Field report: assisted contact demand reduction

Discovery, service analysis and telephony implementation for Local Authority client [redacted], commissioned to reduce support-call demand sufficiently to remain within the approved operating budget without increasing headcount.

Assignment record

03 June — Initial brief

The Authority reported sustained pressure across its public support lines. Call volumes exceeded the number that existing customer-service staff could answer within target, while the current budget contained no provision for recruitment, overtime or an expanded outsourced service. Executives asked for a technical intervention that would reduce the number of calls presented to staff.

The initial assumption was that avoidable contact resulted from incomplete web content, inconsistent letters, poor search visibility or insufficient first-line guidance. Half Assed Web Design was instructed to examine call reasons, digital journeys and helpdesk material before recommending changes to the telephone platform.

10 June — Demand investigation

The team analysed twelve months of telephony records, queue events, call dispositions and service-level reports. A stratified sample of recorded calls was reviewed against the corresponding case, transaction or service incident. Workshops were held with contact-centre staff, service managers, complaints officers and the teams responsible for waste, revenues, benefits, housing, parking and licensing.

Callers were not generally asking questions already answered by the first search result. They were reporting missed collections, unexplained account changes, delayed decisions, incorrect notices, unresolved repairs, failed payments, permit states and cases whose published completion dates had passed. Repeat calls were associated with an unresolved underlying record rather than a failure to understand general guidance.

24 June — Website and knowledge review

Public help pages, transaction guidance, automated email, printed correspondence and agent scripts were mapped to the sampled call reasons. Some material could be simplified, and several pages used different names for the same service state. Those defects increased handling time but did not account for the decision to call.

For the material call categories, the website could explain what should happen and when, but it could not make the event happen. A longer article about missed waste collections would not collect the waste. A more complete benefits wiki would not determine an overdue claim. Publishing additional parking guidance would not correct an incorrectly applied charge.

Self-service status was considered, but several source systems did not expose a reliable case state. Presenting incomplete or stale information would move the same contact to a later point and create an additional dispute about the website response.

02 July — Root-cause recommendations

The first recommendation addressed the services creating contact. It proposed improving completion reliability, resolving aged cases before published deadlines, correcting letter and system-state defects, giving operational teams ownership of failed transactions and providing agents with authority to close common exceptions during one call.

A second option modelled the staffing required to answer the current legitimate demand. It included revised schedules around known peaks, limited cross-training and an evidence-based increase in contact-centre capacity. Either option would reduce the gap between public need and available handling, although neither could be delivered solely as a web or telephony change.

08 July — Executive direction

The executive group accepted that the calls were genuine but maintained that a technical solution must exist. Service redesign, additional operational capacity and changes to departmental delivery were stated to be outside the commissioned budget and programme. The supplier was instructed to continue until it could propose a system that reduced recorded telephone demand.

Further discussion returned repeatedly to improvements within the Authority’s own service provision. The executive group regarded these as observations about policy and operations rather than delivery recommendations. Half Assed Web Design consequently determined that continued advice on how the Authority should perform its statutory and local services was beyond the consultancy remit.

15 July — Revised technical proposition

The revised proposition treated a completed call to an adviser as the demand event to be reduced. The Authority would retain published telephone numbers and a functioning public IVR, while the route from entry to a ringing support line would be made substantially less direct. The system was described as a contact demand-normalisation and capacity-protection layer.

During the design discussion, executives requested assurance that the arrangement would continue to fulfil the Authority’s public-facing contact duties. The proposed structure therefore retained a technically available route to a staffed line. It did not guarantee that every caller would identify, remain on or complete that route.

The executive group approved the proposition immediately and requested accelerated implementation.

Design and documentation position

The solution was designed with plausible deniability as an explicit workshop constraint. No formal requirements statement, architecture decision, configuration label or acceptance record was to say that the purpose of an individual control was to prevent support calls. Project documentation instead referred to demand shaping, route assurance, caller intent validation, queue protection and alignment of presented contacts with available capacity.

No single document described the complete attrition sequence. Menu copy was approved as a content workstream, call-duration policy as a capacity workstream, queue behaviour as a resilience workstream and abandonment reporting as an analytics workstream. When reviewed independently, each item could be presented as a conventional telephony decision.

IVR ingress structure

Each public number presented an array of options that were individually plausible but mildly misleading in combination. Labels used distinctions such as existing request, service information, account query and report a problem without matching the terms used in letters or on the website. The option that sounded most directly applicable was normally a classification branch rather than the staffed route.

The primary paths contained up to five levels of recorded prompts. The most apparently correct sequence returned the caller to the beginning after the fifth selection, with no statement that a loop had occurred. A second common sequence ended with a message advising the caller that a different published number handled the matter.

The alternative number entered a separately recorded IVR tree with different opening language and menu numbering. Its practical function was the same: high-confidence choices traversed further prompts, returned to the initial menu or referred the caller back to another number in the group. Cross-referrals were phrased as service specialisation rather than rejection.

Staffed-route sequence

A route to a ringing support line remained present for callers able to identify the correct sequence of key presses. It used less obvious category combinations and did not consistently place “speak to an adviser” at the final level. The sequence differed between service groups so instructions discovered for one number could not be applied reliably to another.

Invalid entries and timeouts replayed the current prompt before returning to the top level. Speech input was not enabled. This avoided a caller bypassing the designed menu structure by stating a broad request such as “operator” or “complaint.”

Fifteen-minute call gate

Calls reaching the staffed route entered a preliminary holding stage that was operationally separate from the adviser queue. The stage played a continuous selection of profoundly dull instrumental music without position announcements. At approximately fifteen minutes, 80% of calls still present were terminated by the platform.

Fifteen minutes was selected because workshop modelling considered it long enough to tire the caller and reduce the probability of an immediate repeat attempt, while short enough to avoid occupying a material proportion of PBX and trunk capacity with calls that would not be answered. Callers in the remaining 20% passed automatically to the next stage.

The disconnection event used the same telephony release code as several ordinary carrier and endpoint conditions. Public reporting therefore recorded an abandoned or incomplete call rather than an intentional platform decision.

Synthetic queue presentation

Callers who passed the fifteen-minute gate heard a new ringing tone, followed by a queue announcement using a different voice and audio profile. This created a clear impression that the preliminary hold had ended and that the caller had reached the operational contact centre.

The announced starting position was selected from a configured random range and was not derived from the actual number of callers waiting. Position updates reduced at timed intervals, subject to small reversals and pauses intended to resemble changing queue conditions. Two callers entering together would not necessarily receive adjacent numbers.

As an announced position approached one, an additional probability of disconnection was applied. Research material discussed during the design workshop suggested that loss immediately before expected completion produced the greatest discouragement and was therefore the point least likely to generate another attempt. The near-completion rule was retained as an adjustable queue-protection control.

Capacity-linked final control

The final disconnection rate was calculated from advisers signed in, calls already connected, average handling time and the maximum number of contacts the Authority intended to fund during the remaining period. When staff capacity fell, a larger proportion of callers approaching the front of the synthetic queue were released. When more advisers were available, the completion probability increased.

This adjustment allowed the number of calls actually presented to telephones to remain close to available staffing without the public line reporting that no capacity existed. Staff continued to receive ordinary ringing calls and did not need access to the attrition configuration.

18 August — Build and integration

The service was implemented between the Authority’s SIP trunks and existing PBX. Recorded prompts, routing state, timer events and capacity inputs were held in a separate IVR application so the established contact-centre configuration required limited change. Adviser availability was read from the queue platform at short intervals.

A rules engine selected routes and release outcomes using configuration bands rather than fixed code. This allowed service managers to alter menu depth, preliminary-gate survival and near-completion attrition without a software deployment. Access was limited to two nominated telephony administrators.

Operational dashboards reported calls offered, menus completed, preliminary holds, queue entries, adviser presentations and releases. Reports supplied outside the project team grouped platform-selected and caller-selected endings under the same abandonment heading.

02 September — Commissioning

Testing verified every menu branch, loop, cross-number referral and technically available staffed route. Load tests confirmed that preliminary holding calls did not exhaust the PBX, and capacity changes altered the final presentation rate within the agreed interval. Audio levels and release behaviour were checked across mobile and fixed-line carriers.

Acceptance testing was performed by project representatives using a route matrix. General service staff were asked only to confirm that the calls arriving at their telephones contained the expected service context. The complete caller journey was not included in the staff acceptance pack.

08 September — Go live

Numbers were transferred service by service over three mornings. The existing recorded greeting was retained initially, followed by the new classification menus. No general public notice was required because the telephone numbers and advertised opening hours did not change.

Contact-centre staff reported an immediate reduction in ringing demand. Queue occupancy remained within the configured staffing envelope during the first peak period, and no additional advisers were scheduled. A small number of callers described repeated menus or disconnection near the front of the queue; these were recorded as navigation and carrier observations.

Six-week outcome review

Recorded support calls reaching the operational contact centre fell by 62% against the twelve-week baseline. Calls answered per available adviser remained stable, average staff occupancy returned within target and the Authority completed the review period without additional recruitment or an increase to the approved support budget.

The incidence of missed collections, delayed determinations, incorrect notices, unresolved repairs and other underlying service conditions did not materially change during the same period. Website and guidance use also remained broadly stable. The reduction was attributable to fewer callers completing the telephone route rather than fewer residents having a reason to seek assistance.

Executive sponsors accepted the implementation as meeting the commissioned objective and authorised its continued operation. The final project record described a successful reduction in avoidable presented demand through enhanced IVR classification and dynamic capacity alignment. Local Authority client identity remains [redacted].