Genesys Cloud workforce management: the workforce planner's handbook

1.1 What it does

Genesys describes its workforce management features as simplifying the forecasting of interactions and the scheduling of agents in multichannel contact centres. In practice that means six jobs: developing forecasts; comparing forecast and actual values during the day; generating and managing blank schedules, forecast-based schedules and schedules without a forecast; monitoring adherence to schedules; showing agents their schedules; and managing time-off requests.

WFM is a planning loop, not a single tool. Forecast the work, size the workforce, build the schedule, publish it, run the day, measure the gap, and feed what you learn into the next forecast. Genesys's own resource management use case follows the same order, from business unit configuration through forecasting, scheduling, publishing and intraday management to historical adherence. If you only ever do the middle step, the whole thing drifts.

1.2 The channels it plans for

Workforce management forecasts, schedules and monitors six media types: voice, email, chat, callback, messaging and workitems. A workitem is a task routed to an agent that does not necessarily involve a customer conversation, such as a back-office case.

Concurrency matters for digital work. By default, Genesys Cloud lets the routing engine give an agent one callback, up to four chats, one email, up to four messages, one voice call and up to four workitems at a time, configurable for the organisation and for individual agents. A staffing number for chat is therefore not the same thing as a staffing number for voice. Genesys handles chat concurrency when it generates staffing requirements, not in the forecast: the forecast records volume and handle time as they happened, including time spent on concurrent chats, and schedule generation and capacity planning then remove the overlapping time to estimate real agent effort. Read the Genesys article on chat concurrency before you plan a large chat or messaging queue.

1.3 Licensing: what you need

On the Genesys Cloud CX pricing page, checked on 13 November 2026, workforce management (AI-powered forecasting, AI scheduling, capacity planning, time-off management, and monitoring and adherence) is included in CX 3 and CX 4 and is an add-on for CX 1 and CX 2. Employee performance features (gamification, coaching, learning) follow the same pattern. The Genesys Tempo mobility app is an add-on for CX 1 and included from CX 2. Several individual WFM features list the subscriptions they need in their release notes and articles, so check before you promise a feature to the business.

On the add-on route, billing follows use. Genesys bills an agent for the WEM add-on when a schedule is created for them in the WFM scheduler, among other triggers, and some administrator roles are billed immediately. Do not add agents to management units or schedules casually, and test in a non-production organisation. Our workforce engagement management handbook sets out every trigger.

1.4 The planning hierarchy: the most important thing to get right

Genesys Cloud WFM organises the world into four objects. Nearly every early confusion comes from mixing them up. The table lists them from largest to smallest.

1.5 Route paths and service goal templates

A route path is one unique combination of queue, media type, language and skill set. A service goal template is a reusable, business-unit-level definition of what good looks like, in three parts: service level (the percentage of interactions answered within a time), average speed of answer and abandonment rate. You attach a template to planning groups; the template itself applies goals to nothing until you do.

The chain, in one line: route path, planning group, service goal template, forecast, required staff, schedule. When a number looks wrong further down, walk back up this chain.

1.6 Activity codes: how time is classified

An activity code is anything you can put on a schedule. It tells the agent what to do and tells WFM how to count that time. Each code carries settings that decide pay, occupancy, adherence and shrinkage, so treat the list as a controlled dictionary rather than something anyone can add to.

1.7 Work plans: the rules an agent's week must obey

Work plans hold shift types together with weekly, weekend, planning-period and general rules that reflect employment contracts, in one place, assigned to a set of agents. The layers are:

1. Weekly: minimum and maximum paid hours, minimum and maximum days per week, minimum consecutive rest per week, flexible paid hours, flexible shift start and end, maximum shift start variance and minimum time between shifts.
2. Weekend: limits on consecutive working weekends and on working weekends within a planning period.
3. Planning period (weekly or monthly): minimum and maximum days off and paid time across the period.
4. General: daily paid time divisible by a set number of minutes (default 15), maximum consecutive working days, and minimum time between shifts and between shift starts.
5. Daily shifts: the shift patterns themselves, fixed or flexible, and the activities inside them such as breaks and meals.

1.8 Time, intervals and granularity

Several different time grains are in play, and confusing them causes subtle errors.

• Scheduling granularity is set per business unit: one minute or five minutes, with five as the default. At five minutes, work plan shift and activity times are rounded to five minutes. One-minute granularity applies to shift start and end times and to minimum and maximum paid hours, and helps with contracts such as 38 hours over five days (7 hours 36 minutes a day). Flexible shifts still move in five-minute steps: a flexible start between 8:01 and 8:16 pm can begin at 8:01, 8:06 or 8:11.
• Forecasts work in 15-minute intervals. The continuous forecast, short-term forecasts and the daily values and distributions modification all show the day in 15-minute intervals.
• Shrinkage percentages on the management unit can be set in 15-, 30- or 60-minute intervals across a model week.
• Two time zones are always in play. Schedules and limits use the business unit time zone; most pages can be viewed in a different display time zone. In the schedule's vertical view the date column stays on the business unit date while the timeline follows the display time zone, with a purple overlay for time outside the business unit date. Each day is still a 24-hour window at a daylight saving change, so use Day view for shifts that cross one.
• The business unit's start day of the week and time zone cannot be changed after you create it. If they are wrong, you create a new business unit and configure it to match. This is the least reversible decision in WFM set-up.

1.9 Permissions and divisions

WFM permissions sit under Workforce Management and are granted through roles; changes can take a few minutes to take effect. Two dependencies are easy to miss: users need Routing > Queue > Search to see queue names and Routing > Skill > Manage to see skill and language names. Without them, names appear as placeholders such as Unknown1 and Unknown2, consistent within a screen but not across screens.

Where you grant a permission matters as much as which permission you grant.

• Work plans and time-off requests belong to management units; a business-unit grant covers all its management units.
• Forecasts belong to business units; granting forecast permission at management-unit level does nothing.
• Schedules belong to business units with access control at management-unit level. To generate a schedule, reschedule or publish for the first time, a user needs Workforce Management > Schedule > Edit at business-unit level.
• You need a separate division for every distinct access-control building block.

1.10 Configuration limits to plan around

Genesys documents working limits of 10 management units per business unit, 500 management units per organisation, 1,500 agents per management unit and 5,000 agents per business unit. A July 2023 release raised the maximum allowable values to 250 business units per organisation (from 50), 20 management units per business unit, 6,000 agents per business unit and 2,000 per management unit, but Genesys did not change the defaults: raising them in your organisation needs a ticket with Genesys Customer Care.

Plan against the defaults and raise the ticket early if your design needs more. The full list of per-object limits (activity codes, activity plans, forecasts and modifications, planning groups, staffing groups, route paths, schedules, service goal templates, shifts, work plans, bids and rotations) is on the Genesys page Workforce management supported configuration. Check it before committing to a large design rather than relying on any summary, including this one.

2.1 Before you build: Genesys's planning checklist

Genesys publishes a planning checklist for workforce management, and it is the step most teams skip. Work through it with operations before anyone opens the configuration screens.

• Define the roles and responsibilities of the WFM team, and give it real authority over scheduling and agents' time; Genesys treats adherence as a key performance objective the WFM team influences.
• Agree accessibility targets for each queue or interaction type at every level of the contact centre.
• Check queue design: Genesys notes that one consolidated pool of agents is more efficient than several smaller ones, and WFM cannot rescue a routing design that fragments people.
• Document the shifts on offer, with start, break and lunch times, and look for practical flexibility such as a start window instead of a fixed start.
• Train the WFM team on the technology and their responsibilities, and document the processes for forecasting, scheduling, real-time management, time off and training time.
• Identify the patterns that cause variability in your history so they reach the forecast.
• Build a service level recovery plan: the options available in advance and in real time when you are over- or under-staffed (see 8.2).
• Communicate the change and the reasons for it down to agent level. Genesys's own words: gaining buy-in at all levels is critical.

2.2 How to decide planning group granularity

A planning group is limited to one media type and normally holds one route path. You may combine queues, languages and skill sets, and for low-volume route paths combining is a good idea: it reduces planning work and improves the forecast because WFM treats them as one bucket.

The trap is in scheduling. If you combine route paths, the forecast shows only the combined value and the scheduler looks for agents who can handle the entire planning group. An agent who handles only part of it will not be scheduled against that workload but will still be given a schedule, so you can have people on shift whom the engine thinks are doing nothing useful. Combine only where agents handle the full combination, and check the planning group's Agents tab to see who matches.

2.3 Staffing groups, and why you want them

Staffing groups let you apply time-off limits to part of a management unit instead of the whole, for example giving 40 hours of allowance to one group rather than spreading it across everyone, and membership can carry effective dates so changes are planned in advance.

They also feed capacity planning. A capacity plan's staffing requirements come from links between planning groups and staffing groups. Any planning group without a staffing group, or without a forecast, is left out of the plan, and the only sign is a message that some planning groups are not associated. Check the association first when a gap looks suspiciously absent.

3.1 What the forecast is

A forecast predicts, for each 15-minute interval and planning group, Offered (the number of interactions) and Average Handle Time (talk, hold and after-call work).

You can start a forecast up to 26 weeks in the past or 104 weeks in the future, and its end cannot be more than 130 weeks from the start week; if you ask for more, Genesys shortens it and tells you. The 104-week forward reach arrived in January 2026 to support long-term staffing, budgeting and compliance planning.

3.2 The four ways to create a forecast

Use Automatic Best Method unless you have a specific reason not to. Genesys recommends it in one case in particular: with Genesys Cloud EX and externally injected data, because the weighted historical index method permanently caches conversation data older than a day and permanently excludes anything not yet injected, so any delay in injection makes those forecasts unreliable.

Automatic Best Method works with as little as one week of history, but is much stronger with more: it looks at up to 90 days to set daily arrival patterns, and a full season or several seasons give the best seasonality. Genesys also improved its short-term prediction of deferred work, such as email and workitems, in April 2025, so the forecast better reflects when that work is actually done.

3.3 Creating a forecast, step by step

1. Go to Menu > Workforce Management > Forecasting > Forecasts.
2. Choose the business unit at the top right, and change the display time zone if you need to.
3. Click Add and pick the forecast start week.
4. Set the number of weeks, within the 130-week limit.
5. Tick the option to use the short-term forecast for scheduling if you will build schedules from it.
6. Write a description that distinguishes this forecast from others for the same week: what you changed and why. You will not remember.
7. Choose the creation method; for the weighted historical index, set the source day weights, giving most weight to the periods that best represent the future.
8. Click Add Forecast, then review it against history before anyone schedules from it. For Automatic Best Method forecasts, Genesys shows the past 35 days of Offered and Average Handle Time in the short-term view and five years in the long-term view, including imported history.

3.4 Getting history in before you have any

Two features import history, and they are easy to confuse.

• Validated files take up to 24 hours to import; while Pending, they complete at about 3:00 am local time in each region.
• Planning group route paths are never created for you from imported data; build them by hand first.
• The historical end date cannot be in the future. After cutover, import a fresh file up to the cutover date and move the end date, so no gap appears between imported and native data.
• Queues, skills and languages must match exactly. Export from a spreadsheet rather than hand-editing text, and check that names containing commas are quoted.

3.5 Modifying a forecast

Models cannot know about a campaign, a changed bank holiday, a product launch or a letter going to every customer next Tuesday. Forecast modifications are where you add that knowledge. Each one changes Offered or Average Handle Time for chosen planning groups over a chosen time, and all modifications are enabled when created. Up to 200 are allowed per forecast, and in rare cases that limit reduces with the number of planning groups and the impact of existing modifications.

3.6 The main forecast and the continuous forecast

Every night Genesys calculates a continuous forecast using Automatic Best Method and shows it, read-only, on the Main Forecast page: a year ahead at 15-minute intervals, plus weekly forecasts for the first and second years. It cannot be used for scheduling. The page also shows actuals, in historical, trend and forecast views. At 15- to 60-minute granularity it holds the first 59 weeks of forecast and the latest 31 weeks of actuals; daily granularity reaches up to 261 weeks (five years) of actuals.

A Detected Outliers tab lists the actuals Genesys judged abnormal and normalised before using them. Forecast snapshots keep earlier nightly versions so you can see what the forecast said a month ago; they show system predictions only, without manual modifications, and cannot be used for scheduling or capacity planning. Since March 2026 a public API can export continuous forecast data, history and snapshots for a date range for use in other planning tools.

Use the main forecast for trends, budgeting and early warning, and a short-term forecast you created and enabled for scheduling to build schedules.

3.7 Forecast accuracy habits

• Measure forecast error by planning group and by interval, not only by day. A day that is 98% accurate can hide a large miss in the morning ramp, and the intervals are where service is won or lost. A 2018 ICMI study found that only about one in four contact centres measured accuracy at half-hour level.
• Read the Detected Outliers tab weekly. If Genesys normalised something you expect to happen again, add a deliberate modification rather than arguing with the model.
• Keep an events register (marketing, product changes, billing runs, weather-sensitive lines, public holidays, regulatory dates) with the expected effect, and review it at every forecast refresh.
• Keep a log of bad forecast days and their causes; it is the raw material for next year's modifications.
• Do not over-modify. Each modification is your judgement replacing a model that used all the data.

5.1 Three ways to make a schedule

5.2 Generating a load-based schedule

Before you start, check that agents are schedulable, work plans exist and are assigned, a forecast exists and is enabled for scheduling, and time off and activity plans are in place. Then go to Menu > Workforce Management > Scheduling > Schedules, select the business unit, and generate from the forecast, choosing the start week, number of weeks and a description. The schedule period must match the forecast period. The result is an unpublished schedule listed under Last Scheduling Attempts.

• Only one schedule per business unit can be generating at a time.
• A forecast-based schedule may take 20 minutes or longer. Plan your morning around it and do not click Generate repeatedly.
• Only one schedule can be published for a given date range. To replace a published schedule, publish one that fully encompasses it: you cannot publish two weeks over the first half of a published four-week schedule, but you can publish four weeks over a published two weeks if no adjacent published schedule overlaps.
• When an unpublished schedule overlaps a published one, assignments from the published schedule still count towards coverage, so the editor can show understaffing that is not real. Do not fix a phantom gap.
• After you change a service goal template, Genesys uses a default model for new requirements until the overnight build creates a better one. Review requirements, and generate schedules, the next day.
• When you open the current week, the schedule opens on today, not on the first day of the week.
• New hires can be added to a published schedule without rebuilding it for everyone else, through an intraday reschedule, provided they are part of a configured work plan.

5.3 Reading the result: messages, Scheduled, Required and Performance

Judge the schedule before you touch it. On the Schedules page, the Generation Results column shows counts of messages; click it to see the list, filter by severity and copy the details. Genesys ships a set of standard messages, each with a default severity of Error, Warning or Information, and lets you change the severity per business unit, including Ignore, which hides a message. Part 8 sets out how to triage them.

At the bottom of the schedule editor, choose Scheduled, Required and Performance to see, per interval, the people you have, the people the workload needs, the difference and the predicted service. Required Staff excludes shrinkage; Required Staff with Shrinkage adds the unplanned shrinkage configured on the management unit. Daily summaries give agents with shifts, paid hours and total hours.

5.4 Editing and publishing

Editing a generated schedule is normal. In every view you can assign a shift to another agent, swap shifts, replace shifts with activities, copy activities and shifts, remove them, filter, sort, and add or remove agents. When you are ready, publish: publishing makes the schedule visible to agents, and any later change needs republishing.

Agents are notified when a published or republished schedule changes their future activities, but only if you choose the notification type when you publish and their role includes Workforce Management > Agent Schedule Notification > View. Publishing or republishing takes a few minutes to reach real-time adherence.

5.5 Work plans in practice

The quality of a schedule is mostly decided in the work plans, before generation.

• Match the plan to the contract and to the law. Encode rest between shifts, maximum consecutive days and weekly hour limits as constraints and do not override them by hand. In the UK, for example, adult workers are usually entitled to 11 hours' rest between working days, 24 hours' uninterrupted rest a week (or 48 hours a fortnight), a 20-minute break when working more than six hours, and a 48-hour average week unless they opt out in writing.
• Use flexibility where the money is. Flexible start and end times and flexible paid hours let the engine follow the demand curve without adding people. Start with a modest window and widen it only if service gaps persist.
• Mark optional days as optional, so part-time or flexible capacity is used only when the forecast justifies it.
• Put breaks and meals in the shift definitions, not in your head, and use Counts as Work Time so the contiguous work time constraint protects people from long stretches without a break.
• Keep rotating plans constraint-compatible, remembering the minimum-of-minimums and maximum-of-maximums rule.
• Validate before you save. The work plan editor has a Validate button; use it as you go.

5.6 Activity plans: meetings, training and compliance without wrecking service

Activity plans schedule recurring or one-off group and individual activities at the best time, taking account of attendee and facilitator availability and the impact on service goals: monthly one-to-ones, team meetings, training and annual compliance, or one-off events such as product training or a town hall. Attendee and facilitator lists can be rule-based and are updated when the plan runs.

An active plan adds sessions to future schedule generations automatically; to add sessions to schedules already published, open the plan and click Run Now. Making a plan inactive stops future occurrences. Sessions can only be placed during On Queue time or activities marked Is Interruptible; everything else, including work plan activities and manually added activities, is protected. Settings include group size, maximum simultaneous sessions, maximum total sessions, recurrence, a start time increment that sets valid session start times, and transition time for facilitators.

5.7 Time off and how it lands on the schedule

Time off competes with the forecast for the same people, so it is a planner's job. Three objects work together.

• Read the pending statuses: they say why a request did not auto-approve, for example insufficient balance, outside any work plan shift, waitlisted for limit capacity, restricted activity code, removed from the waitlist after a move between management units, waitlist window closed, or blocked date. A request for an agent who is not scheduled and has no work plan cannot auto-approve.
• Process at scale with Evaluate, which works through requests in order. When it cannot approve one it pauses and lets you skip and continue, or stop and review, and can ignore a pending status in future runs. You can export the results as a CSV file.
• Turn on the business unit settings that synchronise payable hours between schedule and request, and that estimate and synchronise start and end times for full-day requests. With the second, publishing updates the request with the actual shift times, preventing conflicts between full-day and partial-day requests and improving limit accuracy.
• Use override dates for peak periods, to block requests or require manual review, and publish them before agents book into those dates.
• Time-off balances can come from an HR system with a JSON REST API through the WFM and HR system integration, built with Architect workflows and data actions.

5.8 Block scheduling: deciding what kind of work, not just who is there

Block scheduling puts planning-group-based work blocks into a shift using non-default On Queue activity codes linked to a specific planning group, so you decide whether an agent blends work types or concentrates on one. It works manually in the schedule and in work plans, where the engine can place blocks and rescheduling is supported. Service level, average speed of answer and occupancy take the blocks into account, and Queue Activation can switch an agent's queues at the start or end of a block. Genesys lists the benefits as closer alignment to demand, control by time of day, less manual editing and less context-switching fatigue, for example separating written from spoken work or keeping quiet periods for deferred work. It was released in October 2025.

5.9 Shift trades, alternative shifts and opportunities

Shift trades let agents exchange complete future shifts within a six-week window that includes the current week. You choose automatic approval or review, a minimum time between approval and shift start (a 24-hour minimum, for example, blocks a trade for a shift starting in ten hours), and matching criteria such as queue, language and skills. Each trade rule, including unequal paid hours, one-sided trades and weekly paid-time violations, can be set to allow, deny or send for review.

Alternative shifts make the business the trading partner: agents browse shifts the system offers, and a request that satisfies the service goal impact you configured is approved automatically. Trades are processed in batches every few minutes; one that no longer meets the impact because of other changes is denied, and a supervisor can override a denial.

Opportunities are structured offers of extra work, such as overtime, extra hours, surge cover or event cover, built on On Queue activity codes with a set number of places, up to 24 hours per opportunity. Agents see opportunities only for times when they have nothing scheduled, and approved requests are added to their schedules automatically. Work plan bids let agents bid for work plans, awarded on criteria you choose, and Genesys Tempo shows agents their schedules, notifies them of changes and shows which days are available for leave and where they are on a waitlist.

6.1 Intraday monitoring

Intraday monitoring compares forecast and actual values per planning group in 15-, 30- or 60-minute intervals, with differences highlighted and red values marking data to investigate. Forecast summaries cover the whole selected range; actual summaries cover the time elapsed. Its stated purpose is to adjust the schedule during the day: add agents to improve service, or send agents home if too many are available.

• It uses the planning group configuration that existed when the forecast was created, so later configuration changes do not show.
• It covers inbound statistics only, and only agents currently in the selected business unit.
• Answered, abandoned and completed percentages are percentages of offered. Forecast averages are weighted by offered; actual averages by completed, except service level (weighted by answered plus abandoned) and average speed of answer (weighted by answered).
• It needs a published, load-based schedule covering the date. Predicted average speed of answer uses only the agents the published schedule says handle the planning group, so make sure the schedule contains everyone who answers that work.
• Service level options, such as including or excluding flow-outs, short abandons and callback wait, are explained in Genesys's workforce management metric definitions.

6.2 Real-time adherence

Real-time adherence compares each agent's current status and presence with the published schedule, so supervisors can act and change the schedule for the rest of the day. Charts show adherence (in, out, unscheduled, ignored or unknown), impact, scheduled activities and actual activities, with a searchable agent list below that includes how long each exception has lasted.

Impact is the column that turns real-time adherence from policing into management. An agent scheduled for a break who is still on queue has positive impact; one scheduled on queue but on an unplanned break has negative impact; one scheduled for training who attends a team meeting is neutral. For scheduled On Queue time, an agent is in adherence when On Queue, or in the Idle, Interacting or Communicating routing status.

• Agents must be in a management unit and on a published schedule, or they have no adherence records.
• Publishing or republishing takes a few minutes to show.
• Moving an agent between management units in the same business unit moves adherence data with them; moving between business units leaves schedule and adherence history with the original unit. Time-off requests follow the agent.

6.3 Adherence thresholds on the management unit

Two numbers are yours to set. The severe adherence threshold is the number of minutes after which an agent counts as severely out of adherence in real time; the default is five minutes. The adherence exception threshold is the minimum length of an exception that appears in historical adherence. Treat the defaults as a starting point, not a policy, and set both with your supervisors so the screen shows what they will act on.

Four more settings on the management unit change every adherence score in it, so decide them once and write the decision down.

6.4 Adherence and conformance

Both formulas are published. Adherence is (1 minus exception duration divided by scheduled duration) times 100, floored at zero: did people do the right activity at the right time? Conformance is on-queue time worked divided by on-queue time scheduled, times 100: did they work the right amount? An agent who works with nothing scheduled shows up in conformance, not adherence.

Adjustments and explanations are different things. An approved adherence adjustment counts the time as in adherence and keeps it in worked time: Genesys's example is an eight-hour shift with 50 minutes out of adherence, 430 of 480 minutes or 89.6%, which becomes 460 of 480 or 95.8% after a 30-minute adjustment. An approved Running Late explanation removes the time from the worked total instead. Agents submit adjustments with a reason code within a look-back period (seven days by default, up to 21), from up to 25 reason codes set on the business unit; Running Late is submitted separately, through Genesys Tempo.

In the schedule editor you can show adherence: a line under scheduled activities marks out-of-adherence time, and you can expand an agent to see the exceptions. Analytics can lag, so the most recent intervals may change as data catches up.

• Agents see their own adherence in an adherence widget in the Activity view (added November 2025): scheduled, actual and exception details for the day.
• Late explanations can currently be submitted only through the Genesys Tempo app; supervisors approve or deny them from the inbox or the schedule editor.
• Treat explanations as data: repeated lateness on the same shift start is a rota design signal, not a conduct problem.

6.5 Intraday reschedule

When the day, or the rest of the week, has moved away from the plan, Genesys can re-optimise it. An intraday reschedule takes manual changes you have made to the short-term forecast and re-optimises shift starts and lengths and activity start times. It is also how you add new hires to a published schedule.

• Reforecast first, then reschedule. Rescheduling against the old forecast only re-optimises the wrong answer; a forecast modification for the rest of the day may be enough.
• Save any manual schedule changes before you run a reschedule.
• Only one reschedule can be active per management unit, and one waiting in Ready to be applied blocks the next until you apply or cancel it.
• Genesys uses the work plan in place when the schedule was first generated; if an agent's work plan has since been replaced by a different one, the new one is not used.
• Agents are notified when the result is published, so reschedule sparingly and not at the last minute.

7.1 Historical adherence

Use historical adherence for trends and for correcting the next plan rather than for one-to-one policing. The view shows adherence and conformance per agent, management unit averages, the target, an adherence-over-time chart you can click into, and exception counts you can drill into. Team reporting is available in the historical adherence and shrinkage views.

7.2 Shrinkage: the number that decides how many people you need

Historical shrinkage records both scheduled and actual shrinkage, so you can refine staffing models and improve scheduling accuracy. Genesys only counts activities whose activity code is marked as shrinkage, which is why activity code governance matters so much.

Unplanned shrinkage, set on the management unit, is the allowance for time scheduled agents will not be available. Genesys's arithmetic is simple: expected agents equal scheduled agents minus scheduled agents times the shrinkage percentage, so 200 scheduled agents at 10% gives 180. Genesys also notes that WFM already accounts for activities in the daily shift, so breaks, meetings and training placed in shifts must not be added to the shrinkage factor as well, or you will count them twice and over-staff.

7.3 Reviewing shrinkage

Shrinkage review is where planning meets reality. Common practice in the industry is to build the shrinkage plan once a year from the forecast, plan holiday inversely to workload so leave is easier to grant in quieter periods, and reconcile planned against actual every month or quarter, starting monthly when the practice is new. Genesys gives you the two numbers to compare: scheduled and actual shrinkage, by category and management unit.

1. Monthly, compare scheduled with actual shrinkage per management unit and by category, by interval band where you can (mornings, afternoons, evenings, weekends).
2. Split the gap into its causes: sickness, late starts and early finishes, meetings that overran, unplanned training, system outages.
3. Decide, for each cause, whether to plan for it (raise the shrinkage assumption on the management unit and accept the cost) or to remove it (an attendance, process or system fix). If actual runs several points above scheduled and the cause is unplanned absence, better scheduling will not fix it.
4. Check the coding before you trust the trend: a new activity code without the shrinkage flag, or a status mapped to the wrong category, moves the number without anything changing on the floor.
5. Record the decision and the new assumption with a date, so next quarter's review starts from facts.

7.4 The weekly cycle, end to end

A typical week for a planner, using only documented features. Adjust the days to your publishing calendar; what matters is that each step happens in order and on a predictable day.

8.1 The daily rhythm

A planning team that runs to a rhythm spends less time firefighting. The morning sets the plan for the day; the day runs to agreed triggers; the evening records what happened.

8.2 An intraday plan agreed in advance

Industry practice is to agree, before the day, an ordered list of actions with the thresholds that trigger each one, how changes are communicated, and a review afterwards. Agreeing it in advance with operations means the planner applies policy rather than negotiating it at 10:15 on a Monday. Avoid knee-jerk changes: a short variance does not always continue, so look at the cause and size of the gap before acting, and make changes incrementally.

8.3 Publishing discipline

When agents see their roster matters to them as much as what it says. Industry surveys suggest most contact centres publish exact shifts between four and eight weeks ahead, with some publishing working days or shift types further out and confirming exact times nearer the date. Genesys Cloud schedules are one to six weeks long, and you can keep several weeks published ahead by generating and publishing each new week on a fixed day.

Late changes are the other half. The platform allows republishing at any time and notifies agents of changes; your policy is what stops churn. A common pattern is a rolling horizon with the nearest weeks locked: changes inside the locked window only for emergencies, and swaps allowed further out through shift trades.

• Publish on a fixed, announced day each week and keep to it.
• Keep a rolling horizon: decide how many weeks ahead agents can see their exact shifts, and do not let it shrink in busy periods.
• Lock the near term. Set an internal cut-off (for example, no changes inside 48 hours, or inside two weeks, without a manager's agreement) and route other changes through trades and opportunities.
• Select agent notifications every time you publish, and use the schedule description to say what changed so supervisors can brief teams without comparing grids.
• Never publish an unreviewed generated schedule: read the generation messages and Scheduled, Required and Performance first.
• Watch UK law. The Employment Rights Act 2025 gives zero-hours and low-hours workers rights to reasonable notice of shifts and of changes or cancellations, and to payment for short-notice cancellations. As of 13 November 2026, Acas says the government has not announced when these changes will happen; regulations will set the notice periods. A disciplined publishing calendar now makes those rules easy to meet when they arrive.

8.4 Triage of schedule generation messages

Every generation produces messages, and the temptation is to ignore them or to hand-edit around them every week. Instead, set each message's severity once on the business unit so the list shows what matters, and treat any message that recurs as a fault in configuration rather than a weekly chore. Genesys lets you set Error, Warning, Information or Ignore per message; the defaults vary from message to message, so review them when you configure the business unit.

8.5 Activity code governance

Activity codes decide pay, occupancy, adherence and shrinkage, so a code added casually by a supervisor can change a month of reporting. Industry guidance is simple: list every shrinkage code in the WFM system and group them, for example into out-of-office time (absence, holiday), discretionary time (coaching, training, meetings) and administrative time (outages, set-up, projects), with operations agreeing the groupings.

• Own the list. One named owner in the planning team approves new codes and changes; supervisors request, they do not create. Use permissions to enforce it.
• Agree the flags in writing with Finance and HR before a code goes live: category, paid, work time, interruptible, shrinkage planned or unplanned, and agent selectable.
• Name codes consistently so reports group naturally, for example by prefix for training, meetings and absence types, and describe each code's purpose.
• Do not change the flags of a code in use. Create a new code with an effective date and retire the old one, so history keeps its meaning.
• Review the list quarterly: unused codes, codes used for the wrong thing, and codes whose shrinkage flag no longer matches how the business treats the time.
• Map presence statuses deliberately. Each category sets a primary presence; if agents cannot select the status the schedule expects, it will look like non-adherence.

8.6 The monthly and quarterly reviews

9.1 Structure

1. Design around shared queues, not the organisation chart. Keep agents who handle the same queue in one business unit; use several management units inside it for permissions or local management, and schedule the whole business unit together.
2. Keep load separate between business units. If you cannot, accept the weekly manual forecast adjustment Genesys says you will need.
3. One route path per planning group by default. Combine only low-volume paths, and only where agents handle the whole combination; check the Agents tab.
4. Add management units for permissions rather than business units, and plan a division for each distinct access-control building block.
5. Set the start day of the week and the time zone once, deliberately. They cannot be changed.

9.2 Forecasting

1. Default to Automatic Best Method.
2. In a new organisation, import up to three years of history on day one, allowing for the 24-hour import and building the route paths first.
3. Let the nightly continuous forecast inform, not drive. Schedules come from the short-term forecast you created and enabled.
4. Modify sparingly; name, date and review every modification. Prefer one daily values change to many interval overrides.
5. Never change a service goal template and generate a schedule on the same day.
6. Measure forecast error by interval and planning group, and publish it.

9.3 Scheduling

1. Encode constraints rather than hand-editing around them. A recurring generation message is a configuration fault.
2. Allow for run time: 20 minutes or more per forecast-based schedule, one generation at a time per business unit.
3. Do not double-count shrinkage: what the shifts contain is already allowed for.
4. Read Scheduled, Required and Performance before you touch a shift, remembering that overlapping unpublished schedules can fake a gap.
5. Publish to a fixed calendar with a rolling horizon, and lock the near term.
6. Put recurring meetings, training and compliance into activity plans.
7. Use block scheduling for real differences in the kind of work, not for hourly control.
8. Use opportunities for overtime and surge cover, so extra hours are capped and land on schedules automatically.

9.4 Time off and self-service

1. Set limits per staffing group where allowances differ.
2. Make Conditional Approval the default. Auto Approval approves beyond the remaining limit, so use it only as a deliberate exception.
3. Turn on start and end time estimation and synchronisation for full-day requests.
4. Explain waitlists to agents: requests approve automatically if capacity frees up before the window closes.
5. Publish override dates for peak periods before agents book into them.
6. Set shift trade rules explicitly, including minimum notice before shift start, and keep trades within the same skills.
7. Treat supervisor overrides of alternative shift denials as a documented exception.

9.5 Running the day

1. Separate the drivers (volume, handle time, staffing) before choosing an action.
2. Reforecast, then reschedule, and save manual edits first.
3. Watch impact, not just out-of-adherence: someone working through a break is not a coaching case.
4. Set adherence thresholds with supervisors; five minutes is a default, not a policy.
5. Make sure the published schedule includes everyone who answers the work.
6. Log every intervention; it is the only honest evidence for next quarter's assumptions.

9.6 People and data

1. Agree definitions in writing with Finance, HR and operations: service level, shrinkage categories and what counts as work time. Most WFM disputes are about definitions.
2. Check the licence before promising a feature.
3. Confirm your organisation's real limits with Genesys before designing near them.
4. Use monitoring under a clear policy that agents know about, and within UK and European rules.
5. Make self-service the engagement argument: visible schedules, notified changes, transparent waitlists and fair trading are what make rigorous adherence acceptable.
6. Check capability questions against Genesys's own documentation before relying on them; it changes with every release.

10.1 Where things are

10.2 Numbers to remember

10.3 Scheduling more agents than concurrent licences

You may schedule more agents than you hold concurrent licences, provided they have the right permissions: Genesys's example is 500 agents against 250 concurrent licences, 250 in the morning and 220 in the afternoon. Billing is based on the concurrent peak of users of workforce engagement features in the month, and concurrency is measured on when users are online, not when they use a feature. This matters when you decide who gets WFM access.