How Collectly Improved Revenue Forecast Accuracy to 95% and Accelerated Growth
We went from not forecasting at all to running the business predictably. Tim Salikhov built the forecast infrastructure that enabled us to invest in growth – with confidence.
— Levon Brutyan, Co-Founder & CEO
Challenge
Why Collectly Couldn't Forecast Revenue Built on Usage, Not Contracts
Collectly was founded in 2016 to fix a problem healthcare providers had lived with for years: patients weren't paying their bills, and providers had no modern way to collect from them. Collectly built a patient payments platform that worked, and growth followed. By 2023 the company had raised a $25 million Series A led by Sapphire Ventures, reported by TechCrunch.
Unlike traditional subscription SaaS companies with fixed annual contracts, Collectly runs on a usage-based revenue model: providers process payment volume that swings month to month, client by client, tied directly to how much each healthcare practice actually collects. There's no flat ACV to anchor a forecast against. The company had never forecasted that revenue at all. It simply watched volume move and reacted.
That worked when there was no capital to deploy. It stopped working the moment there was.
"Raising the round was the easy part. The hard part was knowing how much of it we could actually afford to spend on growth in any given month."
Solution
How Bridges Built a Forecasting Model From Scratch, Client by Client
Tim Salikhov, CEO of Bridges, took on the finance function at Collectly five months after the round closed, and the mandate was immediate: figure out how to deploy capital into growth, prudently. In a B2B SaaS business built on consumption rather than subscriptions, that meant building forecasting infrastructure from zero. He structured the work around three pieces: forecasting existing clients, forecasting new clients, and a way to keep both honest over time.
A Data Analytics Muscle Built on Seven Years of Client History
Tim pulled seven years of payment and billing data into a single model and rebuilt it around how each client actually behaved.
- Segmented clients into enterprise, mid-market, and SMB, since each group moved differently.
- Mapped payment patterns by healthcare specialty, since a dermatology practice and an orthopedic group don't bill the same way.
- Layered in seasonality, identifying February–April as consistently strong months and August–October as consistently slow ones.
Once the data was organized this way, the patterns that had been invisible for years became obvious: enterprise clients were stable and grew predictably except around M&A events, while SMB clients carried real volatility and churn risk. That distinction became the basis for how every client's revenue got modeled going forward.
A Cold-Start Model for Clients With No Payment History Yet
The segmentation work exposed a second problem common to consumption-based B2B SaaS businesses: brand-new clients had no usage history to forecast from at all. Tim built a cold-start model using patterns from the more than 200 clients Collectly had onboarded before.
- Used accumulated accounts receivable at signing as the leading indicator of expected volume.
- Modeled a predictable spike in months one through four as backlogged receivables got collected.
- Set month six as the expected point where volume normalized, based on prior cohorts.
This gave Collectly a credible forecast for a client on day one instead of waiting six months to find out how they'd behave. Every forecast was then checked against what actually happened, which fed directly into tightening the model itself.
Scenario Planning That Built in a Margin for Error
With existing-client and new-client forecasts in place, the next problem was making sure the business didn't overcommit on a single number. Tim layered in bear and base case scenarios for every major capital allocation decision, deliberately erring conservative.
- Ran bear and base scenarios side by side rather than relying on one point estimate.
- Pulled in qualitative feedback from Sales and Customer Success on client health and pipeline risk to inform the assumptions feeding the model.
- Reconciled actuals against the forecast every cycle and fed the gap back in, the same way a model improves with more training data and correction over time.
That feedback loop is what pushed accuracy to within 10% consistently, and it's what let leadership trust the conservative case enough to actually act on it.
"We didn't need a model that was always right. We needed one we could trust enough to make a decision and move."
Results
Collectly Now Invests in Growth With a Forecast It Actually Trusts
Forecasting stopped being a finance exercise and became the thing that let Collectly move. The company built out a sales team, set quotas, and committed marketing spend, all against a forecast leadership didn't have to second-guess.
- 95% forecast accuracy — payment volume now lands within 10% of projections, even across volatile SMB cohorts.
- 50% decrease in monthly variance vs. budget — capital allocation decisions are made on data, not instinct.
- From weeks of prep to board-ready in hours — leadership reviews a live forecast instead of waiting on a manual rebuild.
What started as a Series A capital problem became the system this B2B SaaS company now uses to decide where every dollar of growth spend goes — and the forecasting discipline behind Collectly's 3x scale-up since the round.
"The forecast is the reason we can grow on purpose instead of grow and hope."
— Levon Brutyan, Co-Founder & CEO, LinkedIn