Understory ranks No. 140 on Inc. 5000 list of fastest-growing companies in U.S. Read more

Understory Logo
+1 415 903 7062

The playbooks behind our results.

Every case study on this site started as one of these. Here's what we did, step by step, in the order we did it. Take them and run them yourself, or have us run them for you.

Build outbound from zero

FullPACGTM Engineering

FullPAC lived off Meta ads and had no outbound set up. As it moved to selling a white-label platform to political consultants, it needed a direct line to a buyer who is hard to reach.

About 365 positive replies, 72 to 75 meetings in the first three months, closed deals that already cover the whole investment, and what FullPAC's CMO calls potentially eight figures in pipeline.

Watch the FullPAC case study
  1. 01Run a half-automated, half-human intake on positioning, ICP and what makes the product different.
  2. 02Stand up the sending infrastructure in month one while the intake runs.
  3. 03Build every audience and write every script, then hand over a finished first version for feedback.
  4. 04Run cold email and LinkedIn campaigns together, on top of the client's HubSpot.
  5. 05Use what the campaigns surface to close gaps in HubSpot and RevOps as they appear.

Fix the signal, then scale

DoppelPaid Media

Only 36% of Doppel's paid form submissions became qualified leads, and the conversion data feeding Google Ads had no CRM lifecycle import behind it. The platform was optimizing toward an event sales could not act on.

Google Ads produced $3,652,185 in influenced pipeline on $504,169 of spend, a blended 7.2x, with cost per form down 20% quarter over quarter.

Read the Doppel case study
  1. 01Trace paid form submissions into the CRM and move the conversion goal to CRM-qualified leads.
  2. 02Consolidate 16 campaigns into 11 and rename each one to the client's own product taxonomy across all 4 product lines.
  3. 03Cut competitor allocation from 20.7% of Q1 spend to 4.8% of Q2 spend and redeploy it into campaigns producing at $941.
  4. 04Scale the core campaign 7.6x in one quarter, from $15.8K to $120.5K, and hold cost per acquisition at $941.
  5. 05Open the UK once the US is at scale, where cost per lead ran $581 against $941 for the US.

Audit the defaults first

NorthbeamPaid Media · GTM Engineering

Northbeam's LinkedIn account had 74% of 30 days of budget in the Audience Network and no conversion tracking, so campaigns optimized toward website visits.

LinkedIn produced 217 leads, 84 of them qualified, and 37 deals, with $2.3M in pipeline reported by Northbeam, while Google cost per opportunity fell 35% on 44% less spend.

Read the Northbeam case study
  1. 01Switch off the LinkedIn Audience Network first, which held 74% of the last 30 days of budget.
  2. 02Replace max-delivery bidding with deliberate bids and rebuild conversion tracking so the platform and the CRM agree on what a lead is.
  3. 03Reconcile platform leads against CRM records row by row, which exposed 42 leads in the ad platform against 4 in the CRM.
  4. 04Cut Google spend 44% on purpose, adding a negative keyword system, structuring both non-brand campaigns and pulling budget off brand terms.
  5. 05Run executives' own posts as thought-leader ads, then point Google, LinkedIn, Meta and outbound at 1 enriched Clay list.

Audit first, then audit again

RemoFirstPaid Media · GTM Engineering

RemoFirst's top keywords were blocked by its own negative-keyword lists while $17,500 a month went to branded search, and no experiment had ever been run in the account.

In Q1 2026 the account produced 1,040 opportunities on $899,845 of spend, with brand at $275 each and non-brand at $1,437, 15 to 32 percent below its $1,700 to $2,100 starting range.

Read the RemoFirst case study
  1. 01Audit the account first and walk the client through what is in it before selling anything.
  2. 02Rebuild the negative-keyword system and the account architecture around which searches come from people who become customers.
  3. 03Give the US, the EU, the UK, India, and Australia and New Zealand their own non-brand campaigns, alongside competitor conquesting.
  4. 04Audit again at 120 days and pause the 2 campaigns carrying $13,000 of spend with no opportunities, recovering $4,000 a month.
  5. 05Put a number and a date on the record, with non-brand cost per opportunity down 30% in the next 60 days.

Find the zero-conversion campaigns

HiverPaid Media

Hiver put $20K to $25K a month into Google Ads for mixed results, spread across 24 Performance Max campaigns. The account average still looked survivable, which hid the campaigns producing nothing.

Excluding the 4 campaigns with zero conversions, non-brand ran at $739 per conversion, and cost per SQL fell 46% month over month.

Read the Hiver case study
  1. 01Split brand from non-brand and report a cost per conversion for each, so brand volume stops flattering the average.
  2. 02Sort non-brand campaigns by conversion count and recalculate cost per conversion without the 4 campaigns sitting at zero.
  3. 03Rebuild the account from 19 campaigns into 29 so each intent reports separately.
  4. 04Consolidate 18 legacy negative lists into 4 shared lists and catch the 160 negatives blocking live keywords.
  5. 05Move the conversion goal from MQL to SQL and upload qualified SQLs back from the CRM as an offline conversion action.

Build the list nobody sells

PatientNowGTM Engineering

The standard B2B databases do not cover medical aesthetics well, and the client estimated its US market at around 12,000 med spas, so every contact was finite and the list had to be built from scratch.

The program sent 242,750 emails at a 0.8% lifetime bounce rate and produced 2,778 replies and 270 interested practices over 6 months.

Read the PatientNow case study
  1. 01Build a Google Maps lead engine across 5 verticals and more than 50 metros, then enrich in Clay with a tech-stack waterfall.
  2. 02Build 210 branded mailboxes on the client's own domains and run 4 angles across 2 domain pools, refreshed on a cycle.
  3. 03Purge the list to valid-only addresses, taking bounces from 3.6% in June to 1.3% by late July.
  4. 04Segment campaigns by whether the recipient is on Google, which replied at 2.7% and 3.4% against 1% to 1.5% otherwise.
  5. 05Build an AI reply and auto-responder system in the client's Clay instance, with 4-way intent classification routing replies into Salesforce.

Interview once, publish monthly

OperatusLinkedIn Content

Operatus's CEO believed in LinkedIn but posted in spurts, and an algorithm that rewards frequency treated every spurt as a standing start.

Engagement rose 600% on the prior 90 days, most posts landed 5,000 to 10,000 impressions and one reached 50,000, all from 90 minutes of CEO time a month.

Read the Operatus case study
  1. 01Open with a comprehensive onboarding form at signup, when the client is most willing to hand over everything they know.
  2. 02Propose topics and iterate them with the CEO against 7 content pillars that define the territory the profile owns.
  3. 03Run one 90-minute interview a month against a prepared question set, with 2 or 3 follow-up probes per topic.
  4. 04Turn that single call into 10 to 20 posts and send them back every week for edits and approval.
  5. 05Lean on predictions and hot takes, which cleared 60 to 80 likes at minimum at roughly a 100% hit rate.

Enrich before you write

ExpensifyGTM Engineering

Expensify grew for 8 years on product-led signups with almost no outbound, so buyers who never search the category went unreached. Around 70% of leads never engaged sales.

Replies ran at 1.5% where the accounting package was known against 0.14% where it was not, and the program produced 180 to 190 positive replies from 125,000 emails.

Read the Expensify case study
  1. 01Make the prospect's accounting package a known field before every send, so the email has something true and specific to say.
  2. 02Turn the same signal into a 100-point lead scoring model and score 30,000+ leads into the CRM.
  3. 03Get 50 subdomains of Expensify's own domain approved, then stand up 141 inboxes across 47 domains, each warmed for 14 days.
  4. 04Analyze the existing signup base to isolate the technographic, firmographic and geographic patterns that separate converters from users who vanish.
  5. 05Drop the data vendor whose 40,000 contacts yielded roughly 500 after the ICP filter, a 1.25% yield.

Mine a competitor's audience

hapilyGTM Engineering

hapily had proved outbound could work, but with no full-time owner the campaigns started and stopped, so no run of experiments ever compounded.

The flagship campaign ran at 57% connection acceptance and a 30% message reply rate, and the program had created 39 deals, 5 closed won, by mid-May 2026.

Read the hapily case study
  1. 01Pull everyone who engages with 2 direct competitors' LinkedIn posts on 6 event topics, using Apify on an evergreen schedule.
  2. 02Dedupe in Clay, sort each person into 3 personas with an AI prompt, and filter out roles that will never buy.
  3. 03Check at account level that the company runs events and uses HubSpot before anyone gets a message.
  4. 04Run 3 HeyReach campaigns, 1 per persona, each with a note-free connection request, a persona-specific question and 1 follow-up.
  5. 05Never reference the post the person engaged with in any message.

Source every segment separately

WareSpaceGTM Engineering

WareSpace's buyers are small operators with little or no web presence, so standard tools like Apollo cannot find them, and 2 previous attempts at the channel stalled on list building.

The first 2 active months produced roughly 300 positive replies from 2 of 7 ICP categories, including 50 in 10 days from a 7,200-contact send.

Read the WareSpace case study
  1. 01Set the standard databases aside and source each ICP segment from where it leaves a trace, such as Google Maps for contractors.
  2. 02Build geographic lists on 20-mile radii around each facility, using roughly 134 Google Maps search terms per market.
  3. 03Verify website and phone validity in a Clay lead-scoring waterfall before any record enters a sequence.
  4. 04Scale sending across 4 locations at 34 inboxes each, taking daily volume from 1,200 to over 3,000.
  5. 05Add an AI drafting assistant and a HubSpot sync with instant alerts once reply volume outgrows the team.

Narrow to the buying titles

HyperboundPaid Media

Hyperbound had no paid account, no spend history and no benchmark for what a click should cost, so every early decision had to run on category knowledge.

One targeting change took cost per click down 35%, from $17.66 to $11.48, and 50 of 282 influenced accounts, 18%, booked a meeting.

Read the Hyperbound case study
  1. 01Weight a young category's spend to the top of the funnel, with 42% top, 30% bottom and 28% middle across 90 days.
  2. 02Write CTR targets before buying any clicks, with 3% on LinkedIn promoted posts and a 3% minimum on Google Ads.
  3. 03Tighten the audience to VP, Director and CXO titles and cut $4,500 of off-ICP spend.
  4. 04Promote the founder's own posts as thought-leader ads and track CTR, engagement clicks and dwell time.
  5. 05Move retargeting to video, which ran at $12 per acquisition against $55 to $75 for static demo-request ads.

Promote the founder's best posts

RB2BPaid Media

RB2B's founder had LinkedIn posts that performed well organically, but no paid media of any kind, no audience data and no cost benchmark to start from.

Video creative ran at roughly 20% click-through and blended cost per free signup came in at roughly $22, against the $50 to $100 category range.

Read the RB2B case study
  1. 01Collect the founder's 25 best organic posts and use them as the entire creative inventory, with no new creative.
  2. 02Segment the posts into 4 personas and propose 3 to 9 test audiences for each.
  3. 03Edit a call to action into every promoted post so the ad has somewhere to send the reader and something to price.
  4. 04Run every post against every plausible audience at small budgets, then move budget behind the winning post-audience pairs.
  5. 05Agree the scaling rule before results arrive, which here meant doubling ad spend each month for as long as it kept making sense.

Take the brand back first

KnitPaid Media

A marketing team of 2 checked Knit's ad account about every 2 weeks, so platform defaults took over and paid media produced 7 leads and 1 closed deal in a half-year.

Brand impression share went from 18% to 75%, LinkedIn demo requests came in at roughly $180, and the competitor campaign produced 1 opportunity worth roughly $45,000 with a verbal commitment.

Read the Knit case study
  1. 01Move every keyword to phrase and exact match and set target impression share bidding on the brand campaign.
  2. 02Turn off Search Partners and add 3 negative keyword lists to keep category traffic out.
  3. 03Cut conversion tracking to a single primary conversion action, connect the real CRM and import lead-stage data for value-based bidding.
  4. 04Add dynamic search ads for discovery and competitor campaigns for buyers comparing named vendors.
  5. 05Rebuild LinkedIn with Audience Network off, manual bidding, rebuilt retargeting and an ungated AI research report at the top.

Stop the bleed, then map

VicariusPaid Media

LinkedIn's Audience Network was sending Vicarius traffic measured at over 99% bounce with zero time on page, and one person covered both demand generation and the marketing team.

Between March and September 2025, $100,289 on LinkedIn produced 65 direct leads and 17 website demo requests, and $34,275 on Google produced 40 leads and 10 opportunities.

Read the Vicarius case study
  1. 01Switch off LinkedIn's Audience Network expansion, which ended the spam leads and wasted spend immediately.
  2. 02Rebuild conversion tracking and the forms first, because every later decision reads from them.
  3. 03Ask the client for closed-won accounts by region and build lookalike audiences for each market from them.
  4. 04Write each market's campaigns in its own language and context, and segment every campaign tightly for hit rate.
  5. 05Test 151 ads and read results by region, since lead volume and deal value land in different countries.

Or let us run one for you.

Tell us where pipeline is stuck, and we'll tell you which playbook fits and what it would take.