Support Volume Forecasting: Predict Chat and Ticket Load

Support volume forecasting is the habit of estimating how many chats and tickets will arrive next week, next month and during your busiest season, so you can schedule people before the queue grows instead of after. You do not need a data science team for it. A spreadsheet, twelve months of history and a few honest assumptions will get you within a range that is good enough to plan shifts, set targets and decide when to hire.

This guide shows a simple method that small teams can run in an afternoon: where the numbers come from, how to split them by day and hour, how to add known events, how to turn a forecast into a number of agents and how to check whether your forecast was any good.

Why Forecast Support Volume at All

Most small teams staff by feel. Someone notices that Mondays are heavy and puts the most experienced agent on that day. That works until a product launch, a shipping delay or a holiday sale doubles the traffic and nobody saw it coming. A forecast turns the same instinct into something you can write down, check and improve.

A good forecast helps in four concrete ways:

Schedules match demand. You put people where the hours are busy and avoid paying for quiet hours.
Response times stay stable. A queue grows when demand beats capacity for even a few hours, and the damage to your first reply time lasts longer than the spike itself.
Hiring is justified. Showing that volume has grown by a steady percentage for six months is a stronger argument than saying the team feels stretched.
Leaders hear about spikes early. If marketing plans a campaign, support can say what it will cost in conversations before the campaign starts.

Support Volume Forecasting Starts With the Data You Already Have

You need three series, ideally for the last twelve months: conversations per day, tickets per day and the hour of day each one arrived. Keep chats and tickets apart, because they behave differently. Chats follow visitors on your site and peak when traffic peaks. Tickets that arrive by e-mail follow your customers’ working days and your mailing schedule.

In Talkmio, the Reports section already shows conversations per day split into chats and tickets, busiest hours in your own time zone, hand-offs, first-reply time, the share answered by Mio alone, channels and countries. The pricing page lists how long history is kept on each plan: 30 days on Free, 365 days on Pro and unlimited on Ultimate. That matters for forecasting, because seasonal patterns need a full year of data to show up. If you are on the Free plan, export or note the numbers every month so you build your own archive. CSV export of reports is part of the Ultimate plan.

Before you trust the numbers, clean them. Remove obvious one-off events such as an outage day, test chats from your own team and bot-only conversations that never reached a person. Write down what each unusual day was, because you will want to explain it later.

Separate the questions Mio answers from the ones your team answers

If an AI assistant answers a large share of conversations, total volume is not the same as human workload. Forecast both. Total conversations tell you about demand on your site. Hand-offs, the conversations that need a person, tell you about staffing. Our guide to the AI deflection rate explains how to measure the split.

Find the Patterns: Trend, Season and Time of Day

Almost every support queue is built from the same four layers, and you can estimate them one at a time.

Trend. Is the weekly total growing, flat or shrinking? A rough measure is the average of the last eight weeks compared with the eight weeks before.
Seasonality. Does volume rise before holidays, at the start of a school year or during a sales period? Compare the same weeks in the previous year. The seasonality article on Wikipedia gives a good general overview of the idea.
Day of week. Divide each weekday’s average by the average of all days. If Monday is 1.3 and Saturday is 0.6, you have your weekday weights.
Hour of day. Do the same for hours. Many sites show a morning ramp, a lunch dip and an evening peak, but the shape depends on your country and customers.

Our article on busiest hours for live chat goes deeper on the hourly layer and shows how to read a heat map of your own traffic.

Choose a Forecasting Method That Fits Your Size

The best method is the simplest one you will actually keep updating. This table compares the four approaches small teams use most.

Method	How it works	Needs	Weak spot
Same week last year	Copy last year’s volume and apply a growth factor	12+ months of history	Fails after a big product or pricing change
Moving average	Average of the last 4 to 8 weeks	A few weeks of data	Lags behind sudden growth; ignores seasons
Seasonal index	Base level times weekday, hour and month weights	12 months of history	More setup; weights need an annual refresh
Driver-based	Volume per order, per signup or per campaign visit	Business numbers from other teams	Needs accurate plans from marketing and sales

For most teams with fewer than ten agents, a seasonal index plus a driver adjustment is the sweet spot. A moving average is a fine starting point if you have only a couple of months of history.

A Worked Example With Illustrative Numbers

The numbers below are invented to show the arithmetic. Replace them with your own.

Imagine a web shop whose last eight weeks averaged 400 conversations a week, of which 40 percent reached a person. Last year the same three weeks before the winter holidays ran 35 percent above the yearly average, and orders are expected to be 20 percent higher than a year ago.

Base level: 400 conversations a week.
Season: multiply by 1.35 for the holiday weeks, which gives 540.
Growth and drivers: add 20 percent for expected order growth, which gives 648 conversations a week.
Human workload: 40 percent of 648 is about 259 conversations that need a person.
Weekday weights: if Monday carries 1.3 of an average day, Monday gets about 48 human conversations (259 divided by 7, times 1.3).

That is a plan, not a prophecy. It tells you to expect roughly a quarter more human conversations than today during the holiday weeks and to schedule accordingly. Our Black Friday live chat checklist covers the operational side of such a peak.

Add Known Events and Drivers

History explains the past. Events explain the future. Keep a shared calendar of anything that can change volume and ask other teams to add to it:

Marketing campaigns, newsletters and paid traffic bursts.
Product releases, price changes and new features.
Shipping cut-off dates, carrier delays and public holidays.
Billing dates, renewal runs and invoice e-mails.
Planned maintenance and known outages.

For each event, record what happened last time: how many extra conversations, over how many hours, and which topic dominated. After two or three repeats you will have a rule of thumb such as “every newsletter adds about 30 chats within three hours”.

Turn the Forecast Into Staffing

A forecast of conversations is only useful once you convert it to hours of agent time. You need three inputs: the number of conversations that need a person, the average handle time and the share of an agent’s day that can realistically be spent on live work, often called utilisation.

The formula is short: required agent hours equals human conversations times average handle time divided by utilisation. If 259 conversations take 8 minutes each and agents spend 70 percent of paid time on live work, you need roughly 49 agent hours across the week. Add a buffer for the peak hours, because averages hide the busiest periods. Queueing theory, such as the Erlang unit used in call centres, explains why you need extra capacity to keep waits short.

For help with the handle time input, read our article on average handle time in live chat, and for the ratio of agents to visitors see live chat staffing ratio.

Plan for the shape, not just the total

Two weeks with the same total can need very different schedules. A week with a flat load is easy. A week with a sharp lunchtime peak needs overlapping shifts. Use your hourly weights to place breaks and meetings outside the peaks, and use Talkmio’s Online, Away and Offline statuses so visitors reach a person when one is available and leave an e-mail when nobody is.

Check Your Forecast and Improve It

A forecast you never score will never get better. Each week, compare the forecast with what happened and note the error as a percentage. A simple measure is the absolute difference divided by the actual value, averaged across the days of the week. If you are within 10 to 15 percent most weeks, the method is good enough for scheduling.

When a week misses badly, ask which layer failed:

Trend wrong: growth changed and you used an old rate.
Season wrong: this year’s peak moved to a different week.
Driver missed: an event was not on the calendar.
Data dirty: test chats or an outage distorted the base.

Fix the layer, not the whole model. Keep the review short, ten minutes on Monday morning is enough, and write one line in a log. After a quarter you will see which assumptions keep failing.

Common Forecasting Mistakes

Forecasting only totals. Hour and weekday shape decide whether customers wait.
Mixing chats and tickets. They respond to different drivers and are answered at different speeds.
Ignoring the AI share. If you improve the knowledge base, human workload can fall while total conversations rise.
Chasing precision. A range with a best case and a worst case beats a single number that looks exact.
Never updating. Re-estimate weights at least once a year and after any major product change.

A Simple Monthly Routine

Put the whole process on a calendar so it survives busy weeks.

On the first working day of the month, export conversation counts by day and hour from Reports.
Update the base level with the last eight weeks and refresh the weekday and hour weights if they moved.
Add the events calendar for the next six weeks and apply the driver adjustments.
Convert the human workload to agent hours and compare with the schedule.
Share a one-page summary with the team: expected peaks, planned cover and the biggest risk.

Frequently Asked Questions

What is support volume forecasting?

Support volume forecasting is estimating how many chats, tickets and other contacts you will receive in a future period. It combines recent history, seasonal patterns and planned events, and it is used to schedule agents, set response time targets and decide when to hire more people.

How much history do I need to forecast support volume?

Eight weeks is enough for a basic moving average. To capture seasonal peaks such as holidays or school terms you need at least twelve months. On Talkmio, the Free plan keeps 30 days of history, Pro keeps 365 days and Ultimate keeps unlimited history, so export monthly if you are on Free.

Should I forecast chats and tickets separately?

Yes. Chats follow visitors on your website and peak with traffic, while tickets from e-mail follow customers’ working days and your mailings. They also have different handle times and response targets, so mixing them hides the patterns you need for scheduling.

How accurate should a support forecast be?

For scheduling, an error of 10 to 15 percent on weekly totals is usually good enough, while daily and hourly errors will be larger. Track the error every week and look for the cause of big misses rather than trying to reach perfect accuracy.

How does an AI assistant change volume forecasting?

It splits volume into conversations Mio answers alone and hand-offs that need a person. Forecast both. Total conversations show demand on your site, while hand-offs drive staffing, and improving the knowledge base can lower hand-offs even when total demand grows.

Which numbers do I need to turn a forecast into staffing?

You need the forecast of conversations that need a person, the average handle time and a realistic utilisation rate for your agents. Divide the total handle time by utilisation to get agent hours, then add a buffer for peak hours so customers do not wait.

The Bottom Line

Support volume forecasting works when it stays small: twelve months of daily and hourly counts, a base level, weekday and hour weights and a calendar of events. Convert the result into agent hours, score it every week and fix the layer that missed. Talkmio’s Reports give you the daily, hourly and hand-off numbers to start with. Create a free Talkmio account and begin collecting the history that your next forecast will rely on.