Forecasting Experiences, Not Just Contacts: The Future of Service Operations

By Brian Keiner, Head of Data, serviceMob

For more than four decades, enterprises have forecasted contact center demand the same way: pull historical volumes, apply a rolling average, layer in a seasonal adjustment, and hand it to a WFM team that turns it into schedules. The tools have gotten shinier. The math has not.

That matters because forecasting is not a back-office exercise. It is the single largest lever on your cost to serve. Get the forecast wrong by even a few percentage points and you are either bleeding payroll on idle agents or bleeding customers through long wait times. Most enterprises are doing both simultaneously — they just cannot see it.

At serviceMob, we rebuilt forecasting from first principles. Instead of counting contacts, we forecast experiences — and the difference is not semantic. It is structural, statistical, and worth millions of dollars a year.

The 40-Year Forecasting Problem

Traditional WFM-based forecasting relies on a simple premise: tomorrow will look roughly like yesterday. Platforms ingest historical contact volumes, apply Erlang-C queueing models, and produce staffing requirements. The entire pipeline optimizes for one thing — scheduling bodies against projected traffic.

Here is what that approach ignores:

• Why contacts happen in the first place
• Which contacts are repeats driven by unresolved upstream failures
• What behavioral signals — product launches, billing cycles, outage patterns — actually drive demand density
• Whether the forecast itself has any statistical validity

As Gartner has noted, most WFM tools "do not incorporate real-time behavior, multichannel demand signals, or predictive modeling." Deloitte echoes the point: traditional models "do not account for repeat contacts, upstream failures, or digital interaction leakage."

The downstream consequence is massive. According to industry research from Gartner and Deloitte, most enterprises retain 15 to 30 percent more staff than necessary because their forecasts lack the observability to tell them otherwise. The vast majority of WFM teams never receive the critical inputs — demand drivers, repeat contact rates, channel-specific behavioral patterns — needed for effective demand prediction.

The forecast is wrong. The schedule is built on a wrong forecast. And the budget is built on a wrong schedule. That is the compounding cost of average-based planning.

Contacts vs. Experiences: Why the Unit of Measure Matters

When you forecast contacts, you are counting transactions. When you forecast experiences, you are modeling the conditions that produce those transactions.

Consider a customer who calls about a billing error, gets a partial resolution, calls back two days later, then opens a chat to escalate. Traditional forecasting sees three contacts. serviceMob sees one experience with a repeat contact signal, an unresolved root cause, and a measurable increase in customer effort.

That distinction changes everything about how you predict demand. If 40 to 60 percent of your contacts are repeat or avoidable — and across enterprise service operations, they typically are — then any forecast that treats each contact as an independent event is structurally overestimating required capacity. You are staffing for failure demand, not actual demand.

Experiential forecasting separates signal from noise. It accounts for the inputs that drive your forecast, not just the outputs your WFM tool can see.

FaaS: Forecasting as a Service — Layer 3 of the Answer Engine

FaaS sits within Layer 3 of the serviceMob Answer Engine architecture. While Layers 1 and 2 build the data ontology and quantify the customer experience, Layer 3 operationalizes that intelligence into forward-looking demand prediction.

FaaS is not a module you bolt onto your existing WFM stack. It is a fundamentally different forecasting methodology delivered as a managed service — platform plus Forward Deployed Engineers who learn your business, build your experience model, and continuously refine forecasting accuracy.

What FaaS actually does

The platform ingests data from inside and outside the contact center — interaction records, channel volumes, product telemetry, business calendars, marketing event schedules, and any other demand signal relevant to your operation. It then runs an ensemble of forecasting models and selects the champion based on rigorous backtesting.

The model ensemble includes NeuralProphet, linear regression, STL decomposition, exponential smoothing, dynamic linear models, XGBoost, random forest, SARIMAX, ARIMA, Prophet, Holt-Winters, and vector autoregression — among others. Each model is evaluated against the same holdout data using three primary statistical measures:

• R-squared — how much variance the model explains
• MAE (Mean Absolute Error) — average magnitude of forecast error
• MAPE (Mean Absolute Percentage Error) — average percent difference between predicted and actual values

The champion model is not selected by a human hunch. It is selected by statistical performance across backtested intervals. In production deployments, NeuralProphet consistently ranks among top-performing models for experiential volume prediction — but the ensemble ensures the best model wins for your specific data, your specific channels, and your specific demand patterns.

The output

FaaS delivers 98%+ forecasting accuracy at 15-minute intervals with +/-3% MAPE accuracy against the actual curvature of contact demand. That is not a marketing number. It is a statistically validated, backtested result — and it applies across voice, chat, SMS, email, and messaging channels.

One of the most powerful capabilities: FaaS can forecast demand for brand-new channels, even those with zero historical data. Because the models are built on experiential inputs and demand drivers rather than purely historical contact patterns, serviceMob can project demand density for a channel you launched last week.

The CPx Connection: Why Forecast Accuracy Is a Financial Lever

serviceMob's Cost per Experience (CPx) metric quantifies the true cost of each customer experience across the entire journey — not just the cost of handling a single contact. CPx and forecast accuracy are directly linked in a causal chain that most enterprises never see:

1. When CPx increases — meaning experiences are getting more expensive due to repeat contacts, escalations, or unresolved root causes — forecast error increases because the demand signal becomes noisier and less predictable.
2. When forecast error increases, WFM teams pad schedules to absorb the uncertainty. Staffing inflates.
3. When staffing inflates, cost rises — and the gap between actual demand and staffed capacity becomes a permanent budget drag.

This is why forecasting accuracy is not a WFM problem. It is a financial problem. And it is why forecasting contacts in isolation — without understanding the experiential factors that drive those contacts — guarantees you will overspend.

Reverse-Engineering Service Targets from Customer Tolerance

Traditional service levels are set by internal policy: "Answer 80% of calls in 20 seconds." But that target rarely reflects what customers actually tolerate.

serviceMob reverse-engineers ASA (Average Speed of Answer) targets based on customer abandonment tolerance. By analyzing actual abandonment behavior — not assumed thresholds — we can determine the real-world wait time at which customers leave. For one enterprise engagement, targeting a 5% abandonment rate yielded an ASA threshold of 64 seconds or less, translating to a service level of 70%/64 seconds.

The result: your SLA is based on the tolerance of your customers on any channel, for any product line — optimized for effort and efficiency. You stop staffing to an arbitrary target and start staffing to what your customers actually need.

What This Looks Like in Practice

Healthcare: $7.3M cost deflection

A disruptive healthcare company focused on preventive care and patient outcomes operated approximately 25,000 contacts per month across member support and provider support queues. They had no capability to understand how many agents they actually needed or how to drive positive customer experiences.

serviceMob deployed experiential forecasting with shift optimization and ontology-defined intelligent routing. The results:

• $7.3M in cost deflection
• 110 FTE reduction in required headcount
• 98.3% forecasting accuracy at 15-minute intervals
• 30% occupancy increase
• 70% abandonment rate reduction

Travel and Hospitality: $5M in savings

A major travel and hospitality enterprise was on a growth trajectory that, under its incumbent staffing model, would have required approximately 230 more FTEs than necessary over a nine-month period. serviceMob deployed platform capabilities, analytics, and experiential forecasting to provide the prescriptive data model needed to right-size the operation.

The results:

• $5M in total cost savings
• ~230 FTEs avoided versus the incumbent trajectory
• 52% reduction in voice resolution time
• 36% reduction in chat resolution time
• 54% reduction in SMS resolution time

In both cases, the savings were not achieved by cutting corners or degrading service. They were achieved by forecasting with precision — understanding what drives demand, not just counting what shows up.

Pricing Transparency

FaaS is priced at $12 per agent per month plus a $1,000 monthly hosting fee. That is the full cost — no hidden platform fees, no percentage-of-savings models, no multi-year lock-ins before you see value.

For an operation with 500 agents, FaaS costs $7,000 per month. The healthcare case above deflected $7.3M. The travel case saved $5M. The math is not subtle.

Pricing as of March 2026, subject to change. Contact serviceMob for current pricing.

Key Takeaways

• Traditional WFM-based forecasting has not fundamentally changed in 40+ years. It counts contacts using rolling averages and Erlang-C models — ignoring the demand drivers, repeat contact rates, and behavioral signals that actually determine volume.
• Enterprises retain 15 to 30 percent more staff than necessary because their forecasts lack statistical rigor and experiential context.
• serviceMob forecasts experiences, not contacts — accounting for repeat demand, channel behavior, and upstream root causes to deliver 98%+ accuracy at 15-minute intervals with +/-3% MAPE.
• FaaS uses an AI ensemble of 12+ models with automated backtesting and champion selection. The best model wins based on R-squared, MAE, and MAPE — not human judgment.
• Forecast accuracy directly impacts CPx: higher effort drives noisier demand signals, which inflate staffing, which inflates cost. Fixing the forecast fixes the financial model.
• Proven outcomes: $7.3M cost deflection in healthcare (110 FTE reduction, 98.3% accuracy), $5M savings in travel (230 FTEs avoided, 52% voice resolution reduction).
• Transparent pricing at $12 per agent per month + $1K hosting — no hidden fees.

Stop Forecasting Contacts. Start Forecasting Experiences.

If your WFM team is building schedules on a forecast that has never been backtested, never validated with MAPE or R-squared, and cannot tell you why demand is what it is — you are not forecasting. You are guessing with a spreadsheet.

serviceMob's FaaS replaces guesswork with statistical certainty. We forecast the experiences your customers will have, not just the contacts they will generate. And we do it at a price point that makes the ROI undeniable from month one.

Book a Working Session — we will show you your forecast accuracy gaps, your repeat contact exposure, and exactly how much experiential forecasting would save your operation.