Improvado

Editor’s note: This Improvado review is based on my hands-on test of Improvado’s platform and its MCP server in October 2026. I weighed pricing and plan caps; connectors and extraction; transformation and governance; the AI Agent and MCP server; and security.

Quick verdict: I recommend Improvado for marketing analytics teams and agencies that need harmonized metrics and governance, with Starter from $20,000/year and Growth from $41,700/year. Plans cap rows, AI Agent actions, and governance rules; there is no free plan or trial, and Starter syncs once a day.

Improvado loads marketing and sales data into your warehouse or a BigQuery instance managed by Improvado. The Marketing Common Data Model (MCDM) harmonizes that data, with an AI Agent and an 84-tool MCP server providing conversational access. In this Improvado review, I’ll take a closer look at pricing, features, and setup so you can see exactly whether its managed marketing-data service fits your team.

Key Takeaways

• Four plans: MCP Only costs $100/month, Starter starts at $20,000/year, Growth starts at $41,700/year, and Enterprise requires a quote. No trial or free plan is offered.
• All connectors and destinations are included on every plan, without per-connector fees; destination connection counts remain contractual.
• Starter has tight caps: 20M unique rows/year, daily sync, 20 AI Agent actions/day, and 3 governance rules.
• The MCP server provides 84 tools for clients including Claude, ChatGPT, Gemini CLI, and Copilot.
• SSO requires Enterprise or a Growth add-on. HIPAA is included from Growth, not Starter; managed plans are sold through a demo.

Pros and Cons

How Much Does Improvado Cost?

Improvado’s managed plans start at $20,000/year, but MCP Only costs $100/month for connecting your own AI assistant to marketing data. Starter, Growth, and Enterprise are sold through a demo; none offers a free trial.

• MCP Only ($100/month): For marketing data inside your own AI assistant. It includes 2M unique rows/year, daily sync, 300 MCP actions/week, and 1 workspace, but no AI Agent actions or governance rules.
• Starter (from $20,000/year): For one team needing managed data. It includes 20M unique rows/year, daily sync, 20 AI Agent actions/day, 400 MCP actions/week, 3 governance rules, and 1 workspace.
• Growth (from $41,700/year): For agencies and multi-brand teams needing up to 10 workspaces. Limits range from 60M to 200M unique rows/year, daily to twice-daily sync, 40 to 90 AI Agent actions/day, 800 to 1,800 MCP actions/week, and 10 to 25 governance rules. Premium Customer Support, HIPAA, and 120 to 240 annual Professional Services credits are included; SSO costs extra.
• Enterprise (by quote): For larger deployments needing 1B unique rows/year, custom sync, 600 AI Agent actions/day, 12,000 MCP actions/week, and 100 governance rules. It includes unlimited workspaces, SSO, HIPAA, and 960 annual Professional Services credits.

Limit or feature	MCP Only	Starter	Growth	Enterprise
Price	$100/month	From $20,000/year	From $41,700/year	Quote
Unique rows/year	2M	20M	60M to 200M	1B
Sync frequency	Daily	Daily	Daily to twice daily	Custom
AI Agent actions/day	None	20	40 to 90	600
MCP actions/week	300	400	800 to 1,800	12,000
Governance rules	None	3	10 to 25	100
Workspaces	1	1	Up to 10	Unlimited
SSO	Not included	Not included	Add-on	Included
HIPAA	Not included	Not included	Included	Included

Improvado pricing follows plan and data volume, not a share of ad spend. Its volume allowance counts unique rows per year, and every plan includes all connectors and destinations without per-connector fees.

Growth and Enterprise alone support these add-ons:

• Advanced Analytics bundles, from $20,000/year: Marketing mix modeling, incrementality testing, and multi-touch attribution.
• Creative Analytics, from $4,000/year: Creative analysis with briefing tabs and AI-assisted custom tags added in October 2026.

Is Improvado Good Value for Money?

• Connector breadth favors Improvado: Every connector is included, unlike Fivetran’s per-connection billing. Annual unique rows still constrain the plan.
• Basic reporting favors cheaper options: Supermetrics starts at EUR 39/month billed yearly; Funnel starts at EUR 280/month billed annually. I cannot justify Starter for connector access alone.
• Managed modeling strengthens the case: Improvado earns its place when harmonization and governance are buying requirements, rather than extras for a simple extraction job.
• Growth has variable allowances: The demo establishes the applicable row tier within its 60M to 200M range, alongside action and governance limits.

Reviewer’s Notes: I recommend Starter only for one team that fits daily updates and 20M annual rows. Growth is my agency pick for separate workspaces. For Claude or ChatGPT access without the managed service, choose MCP Only at $100/month rather than treating Starter as the entry point.

My Experience With Improvado

I authorized Google Ads before configuring extraction; the connection alone did not define which reports reached BigQuery. Improvado used a four-step extraction wizard, with loading configured separately.

Connecting My First Ad Account

I opened Data sources > Make a new connection, selected the Google Ads tile, and followed OAuth authorization. Facebook also supports OAuth; other connections use login credentials or API tokens, with Criteo requiring additional tokens.

From Connected sources, I selected Extract on the connection to configure data collection. The extraction templates determined the fields collected from the authorized account.

Setting Up Extraction

1. Select accounts: I chose accounts for extraction. Pagination starts after 10 accounts, and selection persists across pages.
2. Choose templates: I used the prebuilt Default Configuration option. Custom templates are also available, and accounts multiplied by templates must remain at 1,000 extraction orders or fewer.
3. Configure extraction: I set the sync start date for historical data and a daily schedule. I reviewed the time zone and lookback settings before selecting dimensions and metrics.
4. Review orders: I reviewed the generated extraction orders before confirming. Duplicate handling offers Ignore duplicates or Merge duplicates.

Two templates based on the same global extraction template cannot be combined in the selection. Each account-and-template pair creates one extraction order, so adding templates consumes the wizard’s allowance as well as adding accounts.

The lookback options were Live, Default, Max, and Custom. Custom supports a window from 1 day up to the available maximum; the “last day” and “last day (inc)” templates do not support lookback.

I checked the data table name before confirmation because table names cannot be changed afterward. Improvado uses the naming pattern “Data Source – Extraction Template.”

Weekly and monthly schedules are available, but hourly extraction requires a Customer Success Manager (CSM) request. A new table appears in Data Explorer in about 5 minutes, and the Dataflow dashboard shows extraction progress.

Loading Into My Warehouse

I opened Destinations > Add a new destination, selected BigQuery, and completed the connection fields. Improvado also offers a BigQuery instance managed by Improvado.

On the Automation tab, I enabled Use as default destination, which creates a load order for each new data table. Product models are excluded from automatic loading.

I reviewed the historical-load option and destination object name. Setup load also lets you select tables and load time; loading starts about 15 to 20 minutes after the table appears, with batches recommended for 100M+ rows.

Reviewer’s Notes: Settle hourly sync with the CSM and request deeper history or custom connectors through the Service Desk before signing. I see those request-based controls as the main setup trade-off, not self-serve settings.

Connectors, Extraction and Destinations

Improvado includes 1,400+ connectors across 15 categories on every plan, but that catalog does not remove annual row caps or contractual destination limits. I value the breadth for agencies connecting different client platforms.

Custom connectors are built and maintained on request through the Service Desk. The New API connector request requires:

• Data source URL: Identify the system to connect.
• API details: Provide an API reference link or vendor representative contact.
• Sample file: Show the data the connector needs to return.

Credentials, supplied through a shared account or token, are optional. No delivery turnaround is published, so a custom connector is a request-based implementation dependency rather than an immediately available integration.

Historical backfill reaches up to 10 years, with each platform’s API determining the available depth. History beyond the preset maximum requires a Service Desk request, not simply an earlier start date.

Auto-extraction picks up newly added accounts, while Improvado handles pagination, rate limits, and token refresh. I value those managed responsibilities more than catalog size when choosing a service to reduce connection maintenance.

Extraction schedules include:

• Hourly by request: Every 1, 2, 3, 4, 6, 8, or 12 hours through the CSM.
• Daily: The standard frequency on MCP Only and Starter.
• Weekly or monthly: On a selected weekday or calendar day.

One schedule is created by default; additional schedules are capped by contract. Growth’s standard allowance remains daily to twice daily.

Destination choices include:

• Cloud warehouses: Snowflake, BigQuery, Databricks, Redshift, Microsoft Fabric, and ClickHouse.
• Azure targets: Blob Storage, Data Lake, and Azure SQL.
• Databases: PostgreSQL, Heroku PostgreSQL, MySQL, Microsoft SQL Server, and SAP HANA.
• Storage and files: S3, Google Drive, Google Cloud Storage, FTP, and SFTP.

Unlike Fivetran’s per-connection billing, Improvado charges no per-connector fees. However, the destination count remains contractual: Add a new destination becomes disabled when that cap is reached.

Transformation, the MCDM and Governance

Google Ads calls its spending metric “Cost”; Meta calls it “spend.” Improvado’s Marketing Common Data Model (MCDM) harmonizes 46,000+ metrics, resolving those two labels to the same field.

I rate that normalization above raw connector breadth for cross-channel reporting. Improvado also supports custom field mapping, multi-currency normalization, and naming convention decoding.

I opened Recipe templates, picked Paid Social, and named the model and output table. Available templates also cover:

• Cross-Channel: Reporting across marketing channels.
• Paid Search: Search advertising data.
• eCommerce: Commerce reporting.
• SEO: Organic search reporting.
• Paid Ads Google Analytics: Advertising and analytics data.

The Cross-Channel Wizard provides another route into modeling. Within recipes, the no-code operators include:

• Blend and Join: Combine data.
• Filter and Group: Shape the reporting dataset.
• Custom Formula: Define calculated fields.

I used Preview before saving and kept the recipe in test status rather than moving it straight to production. Improvado saves recipe versions automatically, and recipes can become reusable templates across workspaces.

That reuse strengthens my Growth recommendation for agencies: the same modeling approach can serve separate client workspaces. Output can feed Looker Studio, Tableau, or Power BI.

Marketing Data Governance

I used AI chat to create governance rules and assign owners and severity levels. The rule library covers naming conventions, budget pacing, and configuration checks.

Marketing Data Governance provides:

• Pre-flight validation: Check campaign naming, targeting, and budget before launch.
• Notifications: Send rule alerts to Slack, Teams, or email.
• Daily dashboard: Monitor compliance and address violations.
• Naming Conventions: Apply automatic name fixes and sync standardized names back to ad platforms.

Starter includes 3 rules, Growth includes 10 to 25, and Enterprise includes 100; MCP Only includes none.

Reviewer’s Notes: I see Starter’s 3 rules as room for a naming convention and a pacing check, not a full marketing taxonomy. Choose Growth when governance spans separate client workspaces.

The AI Agent and the MCP Server

Starter’s 20 AI Agent actions/day do not mean 20 questions. One tool call consumes one action; a typical question uses 3 to 5 actions, leaving roughly 4 to 6 questions daily before other Agent work.

The AI Agent’s tool types are:

• Query Data
• Visualization
• Web Search
• Image Analysis
• Custom Tools

The Usage tab forecasts “on track,” “at risk,” or “limit reached.” Usage rules can cap Interactive and Automated consumption separately.

The AI Agent shares only metadata, such as table and column names, with LLMs and has a zero-training policy. It can run inside your own Snowflake or BigQuery environment.

Knowledge Graph supplies account-level and workspace-level files the Agent reads before answering:

• Markdown and text
• JSON and YAML
• CSV

Built-in skills include a full marketing account review and a weekly creative performance review. October 2026’s Automations combines scheduled Agent work and data pipelines in one list; pipelines are built through chat.

Connecting the MCP Server

I opened Customize > Connectors > Add custom connector in Claude Desktop, added the Improvado MCP server endpoint, and authorized with my Improvado credentials through OAuth.

The server provides 84 tools with read and write access, with confirmation required before every account change. Its zero-trust proxy keeps credentials hidden from connected agents.

Other connection routes include:

• Claude Code: Use claude mcp add --transport http with the server name and endpoint, then authorize through OAuth on first use.
• ChatGPT and Codex: In Apps, select Improvado AI Agent > Connect. That authorization also works in Codex and Codex CLI.
• Gemini CLI and Antigravity: Both connect; the Gemini web app cannot.
• Microsoft Copilot: Copilot Studio and Microsoft 365 Copilot require an OAuth application, including client ID and secret, from the CSM.

MCP weekly allowances are 300 on MCP Only, 400 on Starter, 800 to 1,800 on Growth, and 12,000 on Enterprise.

Security and Support

Improvado is SOC 2 Type II certified on every plan, but HIPAA is included only on Growth and Enterprise. Request the SOC 2 Type II report through the Trust Center under NDA.

HIPAA BAAs are available, and Improvado requires a BAA before handling protected health information. GDPR and CCPA commitments include a DPA with EU Standard Contractual Clauses and 30 days’ written notice before new subprocessors.

SSO through Azure or Okta is included on Enterprise and available as a Growth add-on. I exclude Starter when SSO is a mandatory purchasing requirement.

• Customer Success Manager: Coordinates requests such as hourly extraction.
• Service Desk: Handles requests, follow-ups, and custom connector submissions.
• Premium Customer Support: Included on Growth and Enterprise.
• Hands-on Professional Services: Included only on Enterprise.

Growth includes 120 to 240 annual Professional Services credits, while Enterprise includes 960. Starter and MCP Only include none; Growth’s credits do not include the Enterprise-only Hands-on Professional Services entitlement.

How Does Improvado Compare to Competitors?

I recommend Improvado for managed, governed marketing data when the budget supports at least $20,000/year, not as the cheapest way to connect an ad account.

• Fivetran: My pick when a free entry point and connection-based usage billing fit the purchase. It has a free plan and bills on monthly active rows per connection. Standard adds a $5 base charge per connection; a median Google Ads connection costs about $48.67/month. Improvado instead meters unique rows annually.
• Supermetrics: My budget pick for a small team with a narrow data requirement. Starter begins at EUR 39/month billed yearly for 2 data sources and 1 user. Growth begins at EUR 159/month billed yearly for 7 data sources and 2 users; Enterprise requires a quote. Choose it over Improvado when those allowances fit and managed governance is not a buying requirement.
• Funnel: My alternative when a trial and included MCP access matter. Starter begins at EUR 280/month billed annually, including 117 connectors, 13 destinations, MCP, and Funnel AI. Business begins at EUR 560/month billed annually with 579 connectors and 46 destinations. Its 30-day free trial offers an evaluation route that Improvado’s managed plans lack.
• Windsor.ai: My entry-price pick. Its free plan supports 1 user and 1 source, Basic costs $19/month billed annually, and its catalog contains 350+ sources. Start here when affordable connectivity matters more to your purchase than Improvado’s managed modeling and rule-based governance.
• Adverity: My quote-based marketing ETL alternative. Adverity Connect covers marketing ETL, and pricing is quote only. Request an Adverity proposal alongside Improvado when your procurement process already includes a custom enterprise evaluation rather than a subscription with a fixed entry price.

For an Improvado vs Supermetrics decision, I prioritize managed governance versus a small-team starting price. For Improvado vs Fivetran, the billing distinction is annual unique rows versus monthly active rows per connection; count your expected usage against the relevant allowance.

Comparison	Improvado	Funnel	Supermetrics	Fivetran
Entry price	MCP Only $100/month; managed Starter from $20,000/year	Starter from EUR 280/month billed annually	Starter from EUR 39/month billed yearly	Free plan; Standard $5 base charge per connection plus usage
Paid-plan example	Growth from $41,700/year	Business from EUR 560/month billed annually	Growth from EUR 159/month billed yearly	Median Google Ads connection about $48.67/month
Allowance or billing basis	Starter: 20M unique rows/year	Starter: 117 connectors, 13 destinations	Starter: 2 data sources, 1 user	Monthly active rows per connection

How I Reviewed Improvado

I followed Google Ads authorization through extraction and BigQuery load setup, then used the Paid Social recipe and governance-rule controls. My walkthrough covered the four-step extraction wizard and default destination behavior, including the exclusion of product models.

I weighed pricing against annual row allowances and workspace limits, rather than making the 1,400+ connector count the deciding factor. AI Agent action metering and weekly MCP allowances informed my recommendations for conversational access.

I queried my data through the AI Agent and the Improvado MCP server in Claude Desktop, where I connected through OAuth and assessed its read/write scope. Security and support influenced my plan choices, particularly HIPAA and SSO eligibility.

My competitor comparison covered:

• Fivetran’s connection-based billing.
• Supermetrics and Windsor.ai entry prices.
• Funnel’s trial and plan allowances.
• Adverity’s quote-based marketing ETL.

Prices current as of October 2026.

Should You Use Improvado for Your Marketing Data?

I recommend Improvado when managed harmonization and governance justify the contract, not merely because it has 1,400+ connectors. The MCDM and reusable recipes address cross-channel modeling, while the AI Agent and MCP server add conversational access with separate usage caps.

Growth is my agency pick for separate workspaces and more than 3 governance rules. It includes HIPAA, although SSO remains an add-on; Starter fits one team that can stay within 20M annual rows and daily updates.

MCP Only at $100/month suits paid access through an existing AI assistant. It includes neither AI Agent actions nor governance rules, so it does not replace a managed-plan evaluation.

Skip Improvado if a self-serve trial or hourly updates without negotiation are mandatory. Supermetrics and Windsor.ai suit smaller budgets; Funnel offers a 30-day trial.

Before contacting Improvado, count your connected platforms and annual unique rows. List the workspaces and governance rules you need, and identify SSO or HIPAA requirements. Use those figures to choose MCP Only or request a managed-plan demo with the relevant limits already in hand.

FAQs

How much does Improvado cost?

Improvado costs $100/month for MCP Only, with Starter from $20,000/year, Growth from $41,700/year, and Enterprise by quote. Pricing follows plan and data volume, not a percentage of ad spend. Growth covers 60M to 200M unique rows/year, with the applicable tier established during the demo rather than guaranteed by its starting price.

Does Improvado have a free trial?

Improvado offers neither a free trial nor a free plan. Starter, Growth, and Enterprise are sold through a demo, while MCP Only is a paid $100/month option for connecting an existing AI assistant. I recommend Funnel instead when a free hands-on evaluation is mandatory: it offers a 30-day trial.

How many connectors does Improvado have?

Improvado has 1,400+ connectors across 15 integration categories, with all connectors included on every plan and no per-connector fees. Request custom connectors through the Service Desk with a data source URL, API reference or vendor contact, and sample file. No turnaround is published, and annual row limits still apply.

Can Improvado load data into my own warehouse?

Yes, Improvado can load data into your own warehouse, including Snowflake or BigQuery. A BigQuery instance managed by Improvado is also available. The number of configured destinations is capped by contract: including every destination type does not provide unlimited connections, and Add a new destination becomes disabled at the contractual cap.

What is the Improvado MCP Only plan?

MCP Only is Improvado’s $100/month plan for connecting your existing AI assistant to marketing data. It includes 2M unique rows/year, daily sync, 300 MCP actions/week, and 1 workspace. Choose it for assistant-based access rather than managed governance: it includes neither AI Agent actions nor governance rules, unlike the managed plans.