Marco Pellegrini hired two senior ML engineers in September 2024 because his agency was drowning in work. Pipeline was strong, three deals were "about to close," and utilization was running at 95%. By November, one of those three deals had fallen through, another was delayed by the client's budget freeze, and the third closed at half the original scope. Marco now had $45,000 per month in payroll for engineers with no billable work.
Marco did not have a revenue problem. He had a forecasting problem. He confused pipeline activity with committed revenue, mistook current utilization for future demand, and made hiring decisions based on optimistic projections rather than probabilistic analysis.
Revenue forecasting in a services business is different from revenue forecasting in product or SaaS companies. There are no recurring subscription revenues to anchor predictions. Each project has a defined start and end. Pipeline deals convert at wildly different rates. And the agency's capacity to deliver constrains how much revenue can actually be captured, regardless of demand.
Getting forecasting right is not about precision — it is about having enough visibility to make confident decisions about the three things that matter most: hiring, investment, and cash management.
Why Services Revenue Forecasting Is Hard
Before building the forecasting system, it is important to understand why services revenue is inherently harder to forecast than product revenue.
Lumpy revenue. Services revenue comes in discrete chunks — projects with defined start dates, durations, and values. A single large deal can move the needle 20-30% in either direction. This lumpiness makes month-to-month revenue volatile and hard to predict.
Long, unpredictable sales cycles. Enterprise AI deals typically take two to six months from initial conversation to signed contract. During that period, deals stall, budgets change, stakeholders rotate, and priorities shift. A deal that seems certain in March may not close until August — or may not close at all.
Delivery capacity constraints. Even if demand is high, your revenue is constrained by your team's ability to deliver. If your team is fully utilized, you cannot take on additional work without hiring, which has its own timeline and risk.
Client-dependent timing. Project start dates and payment milestones often depend on client readiness — data availability, stakeholder alignment, internal approvals. A project scheduled to start in April may not actually start until June because the client's data migration was delayed.
Variable deal sizes. AI agency deals range from $15,000 discovery engagements to $500,000+ implementation projects. This variance makes averages misleading and requires segmented analysis.
The Three-Layer Forecasting Model
An effective services revenue forecast has three layers, each covering a different time horizon and using different data.
Layer One — Committed Revenue (0-3 Months)
Committed revenue is the most reliable layer. It includes revenue from signed contracts, active projects, and ongoing retainers that are virtually certain to generate income in the forecast period.
What counts as committed revenue:
- Active projects. Revenue from projects currently in progress, based on the remaining contracted value and the expected billing timeline. If a $120,000 project is 40% complete and you are billing monthly, the remaining $72,000 is committed revenue spread over the expected remaining timeline.
- Signed contracts not yet started. Projects with signed SOWs and scheduled start dates. The revenue is committed but timing may shift.
- Active retainers. Monthly retainer revenue from ongoing clients. Include only the remaining term (if the retainer has a defined period) or the expected duration (if it is month-to-month, assume three to six months unless you have reason to expect otherwise).
- Approved change orders. Additional scope that has been approved and will be billed beyond the original contract value.
What does not count:
- Verbal commitments without signed contracts
- Expected upsells that have not been approved
- Renewals that have not been confirmed
Forecast confidence: Committed revenue should be 85-95% reliable. The primary risk is project delays, scope reductions, and (rarely) contract cancellations.
Layer Two — Weighted Pipeline (1-6 Months)
Pipeline revenue is less certain than committed revenue and requires probability weighting. Each deal in your pipeline has a probability of closing based on its stage in the sales process.
Typical stage-based probabilities:
- Initial conversation / discovery: 10% probability
- Proposal submitted: 25% probability
- Proposal in review / negotiation: 50% probability
- Verbal commitment / awaiting signature: 75% probability
- Contract sent for signature: 90% probability
Calculating weighted pipeline revenue: For each deal, multiply the deal value by the probability. A $100,000 deal at the proposal stage has a weighted value of $25,000.
Critical adjustment — calibrate your probabilities: The default probabilities above are starting points. Your actual conversion rates by stage may be different. After twelve months of tracking, you will have enough data to calibrate your probabilities based on actual outcomes. If only 15% of your submitted proposals convert to signed contracts (instead of the assumed 25%), adjust your probability accordingly.
Timing adjustment: In addition to probability weighting, estimate when each deal will convert and when revenue will be recognized. A $200,000 deal at 50% probability that you expect to close in two months and deliver over four months generates approximately $25,000 per month in weighted revenue starting in month three.
Layer Three — Projected New Business (3-12 Months)
Beyond your current pipeline, you need an estimate of new business that will enter and convert through your pipeline. This is the least reliable layer but essential for longer-term planning.
Estimating projected new business:
- Historical conversion rates. If your pipeline generates $500,000 in weighted value per quarter and your close rate on weighted pipeline is 60%, you can expect approximately $300,000 in new revenue per quarter from pipeline conversion.
- Lead generation trends. If your lead volume is increasing, your pipeline should grow. If it is flat or declining, future pipeline will be thinner.
- Market conditions. Macro factors — economic conditions, industry trends, competitive dynamics — affect future demand. Factor these in qualitatively, not quantitatively. A recession does not mean zero demand, but it may mean a 20-30% reduction in deal velocity and average deal size.
- Seasonal patterns. Most agencies experience seasonal patterns in their pipeline. Q4 and Q1 are typically slower for new business in enterprise markets. Factor these patterns into your projections.
The honest range: For projected new business, express the forecast as a range rather than a single number. "We expect $250,000-$400,000 in new business revenue in Q3" is more honest and more useful than "$325,000."
Building the Monthly Forecast
Combine the three layers into a monthly revenue forecast:
Month 1-2: Primarily committed revenue (80%+), supplemented by near-term pipeline deals (15-20%).
Month 3-4: A mix of committed revenue (50-60%), pipeline deals (25-35%), and a small amount of projected new business (10-15%).
Month 5-6: Committed revenue decreases (30-40%), pipeline deals carry more weight (30-40%), and projected new business becomes significant (20-30%).
Month 7-12: Primarily projected new business with a wide confidence range.
Visualize the forecast: Present the forecast as a chart showing committed, weighted pipeline, and projected layers stacked on top of each other. This visual makes it immediately clear how much of the forecast is reliable versus speculative.
Using the Forecast for Decisions
The forecast's value is not in its precision — it is in the decisions it enables.
Hiring decisions. Only hire based on committed revenue or very-high-probability pipeline. If your committed revenue supports your current team and your weighted pipeline (at conservative probabilities) supports one additional hire, that is a defensible hiring decision. If you need projected new business to justify the hire, wait until the pipeline develops further.
Cash management. Use the committed revenue layer to plan cash flow. Project when invoices will be sent and when payments will be received. Compare expected cash inflows to committed outflows (payroll, rent, tools, taxes). Identify any months where outflows exceed inflows and plan accordingly.
Investment decisions. Investments in tools, marketing, infrastructure, and training should be funded by committed and near-term weighted revenue, not projected new business. When revenue is strong and cash reserves are healthy, invest. When the forecast shows uncertainty, conserve.
Pipeline development. The gap between your revenue target and your committed plus weighted pipeline revenue tells you how much new business you need to develop. If your Q3 target is $600,000 and committed plus weighted pipeline shows $400,000, you need $200,000 in new business. That tells you how aggressively to invest in sales and business development.
Pricing decisions. If your forecast shows the team will be highly utilized in the coming months, you have pricing power — demand exceeds supply, and you can price at a premium. If the forecast shows light utilization ahead, you may need to be more competitive on pricing to win deals and keep the team busy.
Forecasting Disciplines
Update weekly. Revenue forecasts decay rapidly. Deals move forward, stall, or die. Projects accelerate or slip. Update your forecast every week to maintain accuracy.
Track accuracy. Compare your forecasted revenue to actual revenue every month. Calculate the variance and identify patterns. Are you consistently over-forecasting (optimistic bias)? Under-forecasting? Missing on timing? Tracking accuracy reveals systematic errors that you can correct.
Separate forecasting from goal-setting. A forecast is a prediction of what will happen. A goal is an aspiration of what you want to happen. These are different things, and conflating them produces unreliable forecasts. Your forecast should be realistic, not aspirational. Use goals for motivation and forecasts for planning.
Involve the team. The people closest to each deal and each project have the best information about probability and timing. Involve project managers and sales leads in the forecasting process rather than building the forecast top-down from your own assumptions.
Scenario planning. In addition to your base forecast, create optimistic and pessimistic scenarios. The optimistic scenario assumes higher conversion rates and faster timing. The pessimistic scenario assumes lower conversion rates and delayed timing. Planning for the pessimistic scenario ensures survival. Preparing for the optimistic scenario ensures you can capitalize on upside.
Common Forecasting Mistakes
Counting pipeline as revenue. A $200,000 deal at 50% probability is not $200,000 in revenue. It is $100,000 in expected value with significant uncertainty. Treat pipeline deals as what they are — probabilistic expectations, not commitments.
Ignoring deal decay. Deals that have been in the pipeline for months without progressing are probably dead. Establish a rule: if a deal has not advanced stages in sixty days, reduce its probability by 50%. If it has not advanced in ninety days, remove it from the forecast.
Confusing billing with revenue recognition. When you bill a milestone is not necessarily when you should recognize the revenue. For forecasting purposes, recognize revenue when the work is performed, not when the invoice is sent or the payment is received. This gives you a more accurate picture of when your team is generating value.
Not accounting for delivery constraints. Your revenue forecast should not exceed your delivery capacity. If your team can deliver $150,000 per month at full utilization, a forecast showing $250,000 per month means you either need to hire (with the associated cost and ramp time) or some of that revenue will not materialize.
Over-relying on one or two large deals. If your forecast depends heavily on one or two large deals closing, your risk is concentrated. Diversify your pipeline so that no single deal represents more than 20-25% of your forecasted revenue in any given quarter.
Your Next Step
Build your first three-layer revenue forecast this week. Start by listing all committed revenue — active projects, signed contracts, and ongoing retainers — with their expected billing timelines. Then list every pipeline deal with its stage, probability, expected value, and expected timing. Finally, estimate projected new business based on your historical pipeline performance.
Plot the results on a monthly chart for the next six months. That chart will immediately reveal whether you have a revenue visibility problem, a pipeline gap, or a capacity constraint. And that visibility is the foundation for every confident business decision you will make in the months ahead.