Usage reports and credit cycles in Tigy: comparing consumption

Record intervals

Write report start and end alongside the cycle shown in billing. A calendar month and a cycle starting mid-month are not equivalent. Check displayed time conventions to avoid excluding calls near day boundaries.

Check the agent set

Verify workspace, filters and included resources. A single-agent view should not be compared directly with workspace totals. Identify voice tests and other consumption-bearing executions through available records.

Separate activity from outcomes

Calls, duration and credits describe activity or consumption. Sales, bookings and resolved requests need evidence from responsible systems. Do not divide broad consumption by a smaller outcome sample without explaining the mismatch.

Review financial state

Consult billing for cycle, balance and account conditions. Editorial tables are not the source for current pricing. This procedure promises no fixed concurrency, allowance or limit; verify the conditions actually displayed.

Document discrepancies

If equal intervals and filters still differ, record parameters, times and execution references for investigation. Share necessary information without credentials. Preserve the comparison so analysis can be repeated after adjustments.

Why can reports and billing cycles show different figures?

A report spanning the first through the thirtieth can include activity outside a cycle starting on the fifteenth. Agent filters, time zones and tests also change the compared set. Record each view’s boundaries and resources first; an initial discrepancy does not establish duplicate charging.

After aligning scope, use available runs and transactions to investigate examples. Credits are the billing unit, without a universal conversion to minutes or calls in this guide. To evaluate business outcomes, compare consumption and confirmed tasks within the same period, separately from financial reconciliation.

Start with the question each screen answers

Runs, Reports, and Billing present different views of activity. Runs helps investigate an individual conversation. Reports aggregates activity within a selected range. Billing shows the workspace's cycle, credits, and billing-related records. Comparing those areas without aligning scope may create an apparent discrepancy that is not an error. First choose the question: what happened in a call, how much activity occurred during a period, or how usage appears within a cycle.

In a fictional workspace, staff open a calendar-month report and compare its credits with a cycle starting on the tenth. The intervals differ. Calls from the month's first days may belong to a previous cycle, while days in the following month may fall within the current cycle. Do not classify the difference as incorrect billing before checking dates, workspace, and filters.

To investigate one call, locate the run by agent and time. Review duration and available usage details alongside transcript and actions. For trends, use the report's period and filters. For cycle balance, open Billing and inspect subscription state and transactions. Choosing the appropriate area prevents attempts to reconstruct individual information from aggregated totals. Each area supports a different part of the investigation.

Record comparison scope in a simple note: workspace, agents, interval, time zone, and inspection time. Another person can then reproduce the analysis. A screenshot showing one isolated number may omit the filter explaining the difference. Starting from question and scope reduces premature interpretations and guides the next check when values genuinely remain inconsistent. The note should describe the values observed without claiming a cause that has not yet been established.

Align period, time zone, and agent selection

A call near a date boundary may appear on different days depending on the time zone used for comparison. A report filtered to one agent does not equal the workspace total. A weekly interval does not represent an entire billing cycle. These differences are simple but can appear complicated when an analyst sees only the final value. Check displayed period and filters before adding data or attributing a change to a new agent version.

In a fictional example, staff compare a week of sales-agent activity with a previous week that also included the support agent. Volume and credits changed, but the agent sets are not equivalent. The comparison cannot establish that the sales agent became more expensive. Repeat with the same set, then investigate duration, request types, and outcomes. Changes in composition should be separated from changes in performance.

For Billing reconciliation, use displayed cycle boundaries rather than assuming the first day of a month. Verify the correct workspace, especially when staff use several environments. Tests may generate usage as well; being authenticated does not mean every call is free or that the account permits another run. Include testing activity when it appears in the relevant records. Do not dismiss it simply because the conversation was not with a customer.

If an interval looks empty, confirm activity for the selected agent and data availability. An empty report alone does not prove collection failed. Find corresponding calls in Runs and check scope. This sequence avoids changing the agent when the cause is a filter, period, or workspace. Once those elements align, a persistent discrepancy becomes more concrete and easier to investigate with reproducible evidence.

Read balance and transactions without inferring a new charge

Billing presents the cycle and credit transaction history when available for the workspace configuration. These records help explain how usage is represented, but should be read alongside subscription state and period. A pending payment does not confirm activation of a new plan. An attempted plan change does not prove the change is effective. Confirmed state should guide analysis.

In a fictional scenario, an administrator began a change and expects a different credit allowance. Before treating the balance as an error, check whether payment and activation were confirmed. Documentation describes the change as dependent on payment confirmation. Selecting another plan is not evidence that the cycle has already changed. Checkout and Billing should show applicable values and states for the case. A request in progress remains different from an active subscription condition.

Avoid fixing prices or usage conversions in an informal spreadsheet without checking the current condition shown to the workspace. Operational analysis may track displayed credits and outcomes, while a subscription decision needs to review values in the billing process. Maintaining this distinction reduces the chance of treating an internal estimate as a confirmed commercial term. Do not infer a payment amount solely from a report total when the relevant cycle and plan state have not been checked.

If discrepancies persist, retain related conversation and transaction references. Do not send credentials with evidence. Record observations, period, and comparison steps. An objective description supports investigation without presuming duplicate charges or measurement failure before verification. Reconciliation should explain how values relate within the same scope; payment or plan-change decisions should use confirmed state and values presented in the corresponding process. This creates a concrete question for support rather than a collection of totals drawn from incompatible intervals.

Compare usage with outcomes to assess the operation

More consumed credits may reflect more calls, longer conversations, or a changed service mix. Lower usage may represent efficiency, reduced volume, or premature endings. The number alone does not establish quality. Compare activity with observable outcomes such as correct answers, accepted records, and appropriate routing according to the agent's objective. Obtain outcome evidence from conversations and the relevant system.

At a fictional company, an agent began confirming identifiers before order lookup. Duration might increase slightly while lookup errors decline. Evaluating only time reduction would penalize a useful correction. In another scenario, calls became shorter because the agent stopped asking a necessary question; usage fell but staff had to collect the information again. Interpret usage alongside quality and rework rather than assuming every decrease is an improvement.

When comparing versions, preserve agent selection and request profile. A week of simple questions is not a fair reference for another containing many complex individual cases. Separate categories and inspect a sample. Do not attribute every credit change to the model or prompt without checking service composition and corresponding records. The comparison should make alternative explanations visible rather than selecting the most convenient cause.

An analysis may relate displayed credits to demonstrably completed tasks within the same period when task definitions are clear. Do not call every ended conversation a resolution. A recorded request may be the agent's correct outcome even when staff perform the work later. An informational answer may resolve a question without a write operation. Useful comparison respects those differences and shows where usage supports real work, instead of producing one ratio that combines incompatible objectives. Keep the calculation reproducible and its practical limits explicit when sharing it with the team.

Create a short reconciliation and investigation routine

A periodic review can begin with the cycle in Billing, continue through totals for the corresponding period in Reports, and use Runs for examples. Reviewing every conversation is unnecessary to detect a scope mismatch. First confirm boundaries, filters, and workspace. Then choose calls explaining the largest deviation or a category behaving unexpectedly.

Keep a note with period, agent selection, observed values, and references for reviewed examples. Someone taking over operations can reproduce the analysis and distinguish actual changes from filter changes. Avoid copying complete transcripts when an identifier and minimal excerpt are enough. Evidence access should follow organizational rules. The note should support investigation without unnecessarily broadening access to customer information.

When scope does not explain a difference, investigate data availability and conversation states. Not every execution offers identical details, and text chat produces no voice recording. For audio questions, choose a voice run with an available recording. For a tool action, inspect parameters and results where present. Evidence should match the question being investigated. Missing recording from a text interaction is expected rather than proof that the reporting system lost audio.

Turn findings into specific decisions: correct a filter, review a category, revise instructions, investigate integration, or request support with evidence. Do not change an agent simply because different totals appeared on separate screens. Explain the difference before changing operations. After a correction, compare the same scope again and inspect several outcomes. This procedure relates usage, quality, and billing without conflating them, helping staff follow service growth through reproducible information. Keep unresolved questions separate from confirmed findings so a later review can continue from clear evidence instead of repeating assumptions about what the totals must mean.