Performance issues

Performance problems in Mammoth almost always trace back to one of three root causes: large unfiltered datasets being queried inefficiently, pipeline steps that process more data than necessary, or over-broad scheduled automations. This guide helps you identify the source and fix it.

Quick diagnostic: What's slow?

Symptom	Most likely cause	Go to
Dashboard takes a long time to load	Source View is too large or unoptimized	Slow dashboard loading
Pipeline times out or runs for a long time	Large dataset + expensive operations	Slow pipelines
Scheduled refresh takes hours	Too many datasets refreshing simultaneously	Automation performance
AI dashboard generation is slow	Wide Views or a complex prompt	AI generation timing
Export to BigQuery or SFTP is slow	Export volume or destination configuration	Slow exports

Slow dashboard loading

Understand what makes dashboards slow

Dashboards load data by querying the underlying View at render time. The dashboard rendering itself is fast—the bottleneck is almost always the View query.

What makes a View query slow:

• The View is connected to a large dataset (millions of rows) without aggregation
• The View applies complex multi-step pipelines that run at query time
• Multiple charts on the dashboard each trigger separate View queries

Step 1: Check view row count

1. Open the source dataset in your project
2. Navigate to the View your dashboard uses
3. Note the row count shown at the bottom of the View

The more rows the View holds, the longer each chart query takes. Views with millions of rows and no aggregation are the most likely to be slow.

Step 2: Aggregate before the view

For dashboards showing KPIs, totals, and summary charts, the View should contain aggregated data, not row-level records.

Pattern: Add a Group & Pivot task at the end of your pipeline that summarizes data to the grain your dashboard needs. If your dashboard shows monthly revenue by region, the View should have one row per month-region combination—not one row per transaction.

Step 3: Pre-filter the view

Add a Conditional Filter task to the pipeline before creating the View, limiting data to the time window or subset relevant to your dashboard.

Example: If your dashboard shows the last 12 months of sales, add a date filter to the pipeline so the View only contains the last 12 months—not all historical records.

Step 4: Limit chart count per dashboard

Each chart on a dashboard triggers a query against the View. A dashboard with 20 charts on a 500,000-row View will load noticeably slower than one with 6 charts on the same View.

Recommendation: Break large dashboards into multiple focused dashboards (e.g., "Revenue Overview" and "Product Performance") rather than building a single all-in-one dashboard. This also improves readability.

Slow transformation pipelines

Identify the slow step

Use Preview Data at this Task on each task in your pipeline to see the row count at that point, and look for:

• Tasks where row counts are unexpectedly high (a filter that should reduce rows but doesn't)
• Tasks that work across the whole dataset (joins and group & pivot on large datasets)
• Math Function columns with complex calculations applied across millions of rows

Fix the slow step

Once you know which step is slow, apply the matching technique from the Performance guide:

• Row counts that stay high: move Conditional Filter tasks as early as possible, ideally right after the data source
• A slow Join: check whether Lookup is enough, and whether the join key has duplicates
• Many unused columns: add a Remove Column task right after the source
• Heavy rebuilds while you edit: turn on Draft mode

See Pipeline optimization for the full guidance on ordering tasks.

Automation and scheduled refresh performance

Understanding automation scheduling

How often an automation can run depends on your plan, and the units you can choose are listed in the schedule editor. If your workflow needs more frequent refreshes than your plan allows, contact support@mammoth.io to discuss options. See Scheduling for the frequency options and plan limits.

Staggering scheduled refreshes

If multiple large datasets are scheduled to refresh simultaneously (e.g., all at midnight), they can compete for resources and each may run slower. Stagger refresh times between large datasets.

To change an automation schedule:

1. Open the Automations tab in the Data Library
2. Find the automation and open it
3. Edit the schedule
4. Save the automation

Automation taking longer than expected

If a scheduled automation is taking significantly longer than previous runs:

1. Check if the source dataset has grown significantly. A much larger dataset takes longer to refresh.
2. Check for downstream pipeline complexity. If the pipeline was edited recently, a new task may be processing the full dataset in a slow way.
3. Check whether other automations are running simultaneously and competing for resources (see staggering above).

Reminder: If an automation fails, open it, read Last run result, and fix the root cause. Then use Run now to re-run it.

AI dashboard generation timing

AI dashboard generation takes a short while, because the AI reads your View's schema and data before building charts.

If generation seems stuck:

1. Refresh the page and try again
2. If it keeps happening, simplify the prompt and reduce the number of columns in the source View. Narrow Views with a few focused columns generate faster than wide Views with many columns

Slow exports

Export to cloud destinations (BigQuery, SFTP)

Export speed depends on:

• Volume being exported—large datasets with millions of rows take longer
• Network throughput between Mammoth and the destination
• Destination write throughput—some cloud warehouse configurations have write limits

To improve export performance:

• Filter the dataset before exporting—only export the rows the destination actually needs
• For incremental data, add a date filter to export only new or changed rows since the last export, rather than full table refreshes
• For SFTP destinations, verify that the SFTP server is not the bottleneck (check server capacity and available disk space)

Large datasets

For guidance by dataset size, see Dataset size guidance in the Performance guide.

Still not resolved?

If you've applied these optimizations and performance is still unacceptable:

1. Check Monitor: the Running tab shows whether a task is Stuck, and Needs Attention lists failed runs.
2. Review Join steps—check whether the join key has duplicates (fan-out multiplies rows significantly).
3. Contact support@mammoth.io with:
   • Your workspace name and dataset name
   • The pipeline ID, so the support team can inspect server-side execution logs
   • Approximate row count and column count
   • The specific operation that's slow (dashboard load, pipeline run, export, automation)
   • How long the operation takes vs. how long you'd expect