1. Overview: What the Integration Is & the Value It Delivers
Prism brings together Fospha’s daily, full-funnel measurement with Smartly.io’s Predictive Budget Allocation (PBA) — automating budget shifts across channels, campaigns, and ad sets to drive incremental impact every day.
It’s the only daily MMM integrated directly with Smartly’s PBA — combining scientific measurement accuracy with operational automation.
Why it Matters
Modern marketing teams face three consistent challenges:
Too many decisions, not enough hours — high-value optimisation opportunities get missed.
Siloed or biased data — platform-reported metrics exaggerate performance in their own ecosystem.
Unstable signals — frequent resets in platform learning create inconsistent outcomes.
Prism solves these by turning Fospha’s unbiased, model-based data into Smartly’s optimisation engine — so spend shifts automatically toward the campaigns that actually drive new customers and incremental revenue.
In short:
More growth, less lift — PBA executes the “fifth task on your list” every day.
Smarter budgets — spend moves toward campaigns proven to drive true business results.
Daily confidence — MMM-level accuracy refreshed every day for stable, cross-channel guidance
2. How It Works: The Closed Feedback Loop
Prism operates as a 4-step continuous loop between Fospha and Smartly:
Fospha provides the data
Fospha shares 14 days of attributed performance data — conversions, new-customer conversions, and revenue — mapped to campaign and ad IDs through automated matching.
→ Source: daily S2S feed directly from Fospha.Smartly ingests the data
A secure server-to-server (S2S) API transfers Fospha data into Smartly’s dashboard as a new optimisation signal for PBA.Smartly reallocates budget
Smartly’s Predictive Budget Allocation automatically redistributes spend across campaigns and channels based on Fospha’s MMM-led performance data — optimising toward your chosen KPI (ROAS, revenue, conversions, or new-customer ROAS).Fospha validates the results
Fospha measures the incremental effect of those budget changes and reports Test vs Control outcomes, creating a closed feedback loop that strengthens over time.
3. Best Practice: Identifying the Right Test Opportunity + Guardrails
To ensure Prism delivers consistent, measurable value, start with well-defined pilot tests that balance volume, stability, and learning potential.
How to Spot a Strong Test Opportunity
Start with conversion-focused campaigns.
Immediate results are most visible in conversion-driven activity (e.g. evergreen prospecting or DPA campaigns).
Avoid mixing funnel stages (Awareness, Consideration, Conversion) in one pool; PBA works best when all campaigns share a single optimisation objective.
Ensure sufficient volume and headroom.
Choose campaigns with enough conversions to generate a stable learning signal (≈5–10× expected daily CPA budget in the pool).
Use Fospha’s Incremental Forecasting outputs to identify campaigns with untapped growth potential.
Exclude your very best performers.
Don’t risk early downside on top performers; start with campaigns that are stable but not mission-critical.
Structure pools around markets or clusters.
For country tests: keep each market isolated (e.g. Germany as a standalone pool) or use clusters of comparable markets (e.g. Nordics, Benelux) to learn allocation across regions with similar dynamics.
Avoid mixing markets with radically different baseline CVR/ROAS, as PBA will over-weight cheaper ones.
Explore cross-platform where relevant.
Single-market, cross-platform pools (e.g. Germany across Meta, TikTok, Pinterest) are high-value tests for identifying the optimal channel mix.
Start with one platform if you want a controlled pilot, then expand to cross-platform once you’ve validated results.
Choose the right KPI.
Decide upfront which Fospha KPI to optimise toward: Conversions, Revenue, ROAS, or New-Customer ROAS.
One KPI per pool keeps optimisation clear; at most two can be combined (e.g. ROAS + Revenue for value-based outcomes).
Align with org structure and ownership.
Pools should reflect budget ownership in the team (e.g. by channel, market, or region).
Assign a clear budget owner for each pool to manage alignment with business goals.
Examples of Strong Test Candidates
Always-on conversion campaigns that today only get monthly adjustments but could benefit from continuous optimisation.
Cross-platform spend in a single high-value market (e.g. Germany across Meta/TikTok/Pinterest) to test channel allocation.
Clusters of mid-performing regions (e.g. Nordics, Benelux) where manual optimisation is minimal and PBA can surface hidden opportunities.
Underperforming or overlooked regions where teams lack time to actively optimise, but Fospha signals can guide reallocation.
Guardrails to configure
Parameter | Recommended Setting | Why It Matters |
Budget Mode | Fixed Daily | Safest first step; stable total spend |
Min/Max Caps | 10% min / 60% max | Prevents starvation or over-concentration |
Exploration Budget | 10% | Allows testing new allocations safely |
Optimisation Cadence | Meta/TikTok: every 2 daysPinterest/Snap: every 3–4 days | Avoid manual edits between cycles |
Separate Pools | Prospecting / Retargeting / App Installs | Keeps funnel stages independent |
KPI per Pool | ROAS, Revenue, Conversions, or New-Customer ROAS | Ensures clear optimisation goal |
Avoid these pitfalls:
Editing budgets manually during the first 7 days (resets learning)
Mixing prospecting & retargeting in one pool
No caps or inconsistent objectives
Changing KPI mid-test — start a new 4-week test instead
4. Designing the Pilot Test
4.1 Picking the Right Campaign to Test
Start with conversion-focused, always-on campaigns.
Ensure campaigns share a common KPI (conversion, revenue, or ROAS).
For country tests: split ad set = country.
For channel tests: group campaigns across Meta/TikTok/Pinterest.
Keep creative comparable; limit to 2–3 creative themes so PBA is not confounded by asset mix.
4.2 Budget Pools Setup Best Practice
Setting | Recommendation | Why |
Pool budget size | ≥ 5–10× expected daily CPA | Ensures enough signals for 7-day learning phase |
Min / Max allocation | Min ≥ 10%, Max ≤ 60% per channel/campaign | Prevents starvation or over-concentration |
Exploration budget | Start at 10% | Allows PBA to test new allocations without risk |
Campaign structure | Disable Meta CBO when pooling ad sets | Gives PBA budget control at pooled level |
Objective buckets | Separate Prospecting, Retargeting, App Installs | Prevents PBA over-optimising into remarketing |
4.3 Timeline
Phase | Duration | Key actions |
Warm-up | Days 1–7 | PBA learns variance; avoid manual budget edits |
Stabilisation | Weeks 2–3 | Monitor Smartly Budget Movement logs and Fospha KPIs |
Evaluation | Weeks 4–5 | Compare test vs control PoP in Fospha |
Scale/Pause | Weeks 6–8 | If results ≥10% uplift, scale to more markets/channels; if not, refine guardrails or extend test |
Expect PBA to adjust budgets every ~2 days on Meta/TikTok and every 3–4 days on Pinterest/Snap.
5. Measuring Test Results
Control design: Run a comparable campaign/market outside the pool on platform default optimisation.
Compare Test vs Control in Fospha:
Use PoP to track CPA, ROAS, new-customer share.
Validate if budget shifts matched incremental outcomes.
Directional validation: Look for divergence in trend (e.g. falling CPA in test vs flat control).
Success criteria:
Look out for ≥10% CPA reduction or ≥10% ROAS uplift vs control.
6. What To Do Next After the Pilot
Once the pilot completes, the goal is to decide whether to scale, refine, or re-run the test based on performance, learnings, and operational impact.
If Results Are Positive
Scale gradually.
Add more campaigns into the existing pool or replicate the setup in adjacent markets with similar structure and KPIs.
Expand scope.
Move from single-channel to cross-channel pools once stable results are proven.
Build funnel-based pools.
Separate Awareness, Consideration, and Conversion pools to enable more granular control heading into peak seasons.
Layer in creative and bid optimisation.
Once PBA is stable, introduce creative tests or audience variations to compound gains.
Document uplift.
Capture the quantitative impact (≥10 % ROAS uplift or ≥10 % CPA reduction) and qualitative learnings (e.g. improved channel mix, reduced manual edits).
If Results Are Neutral or Negative
Extend the learning window.
Keep the test running for 6–8 weeks to allow the model to stabilise and gather more signal.
Re-check setup integrity.
Was the KPI consistent across campaigns?
Were creatives comparable?
Was there enough daily volume for learning?
Adjust guardrails.
Increase exploration budget (e.g. 10 → 15 %), widen min/max constraints, or slow the optimisation cadence to reduce volatility.
Simplify scope.
Split large pools into smaller, clearer tests (e.g. per channel or market) to isolate issues.
Re-baseline expectations.
Performance stabilisation may take longer for smaller budgets or upper-funnel campaigns; focus on directional improvement first.