Intranet business cases fail because they are built entirely on soft benefits. EPC Group builds them on four ledgers, only two of which a CFO will actually underwrite: retired cost and measured time recovery. Price the full cost side — including ongoing ownership — and phase the ask so the first tranche is small.
Last updated: 2026-07-31
EPC Group is a Houston-based Microsoft consulting firm operating since 1997, with six Microsoft Solutions Partner designations and 216+ M&A tenant migrations covering 1.83M users. Governed Microsoft AI, Data & Cloud — since 1997.
Key facts
- A finance function funds retired cost and measured time recovery. Everything else is supporting narrative. Build the case in that order or it dies in review.
- SharePoint tenant storage is 1 TB plus 10 GB per licence. Extra storage bills at $0.20/GB/month, Microsoft 365 Archive at $0.05/GB/month, Backup at $0.15/GB/month (as published at time of writing (July 2026)).
- The intranet has no separate licence. Communication sites, hub sites, the home site, global navigation and Viva Connections come with Microsoft 365 — the spend is people, content and time.
- The Portal launch scheduler requires phased waves by audience size: two under 10,000 users, three for 10,000–30,000, five above 30,000, and a support request above 100,000.
- Platform limits shape the design: 2,000 hub sites per organisation, 2,000 lists and libraries per site, 500 child links per navigation level.
- Load testing SharePoint Online is not permitted. Your only pre-launch performance evidence is a healthy Page Diagnostics score, which the launch scheduler requires before it will schedule.
- Never put an unattributed vendor percentage in the model. Even Microsoft's Adoption Score documentation cites "Forrester research" for a 100-minutes-per-week saving without naming the study or its date.
Quick facts
| Question | Answer |
|---|---|
| Does an intranet need its own licence? | No. It is built from Microsoft 365 capabilities you already own |
| Where does the money go? | Build, content migration, change management, ongoing ownership |
| Which benefit gets funded? | Retired third-party cost, and measured before-and-after time recovery |
| Which benefit does not | "Improved engagement", "better culture", "modern experience" — supporting only |
| Biggest omitted cost | Ongoing ownership: editors, governance, content review, platform admin |
| Typical first-tranche scope | Home site, global navigation, 3–5 top-task flows, one measured population |
| Measurement window | 6–8 weeks baseline before build; the same measures 90 days after launch |
| Launch method | Portal launch scheduler in waves; a healthy Page Diagnostics score is required |
| Most common failure | The business case is written after the design instead of before |
| Where it pays for itself twice | Copilot answer quality — current, owned, well-structured content |
Why intranet projects fail to get funded
They fail for four reasons, and only one is about money.
They lead with soft benefits. "Improved employee engagement", "one place to find everything", "a modern experience". None has a defensible unit, a baseline, or an owner. A CFO is not being obstructive when they decline; they are declining to book a number nobody can audit.
They price the build and not the operation. A proposal covering design, information architecture and migration, with no line for the editorial and governance capacity to keep it alive, is not a business case. It is a capital request creating an unfunded operating liability, and finance people recognise that pattern instantly.
They ask for one large number. A $600,000 request needs a steering committee, a competitive process and a budget cycle. A $90,000 first tranche with a defined measurement gate needs a director's signature. The second is not a smaller project, it is a differently shaped one.
They have no baseline. The team proposes to improve something it never measured, and nine months later cannot prove it improved, which poisons the well for the next request. Baseline capture is cheap and takes weeks — the highest-return hours in the programme.
A fifth reason is worth naming plainly: intranet projects are usually owned by Internal Communications, whose success measures are engagement measures, and funded by Finance, whose success measures are cost measures. If nobody translates between the two, the proposal is written in a language the approver does not read.
The Four-Ledger Intranet Business Case
The Four-Ledger Intranet Business Case is EPC Group's model for structuring the ask. Its governing rule: only Ledgers 1 and 2 get funded; Ledgers 3 and 4 get you a hearing. Put them in this order and state the distinction explicitly. Approvers trust a case that tells them which parts are soft.
| Ledger | What it contains | CFO treatment | Evidence required |
|---|---|---|---|
| 1 — Retired cost | Third-party intranet or portal licences, hosting, a search appliance, a separate mobile app, an on-premises farm and its infrastructure | Accepted at face value. A contract you stop paying | Signed contract, renewal date, invoice |
| 2 — Measured time recovery | Time to find a policy, a form, a benefits answer — measured before and after on the same population, valued at a loaded rate | Accepted with discipline. Needs a real baseline and usually a haircut | Timed task study or ticket-deflection counts |
| 3 — Avoided risk | Fewer stale-policy errors, cleaner audit evidence, reduced oversharing, controlled retention | Discounted, but persuasive. Rarely booked; often decisive when a real audit finding exists | Audit finding, regulator letter, incident record |
| 4 — Option value | Copilot answer quality, Viva Connections reach to frontline staff, faster future rollout | Not booked. A tie-breaker, never the justification | Directional only — say so |
Two rules follow. Never total Ledgers 3 and 4 into the headline number — the moment you sum a soft benefit into the ROI line, a competent reviewer discounts the whole model, including the sound parts. And write Ledger 1 first, because a case opening with a contract you can cancel earns the attention the rest needs.
A worked model — with illustrative inputs you replace
Every figure below is illustrative — a placeholder chosen to show the arithmetic, not an EPC client result, not a benchmark, and not a claim about your organisation. Replace every one.
Variables
| Variable | Meaning | Illustrative input |
|---|---|---|
U | Employees in the measured population | 2,000 |
R | Fully loaded hourly cost | $45 |
T_before | Measured minutes per person per week finding internal information | 22 |
T_after | The same measure, 90 days post-launch | 15 |
H | Confidence haircut applied to recovered time | 50% |
L | Annual third-party licence and hosting cost retired | $84,000 |
C_build | One-time build, migration, change management | $260,000 |
C_run | Annual ongoing ownership | $95,000 |
Ledger 2 arithmetic
- Minutes recovered per person per week:
T_before − T_after= 22 − 15 = 7 - Annual hours recovered per person: 7 × 46 working weeks ÷ 60 = 5.37 hours
- Gross annual value: 5.37 ×
U×R= 5.37 × 2,000 × $45 = $483,300 - After haircut
H: $483,300 × 0.50 = $241,650
Combined
| Line | Calculation | Illustrative result |
|---|---|---|
| Ledger 1 — retired cost | L | $84,000/yr |
| Ledger 2 — time recovery after haircut | step 4 | $241,650/yr |
| Gross annual benefit | L1 + L2 | $325,650 |
| Annual ongoing cost | C_run | −$95,000 |
| Net annual benefit | $230,650 | |
| One-time investment | C_build | $260,000 |
| Simple payback | C_build ÷ net annual benefit | 13.5 months |
| Year-1 net position | net annual benefit − C_build | −$29,350 |
| Year-2 cumulative | +$201,300 |
Four things this model does that a feature tour cannot:
It makes the haircut visible. H is the most contested number in the room, because recovered minutes do not automatically become productive output. Publishing it at 50% and inviting Finance to set it converts an argument into a parameter. Teams that hide the haircut get the model rejected; teams that expose it negotiate one cell.
It separates capital from operating. C_build and C_run sit on different lines because they hit different budgets and approvers. A model that blends them has not spoken to Finance.
It shows a year-1 loss. Almost every honest intranet case is negative in year one, and saying so is the credibility move. A model showing year-1 profit is one somebody massaged, and reviewers know it.
It survives a sensitivity test. Halve the time recovery to 3.5 minutes and payback moves to roughly 25 months — still fundable for a platform with a five-year life. That is the sentence that wins the meeting, so run and publish the downside case yourself.
The cost side, in full
The line most proposals omit is the last one.
| Cost category | What it actually covers | Notes and published rates |
|---|---|---|
| Licensing | Nothing incremental for the intranet itself | These are Microsoft 365 capabilities. Verify your tier in our E3 vs E5 comparison |
| Storage | Growth from consolidating file shares and archives | Tenant quota is 1 TB + 10 GB per licence; overage $0.20/GB/month; Microsoft 365 Archive $0.05/GB/month; Backup $0.15/GB/month |
| Design and build | Information architecture, navigation, home site, hub structure, page templates, branding, accessibility, multilingual, performance tuning | Largest one-time line. Multilingual and accessibility are cost multipliers, not extras |
| Content migration and rewrite | Inventory, ROT analysis, rewrite, re-tagging, owner assignment. The rewrite dominates — moving content is cheap, deciding what it should say is not | Scope with our SharePoint migration services guide; budget editorial hours separately from tooling |
| Integration | Line-of-business links, single sign-on, Teams surfacing, search connectors, HR and ticketing embeds | Each is a support obligation, not a one-time build |
| Change management | Comms plan, phased launch waves, editor training, champion network, launch-week support | Underfunding this is the most common cause of a technically successful, behaviourally failed intranet |
| Ongoing ownership | Editorial capacity, governance forum, content review cycles, navigation upkeep, platform administration, analytics reporting | The omitted line. Express as named FTE fractions, not a percentage. If nobody owns content, the intranet is stale within two quarters |
Two disciplines separate credible proposals from optimistic ones. Price ongoing ownership as headcount, not a contingency percentage — 0.5 FTE of editorial and 0.25 FTE of platform administration is a number a finance business partner can validate against a salary band. And price content by decision, not by gigabyte: the expensive question is not "how many documents move" but "who decides which of the 400 HR pages survive."
Hard versus soft benefits, and which ones survive review
| Benefit | Ledger | Will a CFO accept it? | Why |
|---|---|---|---|
| Retired third-party intranet licence | 1 | Yes | Contract cancellation, verifiable to the invoice |
| Decommissioned on-premises farm and infrastructure | 1 | Yes | Hardware, hosting, patching and support all stop |
| Retired duplicate storage or search tooling | 1 | Yes, with proof of decommission | Only if genuinely switched off |
| Measured reduction in time-to-find, valued at loaded rate | 2 | Yes, with a haircut | Requires a real before-and-after on the same population |
| Deflected HR and IT tickets on published answers | 2 | Yes | Ticket counts are already instrumented and independently auditable |
| Reduced printing and physical distribution | 2 | Yes | Small but hard, and easy to verify |
| Fewer errors from stale policy documents | 3 | Sometimes | Only where a specific incident or audit finding puts a number on it |
| Audit and compliance evidence quality | 3 | Sometimes | Persuasive where a regulator already commented — see our SOC 2 guidance |
| Reduced oversharing exposure ahead of Copilot | 3 | Sometimes | Governance value, rarely booked, frequently decisive |
| Improved employee engagement or sentiment | 4 | No | No auditable unit; useful as narrative only |
| Faster onboarding | 4 | No, unless timed | Becomes Ledger 2 the moment you actually measure ramp time |
| Better Copilot answers from current, owned content | 4 | No | Real, directional, and not bookable — say so plainly |
The honest sentence for your paper: this case is justified by Ledgers 1 and 2; Ledgers 3 and 4 are stated for completeness and excluded from the return. That sentence does more for approval odds than ten more slides of benefit narrative — the discipline we apply in the CFO AI governance conversation.
Phasing: how to get a smaller first tranche approved
Ask for the smallest thing that produces a measured result. We call it Tranche Zero; its only job is to make Tranche One easy to approve.
| Tranche | Scope | Duration | What it proves |
|---|---|---|---|
| Zero — baseline | Measure time-to-find, ticket volumes, portal usage and content ROT on one population. No build. | 6–8 weeks | That you have a number to improve, and the problem is real |
| One — spine | Home site, global navigation, hub structure, 3–5 top-task flows, one department fully migrated | 10–14 weeks | That the design works and adoption is achievable |
| Two — measured gate | Re-run the Tranche Zero measures on the same population at 90 days | 2 weeks | Whether Ledger 2 is real. Fund Three only if it is |
| Three — scale | Remaining departments, integrations, Viva Connections for frontline, multilingual | 4–8 months | Breadth against a proven model |
| Four — operate | Editorial cadence, governance forum, quarterly content review, analytics | Continuous | That the asset does not decay |
Two notes. Tranche Two is a gate, not a milestone — write the abandonment criterion into the paper, because a proposal specifying its own kill condition is markedly more likely to pass. And launch is phased regardless: the Portal launch scheduler requires two waves under 10,000 users, three for 10,000–30,000, five above 30,000, and a support request above 100,000, each wave running one to seven days. The platform itself expects you to go in waves, which is also your best argument for tranche-based funding.
The measurement plan
Everything in Ledger 2 depends on this. Run it before you build anything.
Baseline, weeks 1–8. Pick one named population of 150–500 people. Run a timed task study: give 20–30 people ten realistic tasks ("find the current expense policy", "find the parental leave form") and record time-to-answer and success rate. Pull HR and IT ticket volumes for questions whose answers should be published, plus SharePoint site usage and Microsoft Search usage reports. Count content ROT: pages not updated in 24 months, pages with no owner, duplicated policies.
Instrument during build. Decide the post-launch measures before the design is finished, or the design quietly optimises for something else. Publish the metric definitions.
Re-measure at 90 days. Same tasks, same population, same instrument. Ninety days, not thirty — thirty measures novelty. Re-run at 12 months to test decay, the number that justifies Tranche Four.
Report honestly. If T_after barely moved, say so and diagnose it — the cause is almost always content quality or ownership rather than navigation design, and that finding is worth more than a flattering number. A programme that reports a miss and fixes it earns the next tranche; one reporting a suspiciously round improvement does not.
Instrumentation you already own: SharePoint site and page usage reports, Microsoft Search usage reports (28-day and yearly views), Microsoft 365 usage analytics, and Adoption Score. Treat Adoption Score as directional context, not a benefit — it is a peer-benchmarked composite, not a financial measure.
What breaks — failure modes
| Symptom | Root cause | Fix |
|---|---|---|
| Business case rejected without discussion | Headline built on soft benefits, or Ledgers 3 and 4 totalled into it | Rebuild with Ledger 1 first; state that 3 and 4 are excluded from the return |
| Finance disputes the ROI and the whole case stalls | The time-recovery haircut is hidden inside the model | Expose H as a named parameter and let Finance set it |
| Project approved, then quietly starved | Ongoing ownership never funded, so it competes annually for scraps | Price C_run as named FTE fractions in the original paper |
| Intranet launches, adoption flat | Change management underfunded; no champion network; no launch waves | Fund change management as a line item; use launch scheduler waves |
| Home page slow, launch blocked | Page Diagnostics score not healthy; too many web parts, search requests or iframes | Follow the healthy-portal guidance; load testing SharePoint Online is not permitted |
| Content stale within six months | No owner per page, no review cycle, no attestation | Assign owners at migration; quarterly content review; automate with lifecycle policies |
| Navigation unusable as the estate grows | Limits ignored — 500 child links per level, 2,000 hub sites per organisation | Design the information architecture to the published limits before you build |
| Storage bill appears after consolidation | Tenant quota of 1 TB + 10 GB per licence exceeded | Archive inactive sites at $0.05/GB/month rather than buying storage at $0.20/GB/month |
| 90-day measurement shows no improvement | Content was migrated, not rewritten; users still cannot find the answer | Rewrite top-task content; time-to-find is usually a content problem, not a design one |
| Benefits cannot be proven at all | No baseline captured before the build | There is no retrofit. Always capture the baseline in Tranche Zero |
What changed in 2026
- Viva Connections and SharePoint home sites are complementary, not alternatives. Connections auto-detects an existing home site, they share a permissions model, and multiple home sites are supported. Budget one design effort, not two.
- The news reader replaced the Feed across desktop, web and mobile, pulling news from organisational sites, boosted news, followed and frequent sites, with audience targeting. Editorial planning changes accordingly — reach is now curated, not chronological.
- Microsoft 365 Archive reactivation fees were eliminated (March 31, 2025), with re-archiving of newly reactivated content restricted for four months. Archiving at $0.05/GB/month is now materially cheaper than buying standard storage at $0.20/GB/month for content you are keeping but not using.
- Copilot made content currency a financial variable. Stale, ownerless content now degrades AI answers across the organisation, strengthening the Ledger 4 argument even though it stays unbookable. Sequence this alongside Copilot licensing planning.
- Adoption Score changed shape (January 2026). Technology experiences retired; the maximum moved from 900 to 600 and now equals the People experiences score. Any prior business case citing an Adoption Score target no longer means what it meant.
- Governance moved upstream. Site ownership, inactive-site and attestation policies now enforce the operating model rather than negotiate it, making Tranche Four defensible for the first time.
Where to go next
If your intranet proposal has been circulating for two budget cycles, the problem is the shape of the ask, not the design. EPC Group runs Tranche Zero as a fixed-scope engagement: baseline capture, content ROT analysis, a completed Four-Ledger model in your finance function's format, and a phased plan with a written kill condition. Start with SharePoint consulting or the SharePoint migration services guide.
Related: My Sites in SharePoint · SharePoint vs Google Drive · Box vs SharePoint · SharePoint network drive mapping · Teams Premium features · Data governance firms · Microsoft consulting firms · Delivery-partner risk · Microsoft Frontier Company · EPC Group
Frequently asked questions
Does a SharePoint intranet require additional licensing?
No. Communication sites, hub sites, the home site, global navigation and Viva Connections are Microsoft 365 capabilities, so there is no separate intranet SKU. Your incremental costs are storage above the tenant quota of 1 TB plus 10 GB per licence, plus build, content migration, change management and ongoing ownership.
What is a realistic payback period for an intranet?
Most honest models land between 12 and 30 months, and almost all show a net loss in year one because the build is one-time and the benefits are annual. Any model showing payback inside a year is usually hiding the ongoing ownership cost or applying no haircut to recovered time.
Which benefits will a CFO actually accept?
Two: retired third-party cost you can prove with a cancelled contract, and time recovery measured on the same population before and after, valued at a loaded rate and discounted with a stated haircut. Risk reduction and option value belong in the paper but should never be totalled into the headline return.
What is the haircut and why does it matter?
The haircut is the percentage by which you discount recovered time, because minutes saved do not automatically convert to productive output. Publishing it — 50% is a common starting point — and letting Finance set the value turns the most contested assumption into a negotiable parameter instead of an argument about credibility.
How do I capture a baseline?
Take one named population of 150–500 people. Run a timed task study with 20–30 participants and ten realistic tasks, recording time-to-answer and success rate. Pull HR and IT ticket volumes for questions that should be published, plus SharePoint and Microsoft Search usage reports. Do this over six to eight weeks before any build.
How large should the first funding request be?
Small enough for a single approver. Tranche Zero is measurement only and costs weeks, not capital. Tranche One is a spine — home site, global navigation, three to five top-task flows, one department fully migrated. Scale funding is released only after the 90-day gate confirms the time recovery is real.
Why should the business case include a kill condition?
Because approvers fund proposals that can be stopped. Writing the abandonment criterion into Tranche Two — "if measured time-to-find has not improved by X, we do not fund Tranche Three" — signals that the team is measuring honestly rather than defending a decision already made.
How do I cost ongoing ownership?
As named FTE fractions, not a contingency percentage. Typical components are editorial capacity, governance forum time, quarterly content review, navigation upkeep, platform administration and analytics reporting. A finance business partner can validate a headcount fraction against a salary band; they cannot validate "15% of build cost."
Can I use published industry ROI percentages in my case?
Not safely. Most circulating figures are vendor-commissioned and unattributed; even Microsoft's Adoption Score documentation cites "Forrester research" for a 100-minutes-per-week saving without naming the study or its date. Build the model from your own measured inputs. A reviewer who finds one unsupported number will discount every number.
How does an intranet affect Microsoft 365 Copilot?
Copilot grounds answers in content users can already access, so stale, duplicated and ownerless intranet content degrades answers across the organisation. Modernization improves that materially, but treat it as option value: state it in Ledger 4 and do not book it.
Sources and verification
- Microsoft Learn — SharePoint limits (storage quota, hub, navigation and list limits)
- Microsoft Learn — SharePoint storage planning
- Microsoft Learn — Add more OneDrive and SharePoint storage to your Microsoft 365 subscription
- Microsoft Learn — Pay-as-you-go services and pricing (Archive $0.05/GB/month, Backup $0.15/GB/month, SharePoint storage $0.20/GB/month)
- Microsoft Learn — Pricing model for Microsoft 365 Archive (reactivation fee elimination, March 31 2025)
- Microsoft Learn — Plan, build, and launch a SharePoint home site for your organization
- Microsoft Learn — Overview of Viva Connections
- Microsoft Learn — Plan Microsoft Viva Connections for your organization
- Microsoft Learn — Introduction to the SharePoint app bar and global navigation
- Microsoft Learn — Planning your SharePoint hub sites
- Microsoft Learn — SharePoint information architecture
- Microsoft Learn — Creating and launching a healthy SharePoint portal
- Microsoft Learn — Launch your portal using the SharePoint Portal launch scheduler (wave thresholds)
- Microsoft Learn — Planning your portal launch roll-out plan (load testing not permitted)
- Microsoft Learn — Page Diagnostics for SharePoint
- Microsoft Learn — SharePoint site usage reports
- Microsoft Learn — Microsoft Search usage reports
- Microsoft Learn — Microsoft 365 usage analytics overview
- Microsoft Learn — Microsoft Adoption Score (January 2026 scoring change; the unattributed Forrester citation)
- Microsoft Learn — Manage site lifecycle policies
- Microsoft Learn — Overview of Microsoft 365 Archive
- Microsoft Learn — Review Microsoft 365 communication tools
- SharePoint Maven — Gregory Zelfond, Decisions for you to consider when creating a governance plan for SharePoint Online, published July 15 2024, updated July 18 2025 (the citation this article replaces)