Every signal Mama fires comes from one of these places.

How sources are scored.

A "source" here means an upstream data feed Mama pulls from. A "signal" is what we derive from one or more sources. Most signals fuse 2–4 sources to reduce false positives — the ICP rubric weights the resulting confidence into the score.

The three tiers

• Tier 1 · Contracted partner — data we pay for from a vendor with an SLA and a usage contract. Highest confidence weight. Most expensive per record, used sparingly.
• Tier 2 · Public web — scraped or fetched from sources that explicitly allow it (robots.txt + Terms-of-Service compliant). Refreshed on a fixed schedule. Most of our signals fuse 2+ Tier-2 sources.
• Tier 3 · Customer-provided — data you uploaded or that we read from your CRM. Treated as authoritative for your workspace, never shared with others.

Tiers are not quality rankings — Tier 2 sources are often more current than Tier 1 partner feeds, which is why we mix them. The tier label is about provenance and contractual basis, not signal quality.

Firmographic data · ICP fit.

The "is this account my customer?" data — industry, employee count, revenue band, location, business model. Used for the ICP fit dimension (35% weight) of every account score.

Funding events.

Closed rounds, M&A, debt raises, IPOs. Highest-weight signal type in the working-list calculation. Refreshed every 2 hours — a Series B announced at 9am will be in your dashboard before lunch.

Exec moves.

VP+ joins, departures, lateral moves. Hardest source category to get right — LinkedIn changes happen in real time but data quality varies. We fuse 3 sources to confirm before firing the signal.

Hiring spikes.

Job-posting volume changes, role-type changes, geo expansion. The most-noisy signal category — most teams hire all the time. We threshold for "spike" relative to that team's 90-day baseline, not absolute count.

Tech stack changes.

Software added, dropped, swapped on a target's marketing site, app, or job postings. Powers /lookup and the stack-change signal type. We detect ~1,400 tools across CRM, sequencer, MarTech, observability, data warehouse.

Voice mining.

Podcasts, interviews, conference talks, panel discussions, public blog posts from exec-level speakers at target accounts. Lowest false-positive rate of any signal type — execs only say things publicly that they want quoted.

What we never use.

Sources we've evaluated and explicitly decided not to touch — either because the legal basis is weak, the data is unreliable, or it would betray the trust that makes the rest of this site credible.

If a source you'd expect to see is missing from both lists, email us — we either haven't gotten to it, or we've evaluated and rejected it for a reason we should add here.

Refresh cadence at a glance.

How often each source category gets re-pulled. Cadences are deliberately staggered — funding is fastest because the news window is short; tech stack is slowest because it changes slowly and false positives spike on tighter polling.

The account score itself re-runs every 6 hours regardless — even if no new source data lands, the recency decay shifts the score. Customers on Pro can force a re-score outside the cycle via the API.

The three crawler tiers

"Crawlers" gets thrown around loosely. In Mama, there are three distinct tiers, and they do different jobs. Don't conflate them.

Tier 1 — Built-in 24/7 crawlers. Mama's own 10+ crawlers covering the 1,000+ source feeds listed above. Run continuously, hitting every account in every workspace on the cadences in §9. Free on every plan. Out of your control by design — we run them, we tune them, you benefit.

Tier 2 — On-demand crawls. Pro and Company override: re-run our built-in crawlers on one specific account, right now. Returns in ~30 seconds. Useful for consultancies prepping for a client call, teams responding to a same-day event, or anyone who just heard news they want validated against fresh data. Credit-based — 100/mo on Pro, 500/mo on Company. Available via the dashboard and the API.

Tier 3 — Custom bots. The Pro/Company-only tier where you tell Mama what else to watch. Point at any public source we don't cover by default — a competitor's pricing page, a specific subreddit, a niche Substack, a GitHub repo, a public Discord channel, a custom URL with CSS-selector extraction. Custom bots run on the same crawler infrastructure as our built-in ones (same auto-retry, rate-limiting, proxy rotation, failover). Pro = 25 active; Company = 100 active. Hard rules: public sources only (no scraping behind logins), respect robots.txt, auto-pause after 3 consecutive failures. See /changelog for limits + the dashboard for the builder.

Why this matters: built-in covers the obvious sources (funding wires, job boards, G2, Reddit). Custom bots cover the niche-but-high-signal ones — the specific Substack a buyer reads, the Discord channel where deals get sourced. Compounding signal advantage over tools that only resell vendor data.

When sources change.

Sources change. Vendors deprecate APIs. Press wires re-architect. LinkedIn ships a tighter rate limit (see the April 5 post-mortem for what happens when they do). Three commitments about how we handle source-side changes:

• Adds get a /changelog entry tagged sources. New source, new tier, what it's used for, what changed.
• Removals get a 30-day notice in the customer dashboard before the source stops being used. Plus a changelog entry. Nothing silent.
• Material confidence changes get an alert on accounts where that source was a primary signal driver. "We used to rate this Tier 2; now it's Tier 3 because the vendor changed terms" — flagged on every affected brief.

If you ever see a signal where the source link 404s or the cited article has been pulled, that's a bug — tell us and we'll pull the signal from the affected briefs while we fix the source.