Client portals, dashboards, and embedded BI for small businesses

Client portals and dashboards are the two private, data-bearing surfaces that small and mid-sized businesses (SMBs) most often consider adding to a public website. The two are frequently conflated in vendor marketing but answer different questions, have different adoption profiles, and carry different cost and liability tails. This article catalogues the definitional distinctions, cost ranges, adoption evidence, feature-set patterns, and the security boundary that separates portal and dashboard surfaces from the public marketing site. It connects to related briefs on customer self-service (Customer self-service on small-business websites), interactive engagement mechanisms (Interactive tools and engagement mechanisms), and open data (Open data as competitive moat).

Overview

A client portal (also "customer portal" or "customer login") is a private, authenticated section of a business's website or software in which a specific customer signs in to view and act on information tied to their account. It sits on three layers: authentication, content (files, records, messages scoped to the customer), and permissions (role-based, account-scoped).

Vendor taxonomies converge on this definition across ShareFile, Softr, Wayfront, Cognito Forms, and Moxo.

Source: sharefile.com, softr.io, wayfront.com, cognitoforms.com, moxo.com. Confidence: Industry-consensus.

Caveat: All sources are vendors selling portal/file software; the definitions are nonetheless uncontroversial across competitors.

The portal differs from a public marketing site (open to anonymous visitors, same content for everyone) and from a retail e-commerce account (transaction- and order-history-oriented for discrete one-off purchases). A client/service portal is relationship-, service-, and project-oriented for ongoing engagements.

A portal is private + authenticated + account-scoped — unlike a marketing site (public, anonymous) or an e-commerce account (transaction/order-history oriented).

Source: oski.site, wayfront.com, agencyhandy.com. Confidence: Verified (definitional).

A dashboard is a live, interactive surface that pulls from one or more data sources and supports filtering, drilling, and refresh against a live source. It contrasts with a static report — a frozen, time-stamped snapshot with methodology and context. The two are complementary. The practitioner test: interact with or drill into the data → dashboard; frozen snapshot for later reference or compliance → report.

Dashboard versus static report — Domo and ThoughtSpot framings of the live-versus-snapshot distinction.

Source: Domo, ThoughtSpot framings (industry-consensus across BI vendors). Confidence: Industry-consensus.

Caveat: All sources are BI vendors.

A customer-facing dashboard shows clients their own data; success is measured by retention, activation, or churn deflection. An internal dashboard serves owners, managers, or staff making operational decisions; success is measured by efficient action. The two require entirely different build-versus-buy decision trees.

Internal versus customer-facing dashboards: different metrics, success criteria, and audiences.

Source: Practitioner framing. Confidence: Industry-consensus.

Both surfaces share a deeper capability: account state. The site remembers who the visitor is and what they have done — login, saved items, order history, task progress — so a returning visitor resumes rather than restarts. The defining test is whether a logged-out visitor sees something materially different from a logged-in one.

Account-state capability: returning visitors resume rather than restart; defining test is whether the post-login view differs from the public view.

Source: Industry-consensus definitional framing. Confidence: Industry-consensus.

Caveat: No clean primary dataset ties account-state to conversion or retention for general SMB marketing websites. Business types most served: retailers, service businesses with repeat engagement, and B2B with ongoing relationships.

Adoption and login-friction patterns

The most consistent finding across independent evidence is that portal and dashboard surfaces underperform vendor adoption claims. Two analyst datasets anchor the portal side; two anchor the dashboard side.

For portals and self-service, Gartner's 2019 survey of 8,398 customers found that only 9% of customers reported resolving their issues completely via self-service. A 2024 update, drawn from 5,728 customers surveyed in December 2023, found that only 14% of customer service and support issues are fully resolved in self-service; even for issues customers describe as "very simple," only 36% resolve fully. The 2024 number is bracketed by a striking framing: 73% of customers use self-service at some point in their journey, yet only 14% finish there.

Gartner (2019, 8,398 customers): 9% of customers report resolving issues completely via self-service. 37% of customers call the company before ever reaching the website.

Source: Gartner press release, 2019-09-25. Confidence: Verified.

Gartner (August 19, 2024; survey of 5,728 customers, December 2023): only 14% of issues are fully resolved in self-service; even for "very simple" issues only 36% resolve fully; 73% use self-service at some point.

Source: Gartner press release, 2024-08-19. Confidence: Verified (analyst primary research).

Caveat: Gartner is an independent analyst firm — no portal-vendor incentive.

A companion Gartner finding from the same 2024 release explains the failure mechanism. Forty-five percent of customers who started in self-service said the company did not understand what they were trying to do; in 43% of failure cases, customers could not find content relevant to their issue. Portal surfaces that fail to recognise user intent on entry, and fail to put relevant content within one click of login, will lose roughly that proportion of attempts.

Gartner (August 2024): 45% of self-service starters said the company did not understand what they were trying to do; in 43% of failures users could not find relevant content.

Source: Gartner press releases, 2024-08-19. Confidence: Verified.

Practitioner data frames the same gap. A US Tech Automations practitioner report (2026, accounting vertical) states "most firms find 40-60% of clients haven't logged in in the last 90 days" — the "ghost login" problem. The figure cuts against the vendor's commercial interest (it sells adoption-automation tools), which lends credibility.

Practitioner data (accounting vertical): 40-60% of clients have not logged in in the last 90 days.

Source: US Tech Automations (2026). Confidence: Single-source.

Caveat: Vendor sells adoption-automation tools, but the direction (low adoption) cuts against its interest, which lends credibility.

McKinsey's independent SMB research provides the channel-preference counterweight. A March-November 2022 panel of ~3,500 US SMBs found that small businesses use digital channels 20-30 percent more frequently than analog channels — but "assisted" channels (chat and email) are more popular than pure self-serve. Fewer than 15% prefer phone or voice.

McKinsey, "Winning the SMB tech market in a challenging economy" (~3,500 US SMBs, March-November 2022): SMBs use digital channels 20-30% more frequently than analog; assisted channels (chat and email) beat pure self-serve; <15% want phone/voice.

Source: mckinsey.com, March 2023. Confidence: Verified (primary research, independent).

Caveat: McKinsey sells consulting but is not a portal vendor; the finding cuts both ways: yes to digital, but assisted (chat, email) over self-serve.

The adjacent healthcare benchmark caps the realistic ceiling. A cross-sectional academic study found 32.1% of patients (141/439) used the portal; the literature ranges 26-51%. ER real-time use was 17.4% of 1.28 million patients. US national 2022: 77% offered, 68% accessed. Healthcare-grade incentives still leave most portals topping out at 30-50%; SMB portals without those incentives will underperform that.

Patient-portal usage (peer-reviewed, adjacent benchmark): 32.1% used (range 26-51%); ER real-time 17.4% of 1.28M; US national 2022: 77% offered, 68% accessed.

Source: pmc.ncbi.nlm.nih.gov; healthIT.gov. Confidence: Verified (peer-reviewed).

Password management compounds the friction. NordPass research (March 2024, 1,509 users) found people manage 168 passwords on average; a 2026 update revised to approximately 120; approximately 20% reset a password weekly. Beyond Encryption (which sells secure email and competes with portals) reports that approximately 60% of consumers prefer to receive important information by email.

NordPass research (2024, 1,509 users): people manage 168 passwords on average; 2026 update revised to ~120; ~20% reset weekly; ~60% prefer important information by email.

Source: nordpass.com; beyondencryption.com. Confidence: Verified (NordPass) / Single-source (Beyond Encryption email-preference figure).

Caveat: Beyond Encryption sells secure email and competes with portals; discount the email-preference framing accordingly.

On the dashboard side, two analyst figures anchor adoption. BARC and Eckerson Group's Strategies for Driving Adoption and Usage with BI and Analytics (March 2022; 214 data and analytics leaders surveyed Nov-Dec 2021) found employees actively using BI/analytics tools at 25% on average, with minimal growth over seven years of tracking. Eckerson has noted the rate "has been stuck around 20% for many years."

BARC / Eckerson Group (March 2022): 25% of employees actively use BI/analytics tools, with minimal growth in the past seven years; Eckerson has stated adoption "stuck around 20% for many years."

Source: BARC / Eckerson Group report, March 2022. Confidence: Verified — primary, analyst firm.

Gartner's Survey Analysis: Why BI and Analytics Adoption Remains Low (document 3753469, Howson and Sallam) reports that "pervasive business intelligence remains elusive, with BI and analytics adoption at about 30% of all employees." BI tools have averaged ~25-30% employee adoption for over a decade.

Gartner (document 3753469, Howson and Sallam): BI and analytics adoption at about 30% of all employees.

Source: Gartner document 3753469. Confidence: Verified — primary, analyst firm.

For SMBs specifically, the adoption number is much lower. Software Advice's How Dashboards Impact SMB Confidence in Data Analytics (online survey of 243 US business owners and managers) found that only 5% of SMBs use a BI dashboard; Excel and Google Sheets dominate at 65%; built-in platform analytics at 23%. Top barriers were lack of awareness (39%), system cost (30%), and comfort with existing tools (23%).

Software Advice (n=243 US business owners/managers): 5% of SMBs use a BI dashboard; Excel/Google Sheets dominate at 65%; built-in platform analytics at 23%; top barriers: lack of awareness 39%, system cost 30%, comfort with existing tools 23%.

Source: Software Advice. Confidence: Single-source — vendor-adjacent; treat as directional.

Caveat: Software Advice is a software-comparison site; n=243 is small. Directionally consistent with BARC and Gartner.

Viral statistics that misattribute

Several widely circulated portal and dashboard statistics trace to nothing defensible. The "42% of customers abandon portals out of frustration / 43% prefer email" pairing is misattributed: the 42% traces to Namogoo's e-commerce cart-abandonment research, re-labelled by Beyond Encryption. The "60-70% of dashboards go unused" attribution to "Gartner research" could not be traced to any named Gartner report; it surfaces in vendor blogs and mutates across sources (Fusedash.ai states "between 60 and 80 percent"). The "45% of people with access to BI tools actually use them" line traces to a Logi Analytics report. The "72% of users abandon dashboards for spreadsheets" comes from Luzmo's vendor-run State of Dashboards in 2025 survey of ~200 SaaS/product leaders.

"42% abandon portals / 43% prefer email" pairing is misattributed (42% traces to Namogoo cart-abandonment research, re-labelled by Beyond Encryption). "60-70% (or 60-80%) of dashboards go unused (Gartner)" cannot be traced to any named Gartner report; it traces to vendor blogs and mutates across sources.

Source: Citation-chain traces; Namogoo, Beyond Encryption, Fusedash.ai. Confidence: Verified — the misattribution and missing primary source are reproducible.

"Only 45% of people with access to BI tools use them" traces to a Logi Analytics (BI vendor) report. "72% of users regularly abandon dashboards for spreadsheets" comes from Luzmo's vendor-run State of Dashboards in 2025 survey of ~200 SaaS/product leaders.

Source: Logi Analytics report; Luzmo "State of Dashboards in 2025." Confidence: Single-source / vendor-run.

Caveat: Luzmo sells an embedded-analytics platform; the "abandon for spreadsheets" finding directly supports Luzmo's positioning.

The defensible position on dashboard adoption is the BARC/Gartner numbers (~25-30%), supplemented by Software Advice's SMB-specific 5% figure. The parallel issue for portals is the "selection effect": vendor marketing universally claims "portal users retain better." This is a textbook selection effect — engaged clients both use portals AND retain; the portal does not necessarily cause the retention. Independent academic work on patient portals could "not conclusively determine the causal effect."

"Portal users retain better" is a textbook selection effect — engaged clients both use portals and retain; the portal does not cause the retention. Academic work on patient portals could not conclusively determine the causal effect.

Source: ncbi.nlm.nih.gov; statsig.com; amplitude.com. Confidence: Verified (the principle); Industry-consensus (skeptical).

Portal feature-set patterns

Portals vary widely in scope. At the minimal end, a portal may present a single capability — secure document exchange or invoice viewing. At the rich end, a portal functions as a self-service operations hub combining document sharing, project/status visibility, approvals and e-signatures, invoicing and online payments, scheduling, secure messaging, and self-service account management. Most SMBs need only two or three of the seven feature classes; building or paying for the other four or five is waste.

Portal feature spectrum: from single-document exchange (minimal) to full self-service operations hub combining seven capabilities — document sharing, project/status visibility, approvals/e-signatures, invoicing/payments, scheduling, messaging, account management.

Source: cognitoforms.com, moxo.com, casestatus.com. Confidence: Industry-consensus.

The horizontal portal platforms are the dominant entry point for service businesses without a vertical-SaaS bundle. SuiteDash combines CRM, portal, projects, invoicing, and scheduling at $19/month with unlimited users — the most aggressive single datapoint against quoting a custom build. HoneyBook (creative and service businesses) runs approximately $19 Starter, $36 Essentials, $39 Premium, with payment-processing fees on top. Dubsado runs $20/month (Starter) and $40/month (Premier — which adds a portal subdomain). Bonsai starts at approximately $25/month. Copilot offers a dedicated branded portal at $39-89/month per internal user.

Horizontal-portal pricing: SuiteDash $19/mo unlimited users; HoneyBook $19-79; Dubsado $20-40 with portal subdomain on Premier; Bonsai ~$25; Copilot $39-89/mo per internal user.

Source: suitedash.com, assembly.com, aplosai.com, emakatiraee.com, coliejames.com, agiled.app. Confidence: Verified (SuiteDash); Industry-consensus (multiple 2026 reviews) elsewhere.

Caveat: Prices shift; verify directly before quoting. Payment-processing fees layer on top for HoneyBook.

The Clinked pricing line illustrates the volatility problem: Agiled.app reports Clinked at $77-297/month; Clinked's own pricing page (via Capterra) reports plans from $239/month — a 3× spread requiring re-verification before any quote.

Clinked pricing conflict: agiled.app reports $77-297/month; Clinked's own pricing page (via Capterra) reports plans from $239/month.

Source: agiled.app; capterra.com. Confidence: Single-source per figure; conflict flagged.

The vertical-SaaS bundles are typically the right answer when an SMB sits in a regulated or systematised vertical, because the portal is already in software the SMB needs anyway. Clio (legal) ranges $39-149/user/month with the secure client portal. Jobber (home services) is approximately $69/month. Housecall Pro runs Basic $59-65; Essentials $109-149; Max $169.

Vertical SaaS portal bundles: Clio (legal $39-149/user), Jobber (home services ~$69), Housecall Pro ($59-169) — 2026.

Source: itqlick.com, g2.com, clio.com, housecallpro.com, fieldservicepro.io. Confidence: Verified / Industry-consensus.

Accounting and property-management verticals continue the pattern. TaxDome is approximately $58/user/month with a core portal; Karbon approximately $59/user/month; Canopy approximately $150/month flat. Buildium runs Essential $58-62/month, Growth $183-192/month, Premium $375-400/month, scaling by door count. AppFolio starts at $298/month minimum (50-unit minimum).

Vertical bundles: TaxDome ($58/user), Karbon ($59/user), Canopy ($150 flat), Buildium ($58-400 by door count), AppFolio ($298 floor) — 2026.

Source: taxdome.com, checkthat.ai, aplosai.com, financial-cents.com, capterra.com, thepropertyceo.com, softwareconnect.com, tenantcloud.com. Confidence: Verified / Industry-consensus.

In these verticals, the portal is bundled into software the SMB likely already uses; the marginal cost is effectively zero. The strongest case for not shopping a horizontal portal is the existence of a vertical bundle.

Custom portal builds sit at the top of the cost stack. SPP (2025) reports $20,000-50,000 initial development plus $10,000-25,000/year maintenance; Agency Handy reports $25,000-60,000 all-in over 6-12 months. Wider agency surveys put the custom-web-portal market at entry-level $20,000-40,000, mid $40,000-80,000, enterprise $80,000-400,000; GoodFirms cites $10,000-200,000. A second source set converges on $20,000-60,000 initial, 3-12 months, with maintenance at 15-20% of build per year (approximately $3,000-10,000+/year) for security patching, auth management, and updates. WordPress-and-plugin routes run a membership plugin at $199-399/year (MemberPress, WooCommerce Memberships) plus approximately $1,000-4,000/year of developer maintenance.

SPP (2025): custom client portal $20K-$50K initial + $10K-$25K/yr maintenance. Agency Handy: $25K-$60K all-in over 6-12 months. Wider market: entry $20K-$40K, mid $40K-$80K, enterprise $80K-$400K; GoodFirms cites $10K-$200K.

Source: SPP (2025); Agency Handy; multiple agency sources; GoodFirms. Confidence: Industry-consensus.

Caveat: Both portal-SaaS vendors (motivated to make custom look expensive) and dev shops (motivated to quote high) appear in the source list.

Ongoing maintenance runs at meaningful fractions of build cost annually: $1,800-12,000+/year for SMB sites and $10,000-25,000/year for portals. A client who hears "$25K to build" should also hear "and $10K-25K/year to keep it working."

SMB sites: $1,800-$12,000+/yr maintenance. Portals: $10,000-$25,000/yr — meaningful fractions of build cost.

Source: Industry-consensus across agency sources. Confidence: Industry-consensus.

Caveat: Vendor incentive on both sides — agencies selling maintenance retainers and SaaS vendors selling alternatives both have reasons to surface these figures.

The portal components have become commodity: authentication via Auth0, Cognito, or Firebase Auth; per-customer data in a managed RDS or equivalent; document and invoice access; role-based permissions. Pre-2009 each required a server, a DB admin, and a custom auth build. The bespoke assembly still costs the headline custom-build numbers above.

Enterprise-tier customer account portal: Auth0 + RDS + role-based permissions. Components are commodity; bespoke assembly still costs.

Source: Synthesis of parts + ceiling entries. Confidence: Industry-consensus.

A custom portal is justified in five narrow cases: a core workflow no SaaS models; deep integration with proprietary internal systems; the portal IS the differentiated product; data-residency or custody requirements no SaaS can meet; or scale at which per-seat SaaS pricing inverts versus a flat-cost owned system (usually hundreds-to-thousands of seats). If none of these conditions hold, buying is the right answer.

Five narrow cases for custom: unmodeled workflow; proprietary integration; portal-is-the-product; data-residency; scale-inversion.

Source: oski.site, digiteum.com, strategeonsoftwares.com. Confidence: Directional-Speculative / Industry-consensus.

Portal "market size" projections illustrate why such figures should not anchor decisions. Client-portal-software market estimates for approximately 2022-2024 range from US$1.47 billion (LP Information 2023) to US$10.47 billion (WiseGuy 2023). The approximately 7× spread between estimates for the same period is itself the finding.

Portal "market size" projections range ~7× between sources (US$1.47B-$10.47B for the same period); CAGRs range 6.6%-13.53%.

Source: 360marketupdates.com, researchandmarkets.com, verifiedmarketreports.com, openpr.com, grandviewresearch.com. Confidence: Directional-Speculative.

Caveat: All are report-sellers; the spread is the finding.

The dashboard-as-front-door pattern

A customer-facing dashboard — especially an embedded one inside the website — earns its keep in a narrow band defined by four conditions: the business delivers ongoing, data-rich outcomes to recurring clients who make decisions off that data; the data is genuinely the client's own and updates often enough that live access beats a periodic snapshot; transparency or retention is a real competitive lever; and self-service deflects support or reporting requests.

Customer-facing embedded dashboard earns its keep in a narrow band: recurring clients + frequently-updating data-rich results + retention/differentiation lever + support deflection.

Source: Synthesis across embedded-analytics vendor and practitioner literature. Confidence: Industry-consensus.

This profile fits agencies, SaaS, logistics tracking, portfolio and financial reporting, and utilities — businesses with many client tenants or where analytics is part of the value. It is overkill when clients are few and want a periodic summary, data updates infrequently, no recurring decision exists on the client side, or the SMB lacks maintenance capacity. Real worked examples include agencies' client-results dashboards; SaaS in-product analytics (Shopify, Strava, Zendesk); branded logistics-tracking portals (FourKites, Shippeo, project44, Descartes MacroPoint); and financial/portfolio and utilities-usage dashboards.

Real customer-facing dashboard use cases: agencies (client results), SaaS in-product analytics (Shopify, Strava, Zendesk), logistics tracking portals (FourKites, Shippeo, project44, Descartes MacroPoint).

Source: Vendor product pages + industry coverage. Confidence: Verified.

The default delivery is a static report rather than a dashboard. AgencyAnalytics benchmark research found that a majority of agencies still send static reports to clients while relying on live dashboards internally — evidence that "live and interactive" is not automatically the right client deliverable. Live, not-yet-final numbers invite client misinterpretation; the polished snapshot has narrative control built in.

AgencyAnalytics benchmark: a majority of agencies still send static reports to clients while relying on live dashboards internally; live unfinal numbers invite client misinterpretation.

Source: AgencyAnalytics benchmark survey. Confidence: Single-source — vendor (AgencyAnalytics sells reporting tooling), but the finding cuts against vendor incentive (the vendor would prefer agencies move to live dashboards).

Internal dashboards are worth building when there is a recurring, owned decision tied to a metric with a threshold. They are not worth building as a "single source of truth" wall of vanity metrics. The practitioner discipline qualifies every proposed metric against three criteria: a named owner, a named decision, and a threshold that triggers action.

Used dashboard criteria: built for the daily user (not the executive who requested it); every metric tied to a decision + threshold; embedded in an existing workflow/ritual; trusted (numbers reconcile to source system or users defect to Excel); few metrics (~5-7 per view).

Source: Practitioner synthesis (Domo, ThoughtSpot, Looker, Metabase). Confidence: Industry-consensus.

The sharpest client-facing version of the test is direct: "If this number doubled or halved tomorrow, what specific action would I take?" If none, the metric is vanity (followers, total pageviews, cumulative signups).

"What would I do if this number doubled or halved tomorrow?" — if no action follows, the metric is vanity.

Source: Synthesis of practitioner literature. Confidence: Industry-consensus.

Sector-specific worked examples make the test concrete. Construction operations use weekly job-cost dashboards (labor-productivity variance, committed-versus-actual materials, change orders, billing-versus-percent-complete, backlog) to catch margin erosion before month-end financials arrive. Professional services use Days Sales Outstanding (DSO) dashboards with aging buckets and at-risk-customer flags, plus utilization and realization. A DSO of 30-45 days is a common "good" benchmark; rising DSO is an early cash-flow warning.

Construction: weekly job-cost dashboards (labor variance, committed-versus-actual materials, change orders, billing-versus-percent complete, backlog) catch margin erosion before month-end financials arrive. Professional services: DSO dashboards with aging buckets + at-risk-customer flags; 30-45 days is a common "good" benchmark. Manufacturing: OEE, MTBF/MTTR, PM compliance, planned-versus-reactive ratio (~80/20 target).

Source: Industry practitioner consensus; finance literature; manufacturing-operations literature. Confidence: Industry-consensus (construction, manufacturing); Verified (DSO benchmarks).

Additional internal-dashboard sectors: e-commerce (conversion, inventory turns, CAC, repeat-purchase), healthcare (bed occupancy, admissions, readmissions, department utilization), logistics (on-time delivery, dwell/exceptions, dynamic ETAs), SaaS (activation, weekly active users, churn/NRR, expansion), and hospitality (occupancy, no-shows, repeat guests, days-booked-ahead).

Cross-industry internal dashboards: e-commerce (conversion, inventory turns, CAC); healthcare (occupancy, readmissions); logistics (on-time, exceptions); SaaS (activation, churn, NRR); hospitality (occupancy, no-shows).

Source: Industry-consensus across BI/operations literature. Confidence: Industry-consensus.

Embedded analytics and the BI tools landscape

For SMBs adding dashboard surfaces, four general-purpose BI tools dominate, plus a smaller field of embedded-analytics specialists.

Microsoft Power BI raised Pro to $14/user/month and Premium Per User to $24/user/month (both up from $10/$20) on April 1, 2025. Fabric capacity F2 runs ~$262/month, scaling to F64 at ~$5,000/month where viewers no longer need per-user licenses. Power BI Embedded via A-SKUs starts at ~$735/month (A1). Power BI is the default for organisations already on Microsoft 365 — free on M365 E5. The Fabric tier kicks in when viewer counts make per-seat uneconomic (typically 100+ viewers).

Power BI pricing (April 2025): Pro $14/user/mo, Premium Per User $24/user/mo (both raised from $10/$20); Fabric F2 ~$262/mo to F64 ~$5,000/mo; Embedded A1 ~$735/mo.

Source: microsoft.com, Azure pricing calculator. Confidence: Verified.

Looker Studio offers a widely used free tier. Looker Studio Pro runs $9 per user per Google Cloud project per month — and the per-project detail materially inflates agency costs. A 12-client agency at 5 users pays ~$6,480/year, not the headline ~$540. Separate enterprise Looker (Google Cloud core) runs ~$35,000-60,000+/year.

Looker Studio: free tier widely used; Looker Studio Pro $9/user/Google Cloud project/month (per-project detail materially inflates agency costs); enterprise Looker $35K-60K+/yr.

Source: cloud.google.com/looker/pricing. Confidence: Verified.

Metabase is the strongest mixed-stack default. Open Source is free under AGPL v3 (self-hosted). Cloud Starter is $100/month (5 users included, +$6/user). Pro is $575/month (10 users included, +$12/user) and unlocks white-label, interactive embedding, row-level security, and SSO. Enterprise starts at ~$20,000/year. Metabase Pro at $575/month unlocks embedding cheaper than the dedicated embedded-analytics platforms — a useful intermediate path.

Metabase pricing (2026): Open Source free (AGPL v3 self-host); Cloud Starter $100/mo; Pro $575/mo (white-label + embedded + RLS + SSO); Enterprise from ~$20K/yr.

Source: metabase.com/pricing. Confidence: Verified.

Tableau is the most expensive of the four. Cloud Standard pricing is Viewer $15, Explorer $42, Creator $75 per user/month (annual); Enterprise $35/$70/$115; embedded/OEM custom-quoted with a year-1 floor reportedly $60,000-150,000. Appropriate for SMBs only when the team is already Tableau-fluent or small enough to avoid the Viewer per-seat trap.

Tableau pricing (2026): Viewer $15 / Explorer $42 / Creator $75 per user/mo (annual); Enterprise $35/$70/$115; embedded/OEM year-1 floor reportedly $60K-150K.

Source: tableau.com/pricing. Confidence: Verified — vendor page.

The embedded-analytics platforms define the third path between general BI tools and custom-embedded builds. Luzmo's Embedded plans run Basic $995/month, Pro $2,050/month, Elite $3,100/month. Explo runs $795/month (Growth) and $2,195/month (Pro). Cube offers Cube Core free and open-source, with Cube Cloud consumption-based. Embeddable was in beta as of late 2025. These platforms are typically right when an SMB needs to ship multi-tenant white-labelled analytics to more than ~20 client tenants without writing row-level-security plumbing.

Embedded-analytics platforms (2026): Luzmo $995-3,100/mo embedded plans; Explo $795/Growth, $2,195/Pro; Cube Core open-source + Cube Cloud consumption; Embeddable still in beta.

Source: luzmo.com, explo.co, softwaresuggest, cube.dev. Confidence: Verified.

A custom-embedded build sits at the top of the cost stack. Multiple (mostly vendor) sources converge on a 3-year TCO for production-grade, multi-tenant embedded analytics in the $300,000-630,000 range, with ~20-30% annual maintenance. Building costs ~3× more than teams expect; hidden costs are multi-tenancy and row-level security, performance at scale, and the perpetual maintenance tail. Custom-embedded should be reserved for analytics-as-core-differentiation cases with one to two engineers to own it permanently.

Custom-embedded 3-year TCO converges on $300K-$630K (multiple, mostly vendor sources) with 20-30% annual maintenance; hidden costs are multi-tenancy/RLS, performance at scale, perpetual maintenance.

Source: Synthesis from embedded-analytics vendor literature. Confidence: Industry-consensus — but all major sources are embedded-analytics vendors with incentive to favour "buy."

Caveat: Vendor-modeled, not audited. Still: custom-embedded at $100K-300K + 20-30%/year is the realistic floor.

Self-service reporting versus scheduled report delivery

The strongest argument for the static report as default customer-facing deliverable is that even agencies — the SMB segment most likely to deliver real-time analytics — still default to static reports. The polished snapshot acts as narrative scaffolding: methodology, period, and comparison sit on the page, removing the most common point at which a live dashboard misleads.

Refresh cadence is the operational hinge between dashboard and report. Real-time or near-real-time genuinely matters for operational exception management (logistics ETAs, manufacturing downtime alerts); daily or periodic refresh suffices for most SMB KPI and financial review. Refresh frequency drives cost — Power BI Pro caps scheduled refreshes at 8/day while Premium Per User allows 48/day. Most SMB "we need real-time" requests are actually "we need daily."

Refresh cadence: real-time matters for operational exception use; daily suits most SMB KPIs; refresh frequency drives cost (Power BI Pro 8/day vs Premium Per User 48/day).

Source: Microsoft Power BI documentation; practitioner literature. Confidence: Verified.

The deflection case for self-service has a vendor-sponsored tail. Forrester Total Economic Impact (TEI) studies report self-service or portal deflection of approximately 35% of tickets and cases (SymphonyAI ITSM; Salesforce); Microsoft Dynamics 365 reports approximately 40%. A widely cited unverified "Deloitte: 35% document-retrieval time reduction" figure could not be traced to a primary Deloitte report.

Forrester TEI studies report ~35% support-ticket deflection from portals (SymphonyAI; Salesforce); Microsoft Dynamics 365 ~40%. Every TEI study is commissioned and paid for by the vendor whose product is studied; enterprise-scale, not SMB.

Source: tei.forrester.com; agiled.app; agencyhandy.com. Confidence: Single-source per study.

Caveat: Every TEI study is commissioned and paid for by the vendor whose product is studied; figures are vendor-sponsored and enterprise-scale, not SMB.

The defensible SMB framing is that deflection only matters if the SMB currently fields a steady volume of repetitive "where's my X" messages. The disciplined check is to count repetitive inbox messages over a two-week sample before quoting any deflection rate.

Data freshness, accuracy, and the trust-versus-real-time tradeoff

Dashboards sit on top of a capture layer: operational databases, data warehouses (Snowflake, BigQuery), SaaS APIs, and spreadsheets. The capture layer and data quality form the foundation; abandonment frequently stems from data the user does not trust (numbers that do not reconcile with the source system) rather than from visualization choices.

Dashboards sit on top of a capture layer (DB/warehouse/SaaS APIs/spreadsheets); data quality below the surface determines trust — abandonment frequently stems from numbers that do not reconcile, not from the visualization.

Source: Practitioner/vendor synthesis (consistent across BI vendor literature). Confidence: Industry-consensus.

The build-the-dashboard question is downstream of "is the capture layer right?" — a business cannot dashboard data it has not captured cleanly. The relationship to public datasets is treated separately in Open data as competitive moat.

Dashboards are not build-once assets. Data sources change schemas; metric definitions drift across departments; organisational attention wanes — dashboards spun up for one initiative quietly go unused. Custom-and-embedded builds carry the heaviest burden, with approximately 20-30% of original build effort recurring annually as maintenance.

Dashboard rot: data sources change schemas, metric definitions drift, organisational attention wanes; custom/embedded builds carry the heaviest ~20-30%/yr maintenance burden.

Source: Vendor synthesis. Confidence: Industry-consensus.

The trust-versus-real-time tradeoff favours trust for most SMBs. A daily-refreshed dashboard with reconciled numbers wins user adoption over a real-time dashboard with numbers that do not match the source-of-truth system. The practitioner-consensus ceiling on a single dashboard view is approximately 5-7 metrics; above that, the dashboard becomes a wall and the user defects to a spreadsheet screenshot.

The security boundary — what stays in the portal, what belongs on the public site

The defining decision in the portal-versus-public-site boundary is data custody. Anything authenticated and account-scoped sits inside the portal; anything intended for anonymous discovery sits on the public marketing site. Once the SMB owns a login it owns the regulatory obligations attached to that login.

The Canadian baseline is PIPEDA, which establishes core duties on consent, safeguarding, and access; mandatory breach reporting (since November 2018) with 24-month record-keeping for all breaches; and penalties up to CAD $100,000 per violation for knowingly failing to report, notify, or keep records. The principal organisation stays accountable even when a SaaS processor holds the data. Quebec Law 25, GDPR, and CCPA/CPRA are the equivalents elsewhere.

Every portal recommendation must treat data custody as a project line item, not a footnote: PIPEDA consent + MFA + encryption + breach response plan + 24-month breach records all in scope before launch.

Source: Practitioner discipline synthesised from PIPEDA, Quebec Law 25, GDPR, CCPA/CPRA framings. Confidence: Industry-consensus.

Buying a portal does not transfer accountability. The SMB principal stays liable even when the vendor holds the data. The operational consequence is a line-item checklist before launch: meaningful consent language, MFA on every account, encryption in transit and at rest, a breach-response plan naming who notifies whom on what timeline, a PIPEDA-compliant processor agreement with any SaaS host, and a budget for breach record-keeping. Most US-only portal vendors have weaker EU and Quebec posture; vendor compliance must be confirmed against the SMB's customer geography.

The boundary carries forward to dashboards. Customer-facing dashboards that show account-scoped data sit inside the security boundary; aggregated industry benchmarks the SMB publishes on the public site sit outside it. Anything tied to a customer identifier requires authentication, encryption, and the PIPEDA-equivalent posture.

The downstream design consequence is that login flow, password reset, MFA enrolment, and recovery become product surfaces in their own right. The friction tax (~120-168 passwords managed per user, ~20% weekly resets, ~60% prefer important information by email) operates on every additional authenticated surface added. The default delivery — wherever it holds — should be the lightest: a scheduled PDF or shared reporting link before any embedded dashboard; an inbox-and-text workflow before any portal.

Source-quality summary

The source base for portals and dashboards is dominated by vendors that profit from selling them. For client portals, source incentives are pervasive across pricing, ROI, deflection, and demand claims; the independent anchors are McKinsey (SMB channel preferences) and Gartner (self-service resolution). Market-size figures are unreliable (~7× spread). The viral 42% portal-abandonment figure is misattributed. Forrester TEI deflection studies are vendor-commissioned and enterprise-scale. No causal evidence links portals to SMB retention.

Caveats for the client-portals brief: source-incentives pervasive; independent anchors are McKinsey and Gartner; market-size figures unreliable; the viral 42% stat is misattributed.

Source: Internal source-quality assessment, June 2026. Confidence: Industry-consensus.

For dashboards, the source-incentive flag is similarly pervasive across Microsoft, Salesforce/Tableau, Google, Metabase, Luzmo, Explo, Cube, Logi Analytics, and Sisense. The "60-70% of dashboards go unused" statistic is folklore. The Luzmo and Logi Analytics figures are vendor-run research. SMB-specific data is thin. Build-versus-buy TCO numbers are vendor-modeled and skew toward buy. Retention and churn-improvement figures should be treated as marketing.

Caveats for the dashboards brief: pervasive BI/embedded-analytics vendor sourcing; "60-70%" is folklore; SMB data thin; retention claims unproven.

Source: Internal source-quality assessment, June 2026. Confidence: Industry-consensus.

The broader self-service surface this article participates in is catalogued in Customer self-service on small-business websites. The interactive-tool side, where unauthenticated visitors run calculators or assessments before booking, sits in Interactive tools and engagement mechanisms. The open-data side, where public datasets become a competitive moat, sits in Open data as competitive moat. Those three are the public surfaces; portals and dashboards are the private ones, and the boundary between them is the central operational decision.