From 200+ mailboxes to one governed service model – what is it worth?

One page, two numbers. Today 1.3 M e-mail threads take about 10 minutes each. With S2P, fewer threads arrive (suppliers self-serve on the Supplier Hub) and each remaining thread takes fewer minutes (routed case, one screen, AI assist). Set those two numbers, pick a scenario, and mirror the result against the ServiceNow value case and peer benchmarks. Save a case with name, version and description, or duplicate it to model a variant.

CaseUnsaved working case

Scenario

Case settings

Organisation / unit

CurrencyUsed for all values and charts. Cost per FTE, licences and one-time cost are entered in this currency.

Baseline – what we handle today

One question: how much working time goes into AP mailbox traffic today? Threads × minutes ÷ hours per person = people. Nothing here assumes any change yet.

E-mail threads p.a. (AP mailboxes)Fact. A thread is one conversation (first e-mail plus all replies and forwards). AP Mailbox volume analysis, WD p. 57: 1,314,322 threads = 2,246,674 e-mails, i.e. 1.7 e-mails per thread.

Handling time per thread, end-to-end (min)

Assumption. All time spent on one conversation: reading, finding the invoice or supplier in S/4, answering, forwarding, chasing. ≈ 6 min per e-mail × 1.7 e-mails. The whole case scales with this number (6 min → 60 % of the FTE shown). To be confirmed by AP Operations.

Query threads p.a. (internal + external)Fact. The threads that are questions – “where is my payment”, “please update our bank details” – rather than invoices or documents. 204,169 internal + 159,828 external (WD p. 57). Only questions can move to self-service; used to translate “fewer threads” into a share of questions that peers can be compared with.

Not counted in this case. 1.78 M invoices a year (e-invoicing project, own business case) and ≈ 200k S2P / D2P tickets (later wave, WD p. 59). This case is about the 1.3 M e-mail threads only.

What changes with S2P – two numbers

Everything the solution does ends up in one of these two numbers. Scenario buttons above set both; move either to see your own case.

Fewer threads – share of all threads that never arrive (%)

7 %

Faster per thread – minutes per remaining thread with S2P

6.5 min

Expert view – how the two numbers are composed

The same two numbers, per scenario, with their derivation. “Fewer threads” is easier to judge as a share of query threads (that is what BASF and GSK report); the minutes saved split into what routing removes before anyone works the case and what the agent gains on the case itself. Edit any cell – the rest follows.

	Conservative	Realistic	Optimistic	Custom

Why these values. Peers with a supplier portal on a case backbone report 50–60 % fewer supplier inquiries (GSK, BASF); we plan for 18–32 % of our query threads, i.e. 5–9 % of all threads. BASF reports 2 minutes saved per case from routing alone and 50 % faster resolution; IOFM puts inquiry handling at 5–20 minutes. We plan 7 / 6.5 / 5.5 minutes instead of 10.

Cost and ramp

Turns released time into money and spreads benefit and cost over three years. Released time is capacity that can be redeployed or not re-hired – not a cash saving by itself.

Loaded cost per FTE (CHF p.a.)

Productive hours per FTE p.a.

FTE in scope (RSS)

Fulfiller licences at full build-out (CHF p.a.)450 fulfiller (agent) licences at full build-out, ServiceNow estimate. Suppliers and requesters need no licence.

One-time implementation (CHF, tbd)Assumption. Design, build, integration to S/4, data migration and change. Working assumption 3× annual licence cost – to be scoped with ServiceNow and the implementation partner.

AI consumption (CHF p.a., placeholder)

Benefit ramp Y1 %

Y2 %

Y3 %

Benefit ramp: how much of the full-year value arrives in each year (Y1 build and first waves, Y2 broad rollout, Y3 steady state). Licence ramp: licences grow with the rollout.

Licence ramp Y1 %

Y2 %

Y3 %

Mirror & benchmarks

Two outside reference points to check whether the bottom-up result is plausible: ServiceNow’s own value case and what peer AP / S2P support organisations with a supplier portal achieved in self-service deflection.

ServiceNow value case, year 3 run-rate (CHF p.a.)First pass with ServiceNow's value case analysis: ≈ 3.2 M in year 3. Detail requested.

ServiceNow value case, FTE (year 3)

Peer deflection low – source

% of query threads deflected

Peer deflection high – source

% of query threads deflected

GSK: −50 % supplier calls within 10 months of online invoice-status check (Tungsten Network). BASF: supplier invoice inquiries > 500,000 to < 200,000 p.a. (−60 %) after replacing hundreds of mailboxes with one portal on ServiceNow. Applied to the query threads × minutes per thread ÷ hours per FTE.

Baseline tied up in e-mail handling

–

Capacity released (run-rate)

–

Value p.a. at run-rate

–

Net benefit p.a. after licence & AI

–

Payback on one-time cost

–

Cumulative net, 3 years

–

The business case on one page

FTE released by scenario – two parts

Fewer threads and faster handling add up to the capacity released. Your current setting is “Custom”; the three presets are the deck scenarios.

Three-year view: benefit ramp vs. cost

Benefits ramp with rollout waves; licences are staged; one-time implementation lands in year 1.

Plausibility check – one corridor

Where the bottom-up model lands against the ServiceNow value case, and how much “fewer threads” alone would release if suppliers self-served as much as at peers (GSK 50 %, BASF 60 % of inquiries).

Scenario table

Per scenario	Conservative	Realistic	Optimistic	Custom

Curious what suppliers would see? (demo data).

What a supplier can do on the Hub today – without writing an e-mail

A click-through of the Supplier Self-Service (Supplier Hub) on ServiceNow S2P, entered from suppliers.roche.com. Demo data only. Every action here is one less thread in an AP mailbox – and one more governed case with owner, status and audit trail.

Where is my invoice?

Deflects ≈ 18,000 status asks + 25,700 dunning e-mails p.a. in the sample

9001234900123590012369001237

Today

Supplier writes to one of 200+ mailboxes. A colleague opens SAP, looks it up, answers by hand – 15,828 times a year in 16 mailboxes alone. No record, no SLA.

On the Hub

The supplier is authenticated, the status comes live from S/4 via governed integration. If something is wrong, one click opens a case with the invoice already attached.

Submit an invoice

Structured intake instead of 294,000 invoice e-mails

Today

PDF by e-mail to an AP mailbox, forwarded, printed or re-keyed. Missing PO? Another e-mail. 41,695 “missing information” chases a year.

On the Hub

PO reference mandatory, duplicates and amounts checked on entry, e-invoice ready for the 2026–28 mandates. The supplier sees the case number immediately.

My cases

Owner · status · SLA · audit trail – for every request

Case	Type	Subject	Status	SLA	Owner

Raise a new case

Today

Threads in Outlook. Who owns it? Which status? Was it answered twice? Nobody knows – and nothing can be audited.

On the Hub

Every case has a number, an owner, an SLA clock and a history. Suppliers and colleagues see the same record. AI drafts inside the case, never outside it.

Company & bank details

Ends bank-detail changes by e-mail – 1,201 in the sample

Registered data (MDG)

Legal name: Müller Bauteile AG
VAT ID: CHE-123.456.789 MWST
IBAN: CH93 0076 2011 6238 5295 7
Payment terms: 30 days net
Last verified: 12 Mar 2026

Request a change

New IBAN

Bank confirmation letter

Upload document

Today

“Please update our bank account” arrives by e-mail from an unverified sender. Fraud exposure, SOX finding, manual MDG change.

On the Hub

Authenticated user, document required, four-eyes approval in the case, change written to MDG with full history. Callback verification triggered automatically.

Open items & reminders

Replaces 25,736 dunning e-mails p.a. in the sample

Invoice	Amount	Due	Status	Expected payment
9001231	CHF 8,900.00	28 Aug 2026	Paid 27 Aug	–
9001234	CHF 12,450.00	14 Sep 2026	Approved for payment	Payment run 11 Sep
9001235	CHF 3,120.00	20 Sep 2026	Blocked – GR missing	After goods receipt
9001236	CHF 640.00	02 Sep 2026	Overdue – in clarification	Case S2P0001042

Today

Supplier sends a reminder to three mailboxes. Three colleagues look it up. One answers. Repeat next month.

On the Hub

The supplier sees every open item with the real S/4 status and expected payment date. Reminders become unnecessary – or become a case with an owner.

Saved value cases

Every saved case keeps its full set of inputs. Load one to continue, duplicate it to model a variant, or compare the headline numbers side by side. Shared by everyone who has the access password.

Cmp	Name	Version	Description	FTE released	Value p.a.	Net p.a.	Updated

Tick two cases to compare them side by side.

Assumptions, method and sources

The model is deliberately simple so that every number can be challenged in the room. Yellow items are working assumptions still to be confirmed.

How the model calculates – in plain words

One sentence: count the threads, multiply by the minutes each one costs, turn that into people – then do the same with fewer threads and fewer minutes. The difference is the capacity released.

Today. 1.3 M e-mail threads × 10 minutes ÷ 1,408 hours = ≈ 156 FTE absorbed by mailbox work, about 30 % of RSS. Fact × assumption.
Fewer threads. Suppliers and requesters answer standard questions themselves on the Supplier Hub – payment status, dunning, bank details, proxy requests. Realistic: 7 % of all threads (≈ 92,000) never arrive. That is one in four of the 364k query threads; peers reached one in two.
Faster per thread. The remaining threads arrive as routed cases with owner and data, and the agent works them on one screen with AI assist. Realistic: 6.5 instead of 10 minutes. Expert view: ≈ 2 minutes from intake & routing (BASF: 2 min per case), ≈ 1.5 minutes from agent productivity and Now Assist.
With S2P. 1.22 M threads × 6.5 minutes ÷ 1,408 hours = ≈ 94 FTE. Released: 156 − 94 = ≈ 62 FTE.
Money. FTE released × CHF 55,000 = value; minus licences and AI consumption = net run-rate. Benefit and licences ramp up over three years; implementation is paid once in year 1. Payback = the month in which cumulative net turns positive.
Plausibility. The result is placed next to ServiceNow’s own value case (≈ 58 FTE) and next to peer evidence: BASF and GSK deflected 50–60 % of supplier inquiries with a portal on a case backbone.
Capacity is released, not cut. The case is about redeploying capacity to exceptions, controls and e-invoicing readiness.

Why only two numbers?

Every effect of the solution shows up either as a thread that does not arrive or as a thread that takes less time. Self-service is the first; routing, one-screen case handling and AI assist are the second. Two numbers can be challenged in the room, remembered after the meeting and checked in six months against the mailbox statistics and the case timer.

What is not in this case

1.78 M invoices a year – the e-invoicing programme has its own business case; S2P provides the intake and the status view, not the capture.
≈ 200k S2P / D2P tickets – today on ITSM forms; a later wave once the AP mailboxes are on the platform (volume to be confirmed, WD p. 59).
Risk and control value – bank-detail fraud, SOX findings, audit trail: real, but not counted in FTE.

How to read the inputs

Fact – counted in the AP mailbox analysis or quoted by ServiceNow; not to be debated, only re-measured.
Assumption – our working value, marked as such next to the input; every assumption names who confirms it.
Benchmark – published reference used to choose the conservative / realistic / optimistic values.

Calculation schemes

Exactly what the calculator does, step by step. Symbols: T threads p.a., m₀ min per thread today, Q query threads, d share of threads that never arrive, m min per remaining thread with S2P, h productive hours per FTE, c loaded cost per FTE, L licences p.a., A AI consumption p.a., I one-time implementation, r_y benefit ramp, l_y licence ramp, b peer self-service share.

Today

H₀ = T × m₀ ÷ 60 · FTE₀ = H₀ ÷ h

Fewer threads

Gone G = T × d · FTE_fewer = G × m₀ ÷ 60 ÷ h · as share of query threads: G ÷ Q

Threads that never arrive save their whole handling time. The share of query threads is what peers report (BASF −60 %, GSK −50 %) and is the number to compare with.

Faster per thread

FTE_faster = (T − G) × (m₀ − m) ÷ 60 ÷ h · m₀ − m = Δ_routing + Δ_agent

The minutes saved split into what structured intake and routing remove before anyone works the case (Δ_routing) and what the agent gains on one screen with AI assist (Δ_agent). Expert view only; the model needs just m.

With S2P and released capacity

H₁ = (T − G) × m ÷ 60 · FTE₁ = H₁ ÷ h
Released FTE = FTE₀ − FTE₁ = FTE_fewer + FTE_faster

Run-rate value

Value V = FTE × c · Net N = V − L − A
Break-even capacity = (L + A) ÷ c · Share of baseline = FTE ÷ FTE₀ · Share of organisation = FTE ÷ FTE in scope

Three-year view

Benefit_y = V × r_y · Cost_y = (L + A) × l_y + I (year 1 only)
Net_y = Benefit_y − Cost_y · Cumulative = Σ Net_y

Payback

first month k with −I + Σ_{j ≤ k} (V × r_y(j) − (L + A) × l_y(j)) ÷ 12 ≥ 0

Monthly accumulation using the ramp of the year the month falls in; “> 5 years” if not reached within 60 months.

Mirror and corridor

ServiceNow FTE equivalent = ServiceNow value ÷ c · Peer corridor = Q × [b_low, b_high] × m₀ ÷ 60 ÷ h

b = share of inquiries peers moved to self-service (GSK 50 %, BASF 60 %). The three scenarios, the ServiceNow value case and the peer corridor are plotted on one FTE axis; the deviation shown is (V − ServiceNow value) ÷ ServiceNow value.

Worked example – your current case

Updates live with the inputs and the two numbers on the Value case tab.

Working assumptions to confirm

AP Operations Handling time per thread (≈ 10 min = 6 min per e-mail × 1.7 e-mails per thread); FTE baseline
S2P process owners ≈ 200k S2P / D2P tickets p.a. and their breakdown (WD p. 59) – for the later wave, not in this case
ServiceNow Value case assumptions, headcount and rates behind the ≈ CHF 3.2 M year-3 figure
Programme lead One-time implementation estimate, AI consumption model, licence staging per wave

Sources

Volumes	AP Mailbox volume analysis – RSS Working Doc v7, p. 57 (1,314,322 threads; 2,246,674 e-mails; 912,971 invoice receipt / 204,169 internal / 159,828 external query threads). Common Mailbox Analysis, 16 of 200+ mailboxes, 637,473 classified e-mails, Jun 2025 – Jun 2026 (p. 50–56). E-invoicing volumes (p. 61).
Licence	ServiceNow estimate: 450 fulfiller licences ≈ CHF 498,000 p.a. at full build-out, staged by wave. RSS FTE assumption CHF 55,000 p.a., 1,408 productive hours.
Peer: BASF	ServiceNow, “Self-service transforms supplier management at BASF” (2023): > 500,000 supplier invoice inquiries p.a. across hundreds of mailboxes, 600 employees in three service hubs; after portal + case management on ServiceNow: < 200,000 inquiries (−60 %), tickets −60 %, resolution 50 % faster, +10 % end-to-end efficiency, 2 minutes saved per case. Also: case study PDF.
Peer: GSK, Elanco	Hackett Group, 2026 Innovation Awards: GSK EMEA/APAC Procure-to-Pay Service Desk “HelpHub” – $4.03 M (£3 M) saved in three years, waiting times −40 %, first-contact resolution 88 to 93 %, ROI > 150 %. Elanco – 30,000 P2P queries p.a. at > 10 minutes each, 30–40 % of manual queries eliminated. Tungsten Network, GSK case study (2020): 70 % of supplier enquiries concerned outstanding invoices; −50 % supplier calls within 10 months of online invoice-status check.
AP effort share	Ardent Partners, AP Metrics that Matter 2025: AP staff spend 21.8 % of their time on supplier inquiries; best-in-class 13.4 % vs 26.9 % for all others; USD 9.40 average vs 2.78 best-in-class per invoice.
Handling time	IOFM survey 2021 via Esker (2026): 5–20 minutes per inquiry; 10–20 % of invoices generate an inquiry; ≈ 60 % concern already-paid invoices. InvoiceInfo (2015): 8 minutes average per inquiry; up to 80 % of inquiries can be handled online. Supports the ≈ 10 minutes per thread working assumption.
ServiceNow value method	ServiceNow Value Calculator (Kanini overview); ServiceNow Source-to-Pay. RSS first pass: ≈ CHF 3.2 M value in year 3.

Access administration

Create an account per person and hand over the initial password personally (chat or phone, not e-mail). Users must replace it at first sign-in. Disable or delete accounts to revoke access immediately.

Users

Initial password for
Shown once. Hand it over personally; the user has to replace it at first sign-in.

E-mail	Name	Role	Status	Last sign-in	Created

Sign-in log

When	Event	E-mail	IP	By

From 200+ mailboxes to one governed service model – what is it worth?

The business case on one page

FTE released by scenario – two parts

Three-year view: benefit ramp vs. cost

Plausibility check – one corridor

Scenario table

What a supplier can do on the Hub today – without writing an e-mail

Where is my invoice?

Submit an invoice

My cases

Company & bank details

Registered data (MDG)

Request a change

Open items & reminders

Saved value cases

Compare

Assumptions, method and sources

How the model calculates – in plain words

Why only two numbers?

What is not in this case

How to read the inputs

Calculation schemes

Worked example – your current case

Working assumptions to confirm

Sources

Set your password

Access administration

Users

Sign-in log

Save value case

From 200+ mailboxes to one governed service model – what is it worth?

The business case on one page

FTE released by scenario – two parts

Three-year view: benefit ramp vs. cost

Plausibility check – one corridor

Scenario table

What a supplier can do on the Hub today – without writing an e-mail

Where is my invoice?

Submit an invoice

My cases

Company & bank details

Registered data (MDG)

Request a change

Open items & reminders

Saved value cases

Compare

Assumptions, method and sources

How the model calculates – in plain words

Why only two numbers?

What is not in this case

How to read the inputs

Calculation schemes

Worked example – your current case

Working assumptions to confirm

Sources

Access administration

Users

Sign-in log

S2P Value Case Calculator