CRM Setup and Pipeline Automation for a B2B Consulting Firm

The Client

A Lahore-based supply-chain and procurement advisory founded in 2016, with 32 staff: nine partners and directors, eighteen consultants, and a five-person research and operations team. The firm runs procurement cost-reduction programmes, warehouse and distribution-network design, and supplier diagnostics for mid-market Pakistani manufacturers, importers, and FMCG distributors. A typical engagement runs PKR 1.5 million to PKR 6 million over three to five months, which places every enquiry firmly in considered-purchase territory: multiple stakeholders, a finance sign-off, and a buyer who shortlists two or three firms before anyone commits.

Growth had always been referral-led, and referrals remained healthy. But by early 2026 the founding partners wanted marketing to carry more of the load, and the outward machinery for that already existed: a website with an audit-request form, an annual procurement benchmark report co-authored with a university department, active LinkedIn profiles for two partners, and a presence at chamber-of-commerce events and one annual supply-chain summit. Together these produced roughly 120 raw enquiries a month against a combined demand-generation budget of PKR 740,000.

The machinery to receive those enquiries, however, was three partner-maintained spreadsheets plus whatever lived in individual Gmail inboxes. There was no shared pipeline, no shared definition of a qualified lead, and no honest answer to the founding partner’s standing question: of the 360-odd enquiries from last quarter, which became proposals, and what happened to the rest?

This engagement is an illustrative composite — a representative profile assembled from patterns we see across Pakistani professional-services firms, with figures kept inside realistic ranges so your team can sanity-check fit against its own baseline. Consulting firms usually reach this work through the digital marketing for business consultants track, where CRM readiness is assessed alongside channel performance.

The Problem

A two-week diagnostic at the start of the engagement measured the leak precisely rather than assuming it.

• 38% of enquiries were logged nowhere. Cross-checking form-export data, inbox threads, and all three spreadsheets for a two-week window showed more than a third of enquiries never entered any system. They existed only in whichever inbox received them, and when that consultant was on an engagement site, the enquiry effectively did not exist.
• Median first response: 14 business hours. An enquiry arriving Thursday afternoon was typically answered Monday. 22% of enquiries waited three business days or longer, and the firm could trace two lost RFP shortlist invitations in late 2025 to slow first responses — Pakistani procurement teams compare notes about advisers just as advisers compare notes about clients.
• Lead acceptance rate of 22%. Marketing handed over everything with a pulse; partners accepted roughly one in five as worth a proper conversation. The other four in five were silently shelved in a “later” folder that had no exit, and marketing had no visibility into why.
• 240 duplicate company records. Three trackers keyed by whoever typed the entry meant one Faisalabad textile group existed as three separate prospects, and in March 2026 two partners independently pursued the same Sialkot food manufacturer — the prospect noticed.
• Forecast variance of ±40%. The quarterly revenue forecast was assembled on a Sunday night from partner memory. Actual proposal outcomes bore little relation to it, which made consultant staffing and cash planning guesswork.
• Rising cost per accepted lead. At PKR 28,500 per sales-accepted lead on flat spend, each quarter bought less than the last — not because demand was drying up, but because the conversion of enquiry into accepted conversation was decaying.

The founding partner’s summary at kickoff set the frame for everything that followed: “We are not short of demand. We are short of a system that treats demand as an asset instead of an inbox.”

Phase 1 — Diagnosis and pipeline cleanup (Weeks 1-2)

Phase 1 produced two artefacts: an honest baseline, and a single de-duplicated dataset worth automating.

The baseline came from structured interviews with all nine partners, two weeks of live enquiry shadowing, and the cross-system reconciliation behind the numbers above. It was written on one page and pinned in the partners’ meeting room, because a baseline nobody can quote is a baseline nobody improves against.

Measure	Baseline (Week 0)
Enquiries logged in any system	62%
Median first response	14 business hours
Enquiries unanswered after 3 business days	22%
Duplicate company records across trackers	240
Stale unworked threads older than 60 days	61
Lead acceptance rate	22%
Quarterly forecast variance	±40%

The cleanup consolidated 2,400 raw contact rows from three spreadsheets and four inbox exports. Pakistani phone data being what it is, the same mobile number appeared as 0300-XXXXXXX, +92-300-XXXXXXX, and 0300 XXXXXXX depending on who typed it, so normalization ran before any dedupe logic. Companies were then keyed by email domain rather than by typed name, which caught most of the 240 duplicates and exposed the 61 stale threads — enquiries from as far back as a March event still unworked in June. Those were re-queued into a structured catch-up sequence rather than quietly buried: an acknowledgement referencing the original enquiry, a single question about current status, and a clear exit to the monthly newsletter for anyone who declined. The sequence recovered eleven active conversations and three discovery calls from enquiries the firm had effectively already paid for — a quiet early win that bought the wider programme credibility with the partners before the scoring model even went live.

The second artefact was a definitions workshop: two hours, all nine partners in the room, one page per lifecycle stage — Lead, MQL, SAL, Opportunity, Proposal, Closed — with explicit entry and exit criteria each partner signed. The single most consequential line in the whole document: an enquiry becomes a sales-accepted lead only when a partner accepts it against the written rubric, not when marketing scores it. That sentence later carried the +46% headline, because it made “accepted” a measurable event instead of a mood.

Phase 2 — CRM build and scoring model (Weeks 3-5)

The implementation followed our CRM setup and pipeline automation scope: platform configuration, lifecycle build, scoring, routing, and reporting in one sequence rather than as disconnected projects. HubSpot Professional carried the build, chosen over Zoho CRM for partner-facing mobile logging and the reporting layer rather than any capability gap — Zoho runs the identical architecture at a lower licence cost and remains the rational pick for cost-sensitive teams. The architecture, not the brand, is what moved the numbers.

Configuration discipline mattered more than feature depth. Six mandatory fields only — company, contact, role, enquiry type, revenue band, source — with everything else optional, because CRM adoption in a partner-led firm dies of required fields. The website enquiry form was rebuilt in the same spirit, cut from eleven fields to six, with revenue band and company size added as dropdowns; form completion rose from 61% to 76%, and the new fields fed the scoring model directly.

The scoring design leaned on lead scoring and sales handoff optimization patterns, weighted with the firm’s own history. We pulled 118 projects going back to 2019 and modelled which enquiry attributes predicted a won engagement: manufacturers with 200–2,000 employees had converted at 2.4x the rate of smaller firms; first contacts holding procurement or supply-chain titles converted at 1.7x the rate of generic management titles; food and textiles verticals carried the strongest realization rates. Those ratios became the fit weights — the model’s credibility with partners came entirely from the fact that no number in it was imported from someone else’s benchmark.

The model has two dimensions. Fit score (0–55): vertical, company size band, contact seniority, and declared budget cycle, weighted per the won-deal analysis. Engagement score (0–45): pricing or services page visit +15, audit-form submission +20, discovery call booked +25, benchmark-report download +5 each, reply to any email +10, repeat site visit within seven days +5.

Band	Score	System action
Hot	70–100	Auto-assign to partner by vertical map; 4-business-hour first-touch SLA starts
Nurture	40–69	Enter five-touch nurture track; re-scored after every interaction
General	0–39	Monthly insights newsletter pool with a quarterly re-engagement sweep

Routing assigned hot leads round-robin within a vertical expertise map — two partners per vertical, with absence and conflict rules so the same prospect never landed on two desks again.

Phase 3 — Routing, SLAs and nurture (Weeks 4-8)

Phase 3 made speed and persistence structural rather than aspirational.

The SLA mechanics were deliberately blunt. A first-touch timer of four business hours starts the moment a hot lead is assigned; at eight hours unassigned-or-untouched, the lead escalates to the managing partner’s queue. The business-hours calendar was set to Monday through Saturday 1pm — this firm works Saturday mornings, and a CRM calibrated to a five-day Western week would misfire every weekend. Three first-response templates were written per source type (referral, audit form, event), each framed as a starting draft partners were expected to personalize; template-first responses kept the 95-minute median response possible without letting correspondence go out sounding generic.

The middle band got a nurture track instead of being forced into premature sales calls: five emails over three weeks built from assets the firm already owned — the procurement cost benchmark, a rupee-devaluation sourcing brief, an energy-tariff note on warehouse operating costs, and a summarized distributor-network redesign case. Every reply and click re-scored the contact automatically, so nurture graduates re-entered routing on their behaviour, not on a marketer’s memory. Of the 38 sales-accepted leads counted at day 90, nine had arrived through nurture re-entry rather than direct hot assignment — a quarter of the headline number from leads the old system would simply have lost.

Adoption was managed as seriously as configuration. A weekly 30-minute pipeline stand-up replaced the Sunday-night memory forecast, and a simple partner leaderboard — logging rate, SLA compliance, stage movement — did more for data hygiene than any policy memo. Two partners initially kept private side-lists; the managing partner’s rule that anything absent from the pipeline review was absent from the forecast closed that practice within three weeks. Training ran as three 45-minute working sessions inside live pipeline reviews rather than as a separate workshop nobody attends — partners learned the system by using it on their own week’s leads, with the consultant in the room for the first two sessions. By week 8 the acceptance rate had moved from 22% to 29%, and the partners were quoting stage names in corridor conversation — the quiet signal that a pipeline model has become the firm’s operating language.

Phase 4 — Recalibration and forecasting (Weeks 8-12)

A scoring model is a hypothesis, so phase 4 tested it against outcomes and corrected it.

Eleven weeks of live data justified three recalibrations. Budget-authority signals were reweighted upward by 40% — enquiries mentioning board approval or a budget cycle had won at a markedly higher rate than their previous weighting implied. Content-download-only behaviour was reweighted down; report downloads alone had over-predicted readiness, and several high-download, zero-meeting leads had been cluttering the hot queue. And negative scoring was added for internship and careers enquiries, which an enquiry-type field on the rebuilt form now caught at entry — a small Pakistani-firm reality that imported scoring templates never account for.

Forecasting changed shape entirely. The weighted pipeline dashboard used each partner’s actual stage-conversion rates over the prior two years, so the forecast updated the moment a stage changed rather than the night before the partners’ meeting. Across weeks 9 to 12, the week-four-ahead forecast ran within ±12% variance against the ±40% baseline quarter. A renewal motion also went live: fourteen past clients entered a re-engagement sequence, and two booked follow-on diagnostics within the quarter. That number is modest, but it was a pipeline category the firm had never worked systematically — repeat business had previously depended on a former client remembering the firm at the right moment.

Final Results

Measured at day 90 against the Week 0 baseline, on demand-generation spend held flat at PKR 740,000 a month:

Metric	Baseline	Day 90
Sales-accepted leads per month	26	38 (+46%)
Lead acceptance rate	22%	32%
Cost per sales-accepted lead	PKR 28,500	PKR 19,500 (−32%)
Median first response	14 business hours	95 minutes
Enquiries logged in the CRM	62%	100%
Discovery calls booked per month	19	25 (+31%)
First call to proposal accepted	47 days	39 days
Week-4 forecast variance	±40%	±12%

Every number traces to a mechanism described above: the acceptance lift to scoring plus SLA routing plus recalibration, the cost reduction to flat spend against a higher accepted-lead count, the response-time collapse to SLA automation, and the forecast accuracy to a live dashboard fed by clean stage data. The firm’s new constraint is a healthier one — 38 sales-accepted leads a month against partner capacity for roughly 25 discovery calls means the next investment is associate-level sales support, not more lead volume.

What Made This Work

1. Partners wrote the definitions, so partners worked the system. The lifecycle workshop produced a one-page document the partners had personally signed. Nothing in the CRM contradicted how they already sold, which removed the usual quiet resistance that kills enterprise CRM rollouts.
2. Scoring was weighted from the firm’s own 118-project history. When the hot queue said a 900-employee Faisalabad textile manufacturer mattered more than a 40-person trading company, partners recognized their own experience in the ranking — so they trusted the queue enough to actually work it.
3. SLAs made speed the default. Response time improved not because anyone was told to try harder but because assignment started a timer with escalation attached. The 95-minute median required zero behavioral exhortation.
4. The middle band was nurtured, not discarded. Nine of the 38 day-90 sales-accepted leads were nurture re-entries — enquiries the previous system would have lost to the “later” folder, recovered purely because a track existed for not-yet-ready.
5. The forecast ran on live pipeline data with partner-level conversion rates. Once forecast inclusion required pipeline presence, logging stopped being optional in practice, and the ±12% variance followed from honest stage placement rather than optimistic memory.

What Teams Can Apply

1. Define “sales-accepted” in writing before buying any software. The signed one-page definitions doc cost two hours and did more for measurability than the platform choice. If partners cannot agree what a qualified lead is, no CRM will resolve the disagreement — it will merely automate it.
2. Find where your leads actually live before configuring anything. In this firm, 38% of enquiries sat in inboxes no system could see. A pipeline built on the 62% you can see will quietly misshape every downstream metric, including your cost per lead.
3. Weight fit scores from your own won deals. Two years of deal history is the practical minimum, and it is enough. Imported benchmarks from another market will rank your leads plausibly and wrongly, and partners will sense the wrongness before the dashboard shows it.
4. Automate the SLA, not just the routing. Auto-assignment alone delivers the lead to a busy desk where it waits. The timer plus escalation is what converts routing into response, and response was worth 22 percentage points of acceptance rate here.
5. Re-check scoring weights quarterly. Deal mix shifts, and a model calibrated once drifts into noise. This model’s first recalibration moved budget-authority weighting 40% and demoted download-only behaviour — both corrections visible only once real outcomes existed to test against.