedIn ad signals across 1.17M companies and 7,466 deals.

What we tested

• 🟢 791 threshold rules: every mix of 1 to 5,000 impressions, 1 to 100 engagements and 1 to 50 clicks.
• 🟢 231 ad format rules: 7 formats × impressions, engagements and clicks, plus every pair of formats.
• 🟢 86 hand-picked rules: thought leader ads, and the ABM stage a company reached before its deal.

Examples of signals we tested

• 🟢 200+ impressions AND 3+ clicks
• 🟢 1,000+ impressions AND 50+ engagements
• 🟢 Clicked a carousel ad
• 🟢 Clicked an image ad AND clicked a document ad
• 🟢 Engaged with a thought leader ad AND clicked a document ad
• 🟢 Reached the Interested ABM stage before the deal

What we found

• 🟢 Engaging is what counts. The more a company engages, the more likely a deal. 10 engagements: 3x more deals. 50 engagements: almost 7x.
• 🟢 Clicks count too. 10+ clicks: almost 4x. 20+ clicks: 5x.
• 🟢 Lots of reach plus engagement wins. 1,000+ impressions with 50+ engagements: almost 8x.
• 🟢 Some formats work much better. A carousel click: 7x. An image click plus a document click: over 10x.
• 🟢 Thought leader ads work best with other ads. A thought leader ad engagement plus a document ad click: almost 12x.
• 🟢 ABM stages work as they should. Interested: 3x. Considering: 4x. Selecting: 11x.
• 🟢 The biggest numbers come from the smallest groups. A 10x signal usually covers only a handful of deals. The signals that cover more of your pipeline have smaller lifts.
• 🟢 No signal explains most of your pipeline. More than half of all deals came from companies that never saw a LinkedIn ad. So even the best LinkedIn signal explains only part of your deals.

Top LinkedIn ad engagement signals to track

Top 10 by phi (correlation strength)

Top 10 by lift (highest impact signals)

How to read "Significant at": the number of target accounts you need before that signal can be told apart from luck. If your list is smaller, use the top by phi list, not the top by lift.

How to find your own signals with Claude + ZenABM

You don't need a data team for this. Connect the ZenABM MCP server to Claude, and Claude can read your company engagement (impressions, engagements and clicks by format) and your CRM deals. Then it runs the same test on your own accounts.

Here's the prompt:

Using ZenABM, pull all target companies from the last 180 days with their LinkedIn impressions, engagements and clicks (by ad format) before their first deal, their highest ABM stage before the deal, and whether they opened a deal. For each signal below, build a 2×2 table and calculate lift with a 95% range, coverage, phi and significant at = (1.96 ÷ phi)². Drop signals where significant at is bigger than my number of companies, or where the 95% range crosses 1x. Split companies in half at random, find the best signals on half A, and show me only the ones that still hold on half B. Rank the result by phi, then show a second list ranked by lift. Use these signals:

Reach

• 1+ impressions
• 10+ impressions
• 50+ impressions
• 200+ impressions
• 1,000+ impressions
• 5,000+ impressions

Engagement

• 1+ engagement
• 3+ engagements
• 5+ engagements
• 10+ engagements
• 20+ engagements
• 50+ engagements

Clicks

• 1+ click
• 3+ clicks
• 5+ clicks
• 10+ clicks
• 20+ clicks

Reach + engagement combinations

• 200+ impressions + 3+ clicks
• 200+ impressions + 10+ engagements
• 200+ impressions + 3+ engagements + 3+ clicks
• 200+ impressions + 10+ engagements + 3+ clicks
• 1,000+ impressions + 50+ engagements
• 5,000+ impressions + 50+ engagements + 20+ clicks

Ad formats

• Image ad: 1+ engagement
• Image ad: 1+ click
• Document ad: 1+ engagement
• Document ad: 1+ click
• Carousel ad: 1+ engagement
• Carousel ad: 1+ click
• Video ad: 1+ engagement
• Video ad: 1+ click
• Video ad: 10+ impressions
• Thought leader ad: 100+ impressions
• Thought leader ad: 1+ engagement
• Thought leader ad: 1+ click

Format pairs

• Image click + Document click
• Image engagement + Document engagement
• Document engagement + Image click
• Carousel 10+ impressions + Document click
• Thought leader ad engagement + Document click
• Thought leader ad click + Document click
• Thought leader ad engagement + Carousel click
• Thought leader ad engagement + Image ad engagement

Thought leader ad combinations

• Thought leader ad engagement + any other ad engagement
• Thought leader ad 10+ engagements + any other ad engagement
• Thought leader ad 100+ impressions + Interested stage or higher

ABM stages (highest stage before the deal)

• Interested
• Considering
• Selecting
• Interested or higher

How to pick your influence threshold

Don't rely on the default "10 impressions in 180 days". In our data, that's not a signal at all: those companies opened fewer deals than average.

• 🟢 Set the window to 2x your sales cycle. 90 day sales cycle: 180 day window.
• 🟢 Require engagement, not just reach. Our data points to 200+ impressions AND (3+ clicks OR 10+ engagements). That rule has a 3x to 4x lift and the best phi of all reach + engagement rules.
• 🟢 Check it on your own data. Ask Claude to test 3 to 5 threshold options and pick the one with the highest phi that is significant at your account count (or let the Influence Coefficient tab do it every week).
• 🟢 Write it down and don't change it every quarter. A threshold you move to make the numbers look good is not a threshold.

In ZenABM, you set the impression threshold and the window in your CRM sync settings, and every deal gets a "LinkedIn influenced" flag based on your numbers. The threshold counts impressions only, so set it to 200 and use the Interested ABM stage (10+ engagements or 5+ clicks) for the engagement part.

Finally: our own influence threshold at ZenABM

We ran the same test on our own account: 38,110 companies, 47 first deals in the last 180 days, median deal $199 (self-serve). Small numbers, so read it as a strong hint, not proof.

• 🟢 Our old rule didn't work. We used the default: 10 impressions in 180 days. It marked half of our deals as "influenced". But those companies opened deals at the same rate as everyone else (1.2x, could be luck).
• 🟢 Reach alone did nothing for us. Companies we reached but who never engaged opened fewer deals than average (0.7x).
• 🟢 Engagement did everything. 1+ engagement: 7x more deals. 5+ engagements: 12x. 3+ clicks: 11x.
• 🟢 5+ engagements is our best signal. 532 companies, 15% of our deals, phi 0.040. That's double the phi of the best cross-account signal, and it's significant at 2,351 companies, well under our 38,110.
• 🟢 Image and thought leader ads were the only reach that helped. About 3x. Text ads reached 43% of our companies and went with fewer deals.
• 🟢 Our new rule: influenced = engaged at least once in 180 days. It covers 17% of our deals with a 7x lift. Fewer "influenced" deals than before, but every one of them means something.
• 🟢 Your number will be different. Our deals are $199 self-serve. A $50k enterprise deal has a longer cycle, a bigger buying group and a different threshold. Run the test on your own data.

Conclusion: influenced pipeline matters, if the threshold is right

• 🟢 Influenced deals are bigger. Across all accounts, the median LinkedIn influenced deal was $20.3k. Deals with no influence: $7.0k. That's about 3x.
• 🟢 But a bad threshold makes it meaningless. 10 impressions marks half your deals as "influenced", and those companies don't open more deals. That's the number CFOs rightly laugh at.
• 🟢 A good threshold makes it defensible. Require real engagement (200+ impressions AND 3+ clicks or 10+ engagements). Fewer deals get the label, but each one shows a 3x to 4x lift you can prove.
• 🟢 Last touch still misses most of it. Most of these accounts never click through and fill in a form. Without an influence number, that work shows up as zero.

Why track signals at all

• 🟢 For sales prioritization. Your SDRs can't call 10,000 accounts. Signals tell them which 500 to call this week. Use the top by lift list: few accounts, much better odds.
• 🟢 For automated outreach. Push the signal to your CRM and trigger a sequence the moment an account crosses it. Use the top by phi list: more accounts, steady lift, enough volume to automate.
• 🟢 For retargeting. Feed accounts with a signal into a warm LinkedIn audience, and spend more on the companies already leaning in.

ZenABM can help you with all of the above. It not only detects the signals across all your target accounts on a regular basis, but also passes them on to sales (to your CRM, to Slack) and syncs them as a retargeting third-party audience directly to LinkedIn Campaign Manager for you!

Top signals across all accounts

• 🟢 Best broad signal: 200+ impressions + 10+ engagements + 3+ clicks. 4.0x lift, phi 0.023.
• 🟢 Best stage: ABM stage Interested. 3.3x lift, phi 0.024.
• 🟢 Best single metric: 10+ engagements (3.3x) and 10+ clicks (3.7x).
• 🟢 Best format combo: Image click + Document click. 10.6x lift.
• 🟢 Best thought leader ad combo: TLA engagement + Document click. 11.8x lift.

Top signals for ZenABM

• 🟢 5+ engagements: 12.4x lift, phi 0.040, 15% of our deals. Our best signal.
• 🟢 3+ clicks: 10.8x lift, phi 0.032.
• 🟢 1+ engagement: 7.3x lift, 17% of our deals. Our broadest signal that still works.
• 🟢 Image and thought leader ads: the only reach that helped us, about 3x.
• 🟢 10 impressions: no lift (1.2x, could be luck). Half our deals, zero signal.

The bottom line: influenced pipeline is not a vanity metric.

But lazy threshold setting is not very smart indeed 😝

Set it on engagement, test it on your own data (or let the Influence Coefficient do it for you every Friday), and use the winning signals to tell sales who to call next (or feed your agents the right accounts to spam next 🤖).

So, as per Jack Cleere's post: cheers to influenced pipeline again 😉