A Customer Data Platform That Grew B2B Pipeline 31%

The Company

Cobalt Ridge Software is a B2B SaaS vendor selling an operations platform to enterprise customers on annual contracts, with a land-and-expand motion that depends on widening product usage inside each account. The commercial engine spans an inside sales team, field account executives, a demand generation function, customer success and revenue operations. Growth had been strong enough to attract a Series C, but the company's economics were drifting: customer acquisition cost was climbing 19% year over year while win rates sat flat, and the board was beginning to ask the question every SaaS leader dreads — where exactly is the marketing spend going?

The uncomfortable answer was that nobody could say, because the commercial data lived in six disconnected systems: Salesforce for CRM, Marketo for marketing automation, product telemetry from the application itself, a subscription billing platform, a support desk, and a data warehouse that had become the dumping ground of last resort. Each system held a fragment of the customer. No system held the customer.

Where Things Stood

Sales reps worked from firmographic fields — industry, employee count, contract value — and were effectively blind to what was happening inside the product they were selling. Whether an account had doubled its weekly active users, plateaued, or hit an API wall never reached the CRM, so expansion conversations started when a customer raised them, not when the signals appeared. Marketing routed leads on MQL rules written two years earlier: score thresholds from form fills and email clicks that bore little relationship to purchase readiness. Trial accounts with heavy usage sat untouched for days because their form-fill scores looked unremarkable, while lightly used accounts with glossy demographics got the calls.

Duplicate records made everything worse. Because contacts, accounts and product users were stitched together by hand across systems, roughly 11% of records were duplicates, and two reps occasionally worked the same account in the same week — an embarrassment in front of an enterprise buyer that also silently inflated CAC. Revenue operations spent its weeks reconciling versions of the funnel: the marketing automation tool said one thing, Salesforce said another, and the warehouse said a third. The board deck, assembled manually each quarter, could not be reconciled to the CRM at all. Gartner's research that B2B buyers now spend only about 17% of their buying journey in direct contact with suppliers made the stakes concrete — if the buying signal is usage and nobody is watching usage, the seller is absent for most of the decision.

The Challenge

The challenge was structural: six systems had produced six partial truths, and every downstream process — routing, scoring, forecasting, attribution — inherited the fragmentation. Point fixes had been tried. A consultant had once rebuilt the lead-scoring rules, and an integration tool had been wired between two of the systems, but the underlying identity problem remained: the same customer existed as different entities in different tools, so any score or report built on top of the fragments was wrong in ways nobody could quantify. Leadership did not need a new dashboard; it needed one account-level truth that every system could read from and write to.

There was an organizational dimension too. Marketing owned Marketo, sales owned Salesforce, product owned telemetry, and each team had optimized its own definitions of "qualified" and "ready." The MQL-to-SQL handoff, where two teams' definitions met, was where pipeline quietly died. Any solution had to define lifecycle stages once, centrally, and make the agreed definitions compute automatically — because a definition that lives in a slide deck gets renegotiated every quarter, and McKinsey's B2B Pulse surveys report that roughly two-thirds of business buyers now prefer remote or digital self-serve interactions, which means the handoff moments matter more than ever.

Discovery and Diagnosis

Emerson Technology began with a two-week go-to-market systems audit. Every tool that touched a customer was mapped with the data it held in isolation: Salesforce contacts and opportunities, Marketo activities and campaign membership, telemetry events with feature-level granularity, billing line items, support tickets with severity and sentiment, and three years of warehouse tables of uncertain provenance. The audit quantified the fragmentation — 11% duplicate records, no shared account identifier across product and CRM, and lead routing that depended on a nightly batch job with a failure history nobody monitored.

The diagnosis went deeper than systems. Interviewing reps surfaced the real workflow: when reps wanted to know what an account was doing in the product, they messaged a solutions engineer and waited. Interviewing marketers surfaced the routing problem: hot trials were indistinguishable from cold ones at the moment of assignment because scores were computed from form fills only. A funnel decomposition showed conversion losses concentrated at two points — initial speed-to-lead and the MQL-to-SQL transition — which meant the biggest wins would come from faster routing and shared definitions, not from more lead volume. Forrester's finding that companies which mature their lead-management processes generate more sales-ready leads at substantially lower cost per lead gave the program its economic frame: fix the mechanics, and the same spend produces more pipeline.

The Solution

Emerson Technology deployed a composable customer data platform with Snowflake as the warehouse of record, and deliberately kept the architecture boring where it counted. The design has three layers, each solving one of the diagnosed failures.

Identity resolution first

Before any scoring or segmentation, a Python identity-resolution service stitched contacts, accounts and product users into one graph, using deterministic matching on email domains, account names and billing identifiers, plus fuzzy matching on the messy leftovers — free-mail signups, acquired companies' lists, human-typed account names. Duplicate rate fell from 11% to under 1%, a 94% reduction, and every merged record keeps its history so nothing is lost. This was unglamorous work, and it was sequenced first on purpose: analytics built on duplicated accounts would have poisoned every downstream number and destroyed rep trust in week one.

Account-level scoring and one set of definitions

With identity settled, dbt models compute health scores, lifecycle stages and account-level readiness once, centrally, on a nightly cadence. Scoring models combine product telemetry — feature adoption, seat utilization, API volume — with firmographic and intent data to score both expansion readiness in existing accounts and purchase readiness in active trials. MQL, SQL and health are no longer negotiated per team; they are computed in one governed model that marketing, sales and finance all consume. The models are versioned and their inputs documented, so when a score changes, anyone can see why.

Reverse ETL and minutes-not-hours routing

Crucially, the insight does not stay in the warehouse. Reverse-ETL syncs unified profiles, scores and segments back into Salesforce, Marketo and the support desk, so the systems where people actually work display account-level truth. Segment-based triggers watch for readiness signals — a trial crossing an usage threshold, an existing account's seat utilization climbing — and fire Slack alerts that route the account to the right rep within minutes, with context attached. Power BI gives revenue operations one funnel view from first touch to renewal, and for the first time the board deck, the CRM and the marketing platform agree. The funnel is no longer three reconciled opinions; it is one record.

The First 90 Days

Weeks one and two covered the systems audit and the funnel decomposition, plus a reconciliation exercise that put a number on the chaos: for a sample of 200 accounts, the team traced every duplicate, every mismatched field and every disagreed stage assignment. Weeks three through six delivered the identity-resolution service in stages — deterministic matches first, fuzzy matches reviewed by revenue operations — with a weekly duplicate review that shrank the exception list from hundreds to a handful. Reps saw duplicates disappear from their own account lists before anything else changed, which bought more goodwill than any launch email could have.

Weeks seven through ten deployed the dbt scoring models and the reverse-ETL syncs into Salesforce and Marketo. The rollout followed a pilot-within-a-rollout: one mid-market segment received usage-based routing two weeks before the rest, so the team could compare speed-to-lead and conversion against the control group. The pilot's early numbers — response times dropping from hours to minutes — were shared with the whole sales organization at the weekly standup, which did more for adoption than training sessions. Slack alert thresholds were tuned in a daily loop with the sales managers: too sensitive and reps learned to ignore alerts, too conservative and signals decayed before anyone acted.

Weeks eleven and twelve extended routing to all segments, retired the legacy nightly batch job, and moved lifecycle definitions into the governed dbt models with marketing and sales jointly signing the definitions. A simple rule governed the transition: when a rep disagreed with an account's score by a wide margin, the case was logged and reviewed weekly, and the models were adjusted or the rep was shown the usage data behind it. By the end of the quarter, lead routing was measured in minutes, the duplicate log was nearly empty, and the funnel had a single owner with a single definition.

The Results

Within two quarters, qualified pipeline grew 31% without additional marketing spend — driven mainly by faster routing and better expansion targeting rather than more leads. Lead response time fell from roughly four hours to five minutes, which matters because usage signals decay fast and McKinsey's buyer-preference data shows the first conversation increasingly happens digitally, on the buyer's schedule. Duplicate records dropped 94%, ending the phenomenon of two reps working the same account. MQL-to-SQL conversion improved 18% as reps began working usage signals instead of stale form fills. And revenue operations finally presents one funnel from first touch to renewal — the board deck agrees with the CRM for the first time in the company's history.

The qualitative changes ran deeper than the metrics. Account executives open their pipeline reviews with usage context instead of asking for it. Expansion plays fire off telemetry rather than renewal dates, so customer success and sales collaborate earlier in the account lifecycle. CAC pressure eased without a dollar of additional spend, because the funnel mechanics — routing speed, identity quality, shared definitions — stopped leaking the pipeline the marketing budget had already paid to create.

What Your Business Should Do

The mechanics behind Cobalt Ridge's 31% are portable to almost any B2B company with a product. Start by inventorying your go-to-market systems and the gaps between them — map every tool that touches a customer and what data it holds in isolation, because the gaps are where your pipeline quietly leaks today. Then unify on the account, not the lead: B2B buying decisions happen at the account level, so stitch contacts, product users and billing records to one account identity before you build any scoring. Fix identity resolution before analytics, however tedious it feels — duplicates poison every downstream metric and erode rep trust fast, and deterministic plus fuzzy matching done early decides whether anything else works.

Score accounts on product usage rather than form fills: feature adoption, seat utilization and API volume predict readiness better than a whitepaper download ever will, and those signals belong in the CRM where reps actually live. Use reverse ETL rather than dashboards — warehouse insights that stay in the warehouse change nothing, so sync scores and segments back into Salesforce, your marketing platform and Slack, where daily action happens. Route hot accounts in minutes, not next week's call list; usage spikes decay fast and speed is the entire advantage. Define lifecycle stages once, centrally, in a governed model rather than per-team spreadsheets, because disputed MQL and SQL definitions quietly destroy more pipeline than bad models do. And track pipeline per segment monthly — created, converted and lost — while resisting the attribution-perfectionism trap; directional consistency month over month is what drives decisions.

Close by asking the questions this program forced Cobalt Ridge to answer honestly: if your best trial account doubled its usage today, how many days before a salesperson noticed; how many of your accounts exist more than once across your systems; could marketing and sales agree on the definition of a qualified account right now. Forrester's lead-management research and Gartner's buying-journey data both point the same direction — the vendors who see usage signals first and act fastest win the accounts everyone else is still scoring on demographics. Work through the eight actions above with your go-to-market team, and revisit the reflection questions in this practice book quarterly to keep the funnel honest.