Agency Portfolio Math: Pooling 15 Clients Over One A/B License
Pooling 15 client sites onto one shared A/B testing license drops the per-client cost to about €55/month and dissolves the velocity floor that kills the ROI math for a single brand.
Quick answer
A €700–€900/month A/B testing license that is unjustifiable for a single €2M store becomes ~€47–€60 per client per month when a 15-client agency runs it as a shared seat. The minimum test-velocity floor drops with it: instead of needing 8–10 tests/month from one brand, you need ~0.7 tests/client/month across the book to clear ROI.
Agency portfolio math for A/B tooling
The unit economics of amortising one A/B testing license across a multi-client agency book instead of billing it per brand.
Agency portfolio math is the specific ROI calculation an agency runs when deciding whether to buy an experimentation platform once and use it across every client site, rather than asking each client to license their own. Because most enterprise A/B tools price per domain, per MTU, or per seat — not per test — a 15-client roster spreads a fixed license cost thin enough that the per-client burden drops below what any individual brand would pay for an equivalent standalone plan. The knock-on effect is that the minimum monthly test volume needed to justify the tool also fragments across the book, which is what actually unlocks the purchase decision.
Most standalone brands doing €1–3M in revenue fail the tool-justifies-itself test at typical enterprise A/B pricing. The break-even in the parent calculation, Minimum Test Velocity To Justify An A/B Tool, sits around 6–10 tests per month for a single-site license — a bar most in-house teams never clear.
An agency running 15 concurrent client sites plays a different game. The license cost stays flat; the test count that has to be produced to hit ROI gets divided by 15 client accounts, and each client only needs to contribute a fraction of a test per month.
Why the per-client cost drops so fast
A typical VWO agency plan or Optimizely mid-tier lands around €800/month for the level a 15-site book actually needs (unlimited experiments, multi-domain, one workspace per client). Divided across the book, that's €53 per client per month — cheaper than most clients' Hotjar bill.
Convert.com's agency tier and AB Tasty's partner program work the same way: one contract, multi-domain rights, workspace isolation so a beauty client and an apparel client don't see each other's experiments. The pricing sheet looks expensive until you divide.
The margin the agency actually keeps
If you rebill the tool to clients at €150/month as part of a CRO retainer, the license becomes a ~€1,455/month gross-margin line item on top of service fees. Most agencies stop treating tooling as a cost centre once the math clicks.
The velocity floor collapses too
The harder constraint isn't cost — it's whether you can produce enough winning tests to make the license pay for itself in incremental revenue. A single €2M Shopify store needs to run roughly 8 tests/month to hit a positive ROI on an €800 license at typical CRO win rates (~15–20% of tests produce a lift).
Across 15 clients, the same test throughput is spread over 15 roadmaps. You need ~0.7 tests per client per month — one meaningful experiment every ~6 weeks per brand — to keep the same aggregate volume. That's a realistic cadence for a small CRO pod, not a mythical one.
Better still, winning ideas port between similar clients. A PDP layout that lifts AOV for one apparel client is a validated hypothesis you can retest, not reinvent, on the next apparel client — compounding win rate across the book without proportional research cost.
Per-client cost by tool at 15 sites
Typical A/B testing license cost split across a 15-client agency book (mid-tier plans, monthly rate)
| Tool | Monthly list price (agency/multi-site tier) | Per-client cost at 15 clients | Domains included |
|---|---|---|---|
| VWO Growth (agency) | €780 | €52 | Unlimited |
| Optimizely Web Experimentation | €1,900 | €127 | Negotiated |
| AB Tasty Partner | €900 | €60 | Unlimited |
| Convert.com Agency | €650 | €43 | Unlimited |
| Kameleoon Partner | €1,100 | €73 | Unlimited |
| GrowthBook Cloud Pro | €240 | €16 | Unlimited |
List prices vary by MTU bands and negotiation — the numbers above assume ~250k MTUs per client site and a genuine multi-domain contract. Convert and GrowthBook come in cheaper because they price traffic and features more agency-friendly out of the box.
Where the math breaks
Pooling only works if each client site has enough traffic to reach significance in a reasonable window. A client doing 8,000 sessions/month can't finish a test on their own PDP inside a quarter — their share of the license buys them nothing. Concentrate the seat on clients with ≥40k monthly sessions on the pages you'd test.
The other failure mode is contract terms. Some vendors' 'agency' plans still cap total MAU across all clients, so a viral product launch on one client eats the whole pool. Read the fair-use clause before you commit — and negotiate MTU headroom equal to 2x the sum of your clients' current traffic.
When it's cheaper to skip the enterprise seat entirely
If your 15 clients are all on Shopify and none exceed 200k monthly sessions, a lightweight plug-in-based testing tool per client often beats a pooled enterprise license on both cost and setup time. Zero-dev install, per-store billing, no shared-workspace complexity — the agency layer isn't earning its keep in that scenario.
The pooled-license model wins when clients skew mid-market (€3M+), when at least half the book has a real experimentation program, and when you have a central CRO team producing hypotheses rather than 15 independent client leads each requesting their own tool.
Agency A/B tool pricing questions
Yes, all three sell explicit agency or partner tiers that include multi-domain rights and workspace isolation per client. You need to sign the agency addendum — the self-serve plan on their website does not include multi-brand usage.
€45–€130 per client per month depending on the tool and MTU tier. VWO, Convert and AB Tasty land in the €45–€75 range; Optimizely and Kameleoon are pricier because they charge more for enterprise features you may not need.
About 0.7 tests per client per month across 15 sites, assuming a €800/month license and typical DTC test economics. That's roughly one meaningful experiment every six weeks per brand — well within reach for an agency with a dedicated CRO producer.
Yes, and most agencies do. Rebilling at €125–€175/month per client either as a line item or bundled into a CRO retainer is standard practice, and clients rarely push back because the equivalent standalone license would cost them 5–10x.
You either upgrade the whole license to the next MTU band or move that client to their own seat. Plan for this at the 100k-MAU-per-client mark — that's where per-client economics usually flip in favour of a dedicated license.
Only if the active half generates enough test volume on their own. If only 5 of 15 clients are running experiments, you're effectively paying €160/active-client — still cheaper than standalone plans, but the velocity math tightens. Consider dropping inactive clients from the seat count.
Each vendor implements it slightly differently, but the pattern is the same: separate projects per client domain, role-based access so consultants only see their assigned clients, and separate reporting scopes. Ask specifically for GDPR-compliant data separation — some tools share cookies across the parent account.
You can port the hypothesis and the variant design, but you must re-test on the new client's traffic — audiences, price points and product categories differ enough that lift rarely transfers 1:1. Treat prior wins as high-priority hypotheses, not conclusions.
For clients under 200k monthly sessions on Shopify, per-store lightweight tools with zero-dev install are usually a better fit than pooling them into an enterprise seat. You lose central reporting but save 30–50% on tooling cost per small client.
Not if the vendor supports true workspace isolation and separate data-processing agreements per client. The risk shows up in cheaper tools that treat the account as one tenant — a client's data engineer will spot cross-account cookies during a security review and it becomes a churn event.
See Metricuno on your data
Bring your stack — Google Analytics, Stripe, a CRM, anything — and we'll walk through the metric tree that turns your funnel into one number.