Uclic is a B2B data agency: we build the dashboards and metrics that tell the truth about your business. Data reconciled across your tools, KPIs defined with you, dashboards an executive can read without a manual. To steer on reliable numbers.
12 sub-services to go from manually entered figures to an executable dashboard.
From source audits to reverse ETL, each sub-service can be activated on its own or packaged into the full Stack plan. Tools named explicitly: Airbyte, Fivetran, BigQuery, Postgres, ClickHouse, dbt, Metabase, Looker Studio, Cube, Census, Hightouch.
Diagnostic
Full data audit
Inventory of the 12 to 25 sources, mapping of duplicates (same metric calculated three times), identification of gaps (retention not calculated), validation of the 5 to 10 priority COMEX questions. Costed roadmap, 6-10 weeks.
12+sources audited
AuditHubSpotStripeGA4
Dedicated expert
Alexis Christine-Amara
VP Sales
Scoping
Stack selection & provisioning
Deliberate warehouse selection (BigQuery, Postgres Supabase, ClickHouse), ELT (open-source Airbyte vs managed Fivetran), BI (Metabase, Looker Studio, Cube). Provisioning on GCP, AWS or Supabase depending on your existing infrastructure.
1 wk.stack provisioned
BigQueryPostgresAirbyteMetabase
Dedicated expert
Tuka Bade
Data Expert | Machine Learning, Python & SQL
Ingestion
ELT Airbyte / Fivetran
Connectors for HubSpot, Pipedrive, Stripe, GA4, Mixpanel, PostHog, Aircall, Intercom, and product data. Incremental CDC sync, schema change handling, Slack alerting on failures. Airbyte if self-hosted, Fivetran if managed.
Staging / intermediate / marts layers. Automated tests: unique, not_null, accepted_values, relationships. Auto-generated documentation with dbt docs. CI/CD with GitHub Actions on pull request. Every metric defined exactly once.
100%versioned metrics
dbt Coredbt CloudSQLMeshGit
Dedicated expert
Mathieu Chenal
Expert SEO & GEO
Acquisition
Acquisition dashboard
Acquisition channel, blended and per-channel CAC, multi-touch attribution (first / last / linear / data-driven), lead volume by source, quality by segment. GA4 + HubSpot + Stripe cross-referenced for LTV by channel.
8+channels tracked
GA4HubSpotAttributionCAC
Dedicated expert
Moussa Ball
AI Architect & Full-Stack Engineer
Pipeline
B2B pipeline dashboard
MQL → SQL → meeting → Win funnel, velocity per stage, conversion rate by segment, weighted ARR, quarterly forecast. View by SDR/AE owner, by industry, by deal size. Comparison against monthly/quarterly target.
5+stages tracked
HubSpotPipedriveSalesforceARR
Dedicated expert
Philippe Lacombe
Consultant SEA & Google Ads Senior | Google Partner
Retention
Retention & cohorts dashboard
Monthly retention cohorts, Net Revenue Retention (NRR), Gross Revenue Retention (GRR), churn by segment and by cause, expansion vs contraction, time-to-churn. View by plan, by industry, by acquisition cohort.
Sync of dbt models to HubSpot, Salesforce, Intercom, Marketo. Enriched lead score, churn-risk segment, account properties (ARR, plan, industry). Data finally leaves the warehouse to serve day-to-day operations.
Slack alerts on MRR drop > 5%, acquisition anomaly (channel volume), conversion drop per stage, ELT sync failure. Statistical anomaly detection via z-score or Prophet. No noise, only actionable signals.
10+metrics monitored
SlackAlertingAnomaliesMonitoring
Dedicated expert
Alexis Christine-Amara
VP Sales
Handover
Training & data dictionary
Training for your in-house data team or ops lead on dbt and Metabase. Model creation, test debugging, CI/CD deployment. Complete documentation versioned in your Git. Data dictionary: every column and every metric has a single definition.
100%self-sufficient team
TrainingDocumentationdbt docsGit
Dedicated expert
Tuka Bade
Data Expert | Machine Learning, Python & SQL
What we do differently in data & dashboards
Why a Uclic data stack rather than anywhere else?
The dashboards of 2026 are no longer Google Sheets recopied every Monday, nor Looker Studio plugged directly into HubSpot. Sources have multiplied, metrics need to reconcile, and the exec team wants to decide on one number: not three.
Last-write-wins ou règle métier · DLQ replay manuellive
BI handover·10 criteria before transfer
6/10
60% passed0 in progress · 4 to do
Alerting freshness > 1hOK
Alerting volume drop > 20%OK
dbt tests on critical tables142 tests
Lineage traced end-to-endOK
Data dictionary documented80%
BigQuery cost monitoring<on quote
Roles + IAM permissionsOK
Backup + DR planpending
Data team training2/4
6 / 10 items passed | 6/10 validated: training to be finalizedready
Benchmark12 mois roadmap
Step 01 / 07Data audit & source mapping
01W1 Audit1 week
Data audit & source mapping
1 week· Step 1
We start from a complete inventory: HubSpot, Pipedrive, Stripe, GA4, Mixpanel, product data, Notion, Airtable. We map the duplicates (the same metric calculated three times with three different results) and the gaps (retention that nobody calculates). We validate the priority business questions with the executive committee: blended CAC, LTV by channel, NRR by cohort. If the main question has no clear answer today, that's where we start.
Inventory of 12-25 sources
Mapping of duplicates and gaps
Validation of exec committee questions
Prioritized data roadmap
Deliverable· Data mapping + costed 6-10 week roadmap
02
02W1-W2 Stack1 week
Stack selection & provisioning
1 week· Step 2
Deliberate warehouse choice: BigQuery if volume > 100 GB or you need elastic scaling, Postgres on Supabase or Neon if < 50 GB with transactional needs, ClickHouse for high-volume analytics. ELT choice: Airbyte open source self-hosted (free, full control) or Fivetran (managed, more expensive but zero maintenance). BI choice: Metabase if self-hosted, Looker Studio if you run on the Google ecosystem, Cube.dev for in-product embedding. Provisioning on GCP, AWS or Supabase depending on your infrastructure.
Warehouse: BigQuery / Postgres / ClickHouse
ELT: Airbyte vs Fivetran
BI: Metabase vs Looker Studio vs Cube
Cloud provisioning (GCP/AWS/Supabase)
Deliverable· Provisioned stack + team access
03
03W2-W3 ELT2 weeks
ELT & source ingestion
2 weeks· Step 3
Connecting priority sources via Airbyte or Fivetran: incremental sync, schema change handling, Slack alerting on failures. Typical B2B sources: HubSpot (deals, contacts, companies, engagements), Stripe (charges, subscriptions, invoices, customers), GA4 (events, sessions via BigQuery export), Mixpanel or PostHog (product events), Aircall (calls), Intercom (conversations). Frequency: hourly for HubSpot/Stripe, daily for GA4.
Airbyte/Fivetran connectors
Incremental sync via CDC
Slack alerting on failure
Schema documentation
Deliverable· 12+ sources synced into the warehouse
04
04S3-S5 dbt2 weeks
dbt modeling & data dictionary
2 weeks· Step 4
dbt layers structured into staging (raw data cleaned and typed), intermediate (joins and business logic: customer_id unification, multi-touch attribution) and marts (business tables ready to expose: fct_pipeline, dim_customer, fct_revenue). Systematic dbt tests: unique, not_null, accepted_values, relationships. Auto-generated documentation via dbt docs: every column and every metric has a single versioned definition in Git.
Staging / Intermediate / Marts
Automated dbt tests
Git-versioned data dictionary
dbt CI/CD on PR
Deliverable· Versioned dbt repo + data dictionary
05
05S5-S7 BI2 weeks
Actionable dashboards
2 weeks· Step 5
5 to 10 Metabase or Looker Studio dashboards organized by function. Acquisition: channel, blended CAC, multi-touch attribution. Pipeline: MQL → SQL → Meeting → Win funnel, velocity, conversion by stage, weighted ARR. Retention: monthly cohorts, NRR, GRR, churn by segment. Finance: ARR, MRR, LTV, CAC payback. Every dashboard answers a specific operational question: no decorative charts, no vanity KPIs.
Acquisition dashboard (CAC, attribution)
Pipeline dashboard (funnel, velocity)
Retention dashboard (cohorts, NRR)
Finance dashboard (ARR, LTV, payback)
Deliverable· 5-10 actionable dashboards delivered
06
06S7-S8 Reverse ETL1 week
Reverse ETL & tool activation
1 week· Step 6
Syncing dbt models into operational tools via Census or Hightouch. Enriched lead score (intent + fit) pushed into HubSpot for SDR prioritization. At-risk churn segment sent to Intercom for a targeted campaign. Account properties (ARR, plan, industry) synced into Salesforce. Data finally leaves the warehouse to serve operations: not just to make charts.
Census or Hightouch
Sync warehouse → HubSpot
Sync warehouse → Intercom
Sync warehouse → Salesforce
Deliverable· 3-5 reverse ETL syncs in production
07
07S8-S10 Training1-2 weeks
Alerting, handover & training
1-2 weeks· Step 7
Slack alerting on critical metrics (MRR drop > 5%, acquisition anomaly, drop in conversion by stage). Training for your internal data team or an ops lead on dbt and Metabase: model building, test debugging, CI/CD deployment. Full documentation handed over and versioned in your Git. You keep ownership of the stack, the models and the dashboards: no vendor lock-in, no Uclic license.
Three transparent entry points. A data audit to frame the roadmap, a dashboard build to ship the essentials in 4-6 weeks, or a complete turnkey stack with dbt and reverse ETL in 6-10 weeks. No hidden subscription to a cloud vendor — you keep full ownership.
Diagnostic
Audit
No commitment
0€
We review your 3 pillars (Inbound, Outbound, AI & Dev) and deliver a costed set of recommendations.
Inbound, Outbound, AI & Dev auditAnalysis of your acquisition channels (SEO, Ads, Content), your outbound prospecting and your AI/automation stack.
Score across the 3 pillarsA /100 score per pillar: Inbound (demand capture), Outbound (prospecting) and AI & Dev (industrialization).
Complete audit of your sources, existing models, and dashboards in place. Identification of duplicates, gaps, and priorities. A costed roadmap to decide what comes next.
Mapping of 12-25 sourcesInventory of HubSpot, Pipedrive, Stripe, GA4, Mixpanel, PostHog, Notion, Airtable, and product data. Identification of missing connectors and duplicate data.
Audit of existing dashboardsReview of the Google Sheets, Looker Studio, Metabase, and Tableau setups in place. Identification of metrics calculated multiple times with divergent definitions.
Validation of business questions90-minute workshop with the exec team to frame the 5 to 10 priority questions (blended CAC, NRR, LTV by channel, multi-touch attribution).
Stack recommendationInformed recommendation on warehouse (BigQuery vs Postgres vs ClickHouse), ELT (Airbyte vs Fivetran), and BI (Metabase vs Looker Studio vs Cube), based on volume and budget.
Essential data stack delivered turnkey in 4-6 weeks: ELT + warehouse + 3 to 5 Looker Studio or Metabase dashboards, plus team training.
Data audit includedSource mapping and validation of business questions with the exec team, included in the plan.
ELT with 5-8 sources connectedAirbyte or Fivetran connecting HubSpot, Stripe, GA4, Mixpanel/PostHog, and 1-2 specific sources. Incremental sync, Slack alerting.
Warehouse provisionedBigQuery (free sandbox, 1 TB/month) or Postgres on Supabase / Neon depending on volume. Typed schemas, team access configured.
Complete, industrial-grade data stack in 6-10 weeks: ELT, warehouse, versioned dbt, 5 to 10 dashboards, alerting, and reverse ETL to activate your data in HubSpot/Intercom/Salesforce.
Audit + dashboard build includedEverything in the Build plan, plus the dbt layer and reverse ETL to go industrial.
Industrialized ELT across 12+ sourcesAirbyte or Fivetran connecting your entire toolset: CRM, payments, analytics, product, support, calls. CDC, incremental sync, monitoring.
Git-versioned dbt modelingStaging / intermediate / marts layers. Automated tests (unique, not_null, relationships). Auto-generated documentation via dbt docs. CI/CD on pull requests.
5 to 10 dashboards per functionAcquisition, pipeline, retention, finance, product. Every dashboard answers an operational question — no vanity KPIs.
Reverse ETL with Census/Hightouch3 to 5 syncs from warehouse → operational tools. Lead scoring, churn-risk segments, enriched account properties in HubSpot/Salesforce/Intercom.
Slack alerting & ops handoverAlerts on MRR drops, acquisition anomalies, and conversion drops. Training for your in-house data or ops team. Fully versioned documentation.
Growth and automation expertise followed by B2B leaders.
Founders, executives and growth leads publicly react to our analysis and methods on LinkedIn. Each card opens the original conversation.
“
Interesting. Personally, I did the opposite in a way. That said, I always keep the code to create APIs that will be used by n8n. So in the end it's a mix of both.
I also think it depends on what you want to do in the long term.
On the other hand, one thing is certain: if you want total flexibility, knowing how to code is important. (I include vibe coding)
Completely agree ! The era of "just good enough" is over. In a saturated market, it's essential to stand out from day one. An MVP certainly needs to be viable and minimal, but that doesn't mean it has to be a stripped-down version of the final vision. Investing in a quality user experience from the start can significantly accelerate acquisition
incredible, I just gave it a directory with hundreds of software I want to scrape (across about ten pages in total). Fully autonomously, it retrieved everything and ran a google search for each software to find the url of the LinkedIn company page.
All of that with just 1 single 3-line prompt and zero errors!
Thanks for the reco!
I've been in the biz for 20+ years. Code was always the missing piece for me as an SEO. But I've got some solid knowledge. That said, vibe coding opened up other doors for me — just huge ones. Today I've gotten into React, Vite, Node, Python, and I code whatever I need. 😍
Incredible! AI is truly becoming an accelerator for automating web tasks, freeing up time and daring to take on new projects. For anyone interested in dropshipping, automation and e-commerce, it's a real revolution.
Thank you for this very informative share. It's impressive to see the growing number of developers and how dominant young people are in this sector. It really shows how essential technology and development are for our future.
Géraud L.
AI Entrepreneur & Personal Coach for Founders and Leaders
Interesting! This paves the way for the future (potentially huge) market of AI SEO. Companies and people who position themselves on it quickly and effectively will have a definite competitive advantage over the others.
Maxime H.
I turn your innovation and entrepreneurship programs into concrete success
I love this kind of hack Wladimir 🙏🏻
Do you have any hacks on the student email part?
I was thinking of buying a .school domain name, making a mini landing page for a fake school, generating an email.
Because .edu is out of reach 😅
Great, thanks for the tip! I just tested it and it's Veo 2 that's offered in the student plan on the US site (which is indeed free for 12 months). That said, there is indeed access to Veo 3 (in a limited version) :)
This kind of tool truly changes the game by automating complex tasks with incredible simplicity.
A new era is setting in where AI becomes a truly ultra-efficient personal assistant
Matthieu S.
CEO Mindsales - I generate 8 to 33 meetings/month for you - 📩 DM me “Setter” to lea
Interesting. Personally, I did the opposite in a way. That said, I always keep the code to create APIs that will be used by n8n. So in the end it's a mix of both.
I also think it depends on what you want to do in the long term.
On the other hand, one thing is certain: if you want total flexibility, knowing how to code is important. (I include vibe coding)
Completely agree ! The era of "just good enough" is over. In a saturated market, it's essential to stand out from day one. An MVP certainly needs to be viable and minimal, but that doesn't mean it has to be a stripped-down version of the final vision. Investing in a quality user experience from the start can significantly accelerate acquisition
incredible, I just gave it a directory with hundreds of software I want to scrape (across about ten pages in total). Fully autonomously, it retrieved everything and ran a google search for each software to find the url of the LinkedIn company page.
All of that with just 1 single 3-line prompt and zero errors!
Thanks for the reco!
I've been in the biz for 20+ years. Code was always the missing piece for me as an SEO. But I've got some solid knowledge. That said, vibe coding opened up other doors for me — just huge ones. Today I've gotten into React, Vite, Node, Python, and I code whatever I need. 😍
Incredible! AI is truly becoming an accelerator for automating web tasks, freeing up time and daring to take on new projects. For anyone interested in dropshipping, automation and e-commerce, it's a real revolution.
Thank you for this very informative share. It's impressive to see the growing number of developers and how dominant young people are in this sector. It really shows how essential technology and development are for our future.
Géraud L.
AI Entrepreneur & Personal Coach for Founders and Leaders
Interesting! This paves the way for the future (potentially huge) market of AI SEO. Companies and people who position themselves on it quickly and effectively will have a definite competitive advantage over the others.
Maxime H.
I turn your innovation and entrepreneurship programs into concrete success
I love this kind of hack Wladimir 🙏🏻
Do you have any hacks on the student email part?
I was thinking of buying a .school domain name, making a mini landing page for a fake school, generating an email.
Because .edu is out of reach 😅
Great, thanks for the tip! I just tested it and it's Veo 2 that's offered in the student plan on the US site (which is indeed free for 12 months). That said, there is indeed access to Veo 3 (in a limited version) :)
This kind of tool truly changes the game by automating complex tasks with incredible simplicity.
A new era is setting in where AI becomes a truly ultra-efficient personal assistant
Matthieu S.
CEO Mindsales - I generate 8 to 33 meetings/month for you - 📩 DM me “Setter” to lea
4 years, still number 1 in the rankings despite all the bosses who came after you. Delcros will remain the undisputed master and pioneer, always one step ahead of the market. 👏
Finally an interesting take.
N8N is great for understanding a workflow, it's very graphic, very visual, but if you want to do slightly more elaborate stuff, code is way better.
Christophe R.
Web developer // nextjs :: node :: adonis :: astrojs :: directus and more...
While everyone fights to master algos they only half understand, other channels remain wide open. And strangely enough, no one seems interested in them.
Fatou G.
We SCALE your business thanks to a MultiChannel client acquisition strategy and u
I really like this thinking. No-code has its uses for testing or moving fast, but it's true that code lets you take things to the next level. Thanks for sharing 🙌
Interesting to see this shift from no-code to code. Your thinking makes it clear that the tool matters less than the end goal: automating, optimizing, generating value.
4 years, still number 1 in the rankings despite all the bosses who came after you. Delcros will remain the undisputed master and pioneer, always one step ahead of the market. 👏
Finally an interesting take.
N8N is great for understanding a workflow, it's very graphic, very visual, but if you want to do slightly more elaborate stuff, code is way better.
Christophe R.
Web developer // nextjs :: node :: adonis :: astrojs :: directus and more...
While everyone fights to master algos they only half understand, other channels remain wide open. And strangely enough, no one seems interested in them.
Fatou G.
We SCALE your business thanks to a MultiChannel client acquisition strategy and u
I really like this thinking. No-code has its uses for testing or moving fast, but it's true that code lets you take things to the next level. Thanks for sharing 🙌
Interesting to see this shift from no-code to code. Your thinking makes it clear that the tool matters less than the end goal: automating, optimizing, generating value.
Our B2B data & dashboards agency works with clients across France
100% remote-first with weekly video meetings: we serve B2B SMEs, mid-market companies and scale-ups in Paris, Lyon, Bordeaux, Marseille, Lille, Nantes, Toulouse, Strasbourg, and beyond (Geneva, Brussels).
B2B data & dashboards agency FAQ: stack, tools, modeling, pricing
Everything to know before starting a data engagement with Uclic: BigQuery vs Postgres, Metabase vs Looker Studio, dbt modeling, reverse ETL, project pricing and 6-10 week timelines.
What makes a good KPI dashboard?
A good KPI dashboard answers a CEO's questions at a glance: where the business stands, what's drifting, where to act. That requires data reconciled across tools (CRM, billing, marketing), precisely defined metrics, and automatic updates. At Uclic, every dashboard is connected to your real data sources and delivered with its own documentation: no more debating the numbers in management meetings.
BigQuery vs Postgres: which one for 50 GB of data?
At 50 GB, both work very well and the choice depends on context. Postgres (Supabase, Neon, RDS) is unbeatable if you already run an application-level Postgres, if you need transactional workloads, or if you want to minimize cost (a $25/month Supabase Pro plan is enough up to 100 GB). BigQuery becomes relevant beyond 100 GB, if you need elastic scaling, or if your main source is GA4 (free native export). The BigQuery sandbox is free up to 1 TB of queries/month: more than enough to get started. Our recommendation: Postgres if < 50 GB with transactional needs, BigQuery if > 100 GB or heavy GA4 volume. ClickHouse only for high-volume analytics (product events > 1 billion rows).
Client-side or warehouse-side dashboards?
Warehouse-side, no hesitation. A client-side dashboard (Looker Studio connected directly to Google Sheets, Metabase on HubSpot via native connector) crashes beyond 50,000 rows, breaks with every schema change on the vendor side, and forces you to repeat every metric calculation in every chart: hence three different definitions of the same MQL → SQL conversion rate. The warehouse + dbt layer solves this: every metric defined once in Git, automatically tested, exposed to multiple BI tools at once. The initial overhead (1 to 2 weeks of modeling) pays for itself within 3 months in avoided maintenance.
Looker Studio is free, so why pay for Metabase?
A false dilemma. Looker Studio is free but limited: degraded performance beyond 100k rows, complicated granular sharing, no clean embedding, Google ecosystem lock-in. Metabase Open Source is also free self-hosted (Docker, < on quote on DigitalOcean), with much better performance, group-based sharing, signed embedding, and native SQL for power users. Metabase Cloud Pro ($85/month for 5 users) adds SSO, alerting, and fine-grained permissions. Our rule: Looker Studio if you're on the Google ecosystem with basic business users, Metabase if self-hosted or if you need SQL and granular sharing, Cube.dev for product embedding. No vendor is better in absolute terms.
How many synced sources, and at what frequency?
An average B2B SME has 12 to 25. Priority sources (included in the build): HubSpot/Pipedrive/Salesforce (CRM, hourly sync), Stripe (payments, hourly sync), GA4 (analytics, daily sync via native BigQuery export), Mixpanel or PostHog (product events, daily sync), Intercom or Zendesk (support, hourly sync), Aircall (calls, daily sync). Secondary sources (depending on usage): Notion, Airtable, Google Ads, Meta Ads, LinkedIn Ads, internal product database. Frequency: hourly for CRM/payments (critical deals), daily for analytics and events (volume), real-time only for in-product embedding (Cube.dev streaming). Incremental CDC sync wherever available: no daily full refresh on 10 million rows.
Do you document the data model (data dictionary)?
Yes, and it's non-negotiable. Every table, every column, every business metric is documented: definition, source, formula, owner, last updated. The docs are auto-generated via dbt docs from YAML files versioned in Git, and deployed internally (often on a subdomain like docs-data.votreboite.com). When a sales rep asks "what exactly is NRR in this dashboard?", they get the answer in 3 clicks: no need to open a Slack thread. This documentation is the real barrier against the data mess that sets in within 12 months if nobody writes it down. Versioned in your Git, it remains entirely your property.
Why do you avoid Snowflake?
Not dogma: relevance. Snowflake is technically excellent and unavoidable beyond 5 TB or in multi-cloud setups. But for 90% of B2B SMEs and mid-market companies (volume < 500 GB), it's overkill and expensive: credit-based billing that's hard to predict, a high monthly minimum cost, and governance complexity that requires a full-time data engineer. Postgres or BigQuery are enough at 5 to 10 times lower cost, with an identical developer experience on the dbt side. We use Snowflake when the client has already invested in it, or when volume + multi-cloud justify it: never by default.
How much does a Uclic data stack cost?
Three entry points, all transparent: Data audit (quote-based) (source mapping, audit of existing dashboards, costed roadmap). Dashboard build (quote-based) (audit + ELT for 5-8 sources + warehouse + 3-5 Looker Studio/Metabase dashboards + training, delivered in 4-6 weeks). Full stack (quote-based) (build + versioned dbt + 5-10 dashboards + Census/Hightouch reverse ETL + Slack alerting + handover, delivered in 6-10 weeks). For continuous production after the build, switch to a Modular Stack plan, 2,000 to quote-based (data ops over 6 months). No hidden subscription to a cloud vendor: you keep full ownership of the stack, the dbt models and the dashboards.
How long before you have dashboards in hand?
Standard rollout plan: 1 week for audit and stack scoping, 1 to 2 weeks for ELT and warehouse provisioning, 2 weeks for dbt modeling, 2 weeks for dashboards. For the Dashboard build (quote-based) plan without dbt, plan for 4 to 6 weeks from brief to handover. For the Full stack (quote-based) with dbt and reverse ETL, plan for 6 to 10 weeks depending on the number of sources. First working dashboard usable internally by week 3 in most cases. No magic method: dbt modeling takes time, but it's what separates a stack that scales from a throwaway dashboard rebuilt 6 months later.