91% of CRM records are incomplete & AI forecasting inherits the gaps

Operators and investors,

Before we start, I’m flying in for SuperReturn US West in LA on Sep 14 and 15. If you’re attending as well, hit “Reply” and let’s connect this week.

Meanwhile, let’s move to disjointed data and hallucinations in enterprise today.

The average B2B sales forecast misses by 25% to 40%. Top-performing organizations run at 80% to 95% accuracy while the median team sits at 50% to 70% (2026 forecasting benchmarks).

Why forecasting is increasingly harder in 2026

Every portfolio company I see closing that gap fixes the data before it changes the model, and that order is being reversed across the mid-market right now. AI forecasting reaches 85% to 95% accuracy against 60% to 70% for traditional methods, and the qualifier attached to that finding is doing all the work: the lift holds only where data quality is strong. Buy the model on top of a pipeline nobody maintains and the output is a confident number with the same error inside it.

91% of CRM records carry incomplete fields, and 35% of sales professionals say they trust their CRM data.

This issue covers 4 topics:

• Why AI forecasting inherits the CRM gap

• What a 30% forecast miss costs a PE-backed company

• Where the first 20 points of accuracy come from

• What to fix before buying another forecasting tool

It is also written for operating partners and portfolio CEOs whose board pack carries a forecast the sponsor is underwriting against, drawing on the RevOps work for portfolio companies we run across PE-backed platforms.

1. Why AI forecasting inherits the CRM gap

A forecasting model weights deals on engagement signals, stage history, and close patterns. Each of those is a field somebody either filled in or skipped.

When close dates get pushed by a rep clearing their pipeline view, the model reads a deal that slipped for a commercial reason. When stage definitions differ between two acquired sales teams, the model reads one funnel where two exist. Neither error announces itself, and both survive the upgrade to an AI layer because the layer is reading the same table.

The claim I would make to any sponsor considering a forecasting purchase: the accuracy ceiling is set by the data, and the model determines how quickly you reach that ceiling. A company at 60% accuracy on incomplete records will not reach 85% by changing vendors.

2. What a 30% forecast miss costs a PE-backed company

A forecast miss is an internal disappointment in an owner-operated business, and in a PE-backed company it reprices the relationship with the sponsor and the lender.

A platform forecasting $60M and landing at $45M has missed by 25%, which is inside the average band. That miss lands on a covenant calculation, a management incentive plan, and the sponsor's own LP reporting in the same quarter. The second miss removes the CEO's ability to set the number, because the board starts applying its own haircut to whatever finance submits.

I have watched a $120M software platform lose 2 quarters of strategic discussion to forecast forensics. Every board meeting reopened the same question about why pipeline coverage looked healthy and conversion did not follow, and the answer sat in stage definitions nobody had reconciled after 2 add-on acquisitions.

Lenders evaluating cash generation in 2026 treat forecast volatility as execution risk, which can mean higher pricing or tighter credit terms.

3. Where the first 20 points of accuracy come from

Moving from 60% to 80% accuracy is a data exercise, and the next 10 points need process discipline layered on clean data.

The first 20 points come from improvements that don't require buying new software. Organizations applying basic pipeline hygiene see a 10% to 15% improvement in accuracy within 30 days, which is faster than any implementation timeline a vendor will quote.

Four fixes carry most of that movement:

• One definition of the sales funnel stages across every acquired team, with exit criteria written down and applied backwards to open pipeline

• Close dates that move only with a documented commercial reason, tracked as a slip rate by rep

• Required fields enforced at stage transition, so the record cannot advance while the data behind the forecast stays blank

• Won and lost reasons captured at close, which is the only input that lets a model learn from the last 12 months

That list takes a quarter of RevOps attention. It moves the number more than a platform migration does.

Each block sits on the one below it. A team buying the top block while the bottom is still missing pays for capability the data cannot support.

4. What to fix before buying another forecasting tool

A sponsor evaluating a forecasting purchase should ask the portfolio company to produce 3 months of forecast against actual, by rep, before any vendor conversation starts. Companies that cannot assemble that view are describing the problem the tool was supposed to solve.

The sequence I would run starts with CRM standardization across the portfolio: one stage definition spanning the platform and every add-on, field completion enforced at stage gates, slip rate measured for a quarter, then an evaluation of what a model adds on top of a clean record. The evaluation is cheaper and the answer is better, because the vendor demo runs against data that means something.

This week, pull the last 3 quarterly forecasts for your largest holding and compare each to actual by rep. Two questions to think about:

1. What was our forecast variance in each of the last 3 quarters, and does the error sit with a few reps or across the whole team.

2. Do our acquired businesses share one stage definition with written exit criteria, or does the pipeline report combine funnels that measure different things.

Score each answer red, yellow, or green. A red on the second question explains most of the variance in the first, and no forecasting tool resolves it.

Mario

🔍 Every deep dive pulls skeletons from the closet. Across 25-30 diligence audits and RevOps assessments a year, the pattern repeats: scarce documentation, dashboards with unexplainable KPIs, and processes that work in mysterious ways.

📊 Quota capacity is where most B2B growth plans break. A plan assuming 18 quota-carrying reps by month 18 turns into 4 hitting quota once ramp runs 7 to 11 months, a $4.3M miss against a $22M target the model never stress-tested.

🔌 Tech diligence checklists miss what breaks first after close. One portco had clean financials but a Salesforce instance three years behind and a Google Maps API already at 80% of quota, and integration stalled for nine weeks.

🛠️ Scoping the integration is not the same as rebuilding the engine. A 94-page consultant deck mapped 3 CRMs and 41 duplicate programs, yet six months on the portco missed revenue by 14% because nobody rebuilt the revenue engine.

🎙️ KKR's $17bn sale of USI, a business it bought at $4.3bn in 2017, was built on 90-plus strategic add-on acquisitions and no single transformational bet, compounding revenue near 12% a year over a nine-year hold. - by Nick Bradley

🎙️ Advent has built an IC AI robot, an observer trained on years of investment-committee memos that surfaces the questions the committee always asks, a concrete use of AI inside the fund. - by Hugh MacArthur

🎙️ The language of a "merger of equals" is a common source of value loss, because when both sides still believe they call the shots you build the conditions for post-close conflict. - by Kison Patel

🎙️ Six months into a leadership cohort for Rallyday Partners portfolio executives, the peer group drove more growth than the curriculum, which is the part most talent programs underbuild. - by Dan Cremons

🎙️ AI advisory demand across BluWave's network rose 193% year over year in Q2 2026, and almost none of it asks whether to use AI, since the questions are now about how to execute. - by Sean Mooney

🎙️ Manufacturers say they cross-train to remove key-person risk, but most stop at a box-check where someone shadows an expert for a shift and call it mitigated. - by John Stewart

🎙️ Calling a 13x EBITDA purchase "a good deal" misses that $25M to $100M deals trade at 8.8x while $500M to $1B deals trade at 13.4x, a gap that widened from 3.4 to 4.6 turns. - by Haktan Tuna Yilar

🎙️ Residential roofing has become one of the most active PE consolidation markets in home services, with 40-plus platforms and one deal every 48 hours at the mid-2025 peak. - by Andrew Allred

🎙️ Carlyle's tie-up with Oracle Red Bull Racing shows PE firms using sports to market their portfolio companies, opening the F1 ecosystem to portcos for customer acquisition and B2B relationships. - by Ashley France

🎙️ A meticulous 100-day plan went irrelevant by day 60 once the incoming CEO left and diligence assumptions aged, which is why the best sponsors move management diligence upstream. - by Sachin Bansal

A few kept from Thomas's roundup:

🎙️ Most founder-CEOs get replaced within 18 months when they never adapt to PE board dynamics and reporting cadence. - by Adam Coffey

🎙️ Alvarez & Marsal's latest read puts operational improvement at 47% of buyout value creation against 25% from leverage and deal structuring. - by Mike Trenouth

🎙️ Cash generation and DPI are becoming the scorecard LPs trust as exits stretch longer than planned. - by Eidji Braghin

🎙️ Poorly managed transition service agreements destroy real value during carve-outs, long after the deal closes. - by Chris Barber

Offensive AI is raising the security bar for every portfolio. OpenAI's forthcoming Astra model is described as very good at breaking into computer systems, which pushes breach risk, cyber-insurance cost, and security diligence up across leveraged portcos.

An AI-agent control layer is emerging. AIR raised $50M to help companies vet the skills and add-ons their AI agents use, a sign that portcos deploying agents now inherit third-party supply-chain risk that belongs in governance, not just IT.

Export barriers are reshuffling hardware supply chains. New US restrictions on foreign-made drones and robots collide with China's manufacturing scale, forcing industrial portcos to rethink vendor selection and sourcing before the next add-on.

AI value capture is a CEO mandate. A health-system piece argues CEOs, not the technology, drive AI transformation, the same lesson operators keep relearning: without owned accountability, AI spend stalls short of the value-creation plan.

RYNOSKIN TOTAL insect-protection apparel brand: 28-year-old brand selling a chemical-free, breathable undergarment with 98.9% protection against biting insects, throwing off roughly $337K a year in profit at a 6.8x multiple. $575,470

Federally-mandated ELD compliance SaaS for trucking: 4-year SaaS with $200K ARR, 300+ trucking clients, 80% margins, and a bundled TMS platform, with a regulatory moat and recurring revenue. $1,050,000

BioBlade V2 Pro multi-channel recovery-device brand: DTC recovery device combining heat, vibration, scraping, and micro-current in one unit, on $2M revenue and about $600K profit across channels. $1,342,620

Profitable Web3 marketing agency: 7-year systematized services business with 250+ SOPs and proprietary databases, a services roll-up profile at a 4.0x profit multiple. $1,400,000

For PE partners and operators seeking alpha

🌐 Scaling $50M - $500M+ mid-market companies with value creation through RevOps, data engineering, and WordPress. DevriX provides full RevOps consulting + delivery with GTM enablement for PE-backed portfolio companies, traditional tech, healthcare, finance, and professional service businesses pacing toward revenue growth initiatives. Our standard retainers between $10K and $60K include revenue lifecycle services for marketing and sales leaders, FP&A for financial teams, pipeline enrichment through websites and dozens of lead sources, automations and delivery integrations, CRO and ongoing testing, product delivery and platform integration solutions, and more through our consulting solutions.

🚀 1:1 Advisory retainers. Supporting operating partners, private equity funds, family offices, and mid-market executives in different capacities, from value creation through due diligence to portfolio digital GTM management in my async advisory programs via Growth Shuttle.

📈 GTM while scaling. European and international businesses can opt in for doola LLC and their “Business in a Box” model. Scaling founders can find smaller digital opportunities on Flippa. And additional opportunities across my investments can be found here.