The problem this solves
Pricing pages are where competitors make their strategy legible, and they change more often than anyone tracks. A plan quietly renamed, a seat limit loosened, an enterprise tier that stopped publishing a number - each move says something about what is working for them and what is under pressure. Checking eight pricing pages by hand is an hour of tab-switching that produces notes nobody can compare, because every vendor structures plans, seats, and limits differently.
The absence of a normalized view has a direct cost in deals and in planning. Sales quotes against a mental model of competitor pricing that may be two changes out of date. Pricing committees debate moves without a current map of the market's actual structure. And because nobody records what the pages said last quarter, changes are invisible: you can only notice a delta if you kept the previous state somewhere.
How the mission runs
- Read every pricing page. Web Research visits each of the eight competitor pricing pages and captures the live state: plan names, prices, billing terms, seat rules, usage limits, and the feature gates each tier advertises. Pages that hide numbers behind a sales call are recorded that way explicitly.
- Normalize the structures. The agent maps every vendor's idiosyncratic packaging onto one comparable frame - entry, mid, and top tiers, per-seat versus flat pricing, and what each limit actually meters. Ambiguities are preserved as notes on the cell they affect rather than silently resolved.
- Diff against history. Where the mission has run before, the current capture is compared with the stored previous state and every change is flagged: price moves, renamed plans, added or removed limits, and tiers that appeared or vanished. First runs establish the baseline for the next pass.
- Build the comparison in Sheets. The normalized table lands in Google Sheets: competitors as columns, pricing dimensions as rows, changes highlighted with the prior value noted alongside. A source tab links each page and its capture date so any cell can be checked against reality in one click.
The prompt
This is the exact objective the agent receives. Swap the obvious placeholders for your own domain, segment or channel and run it as-is from the console, Slack, or the API.
What comes back
A Google Sheets workbook with the market's pricing on one screen: eight competitors normalized across plans, prices, seats, and limits, every change since the previous run highlighted with its prior value, and a source tab tying each data point to the page and date it was captured. Run it monthly and the workbook becomes a living pricing history for the whole category.
Make it yours
- Add a positioning row that captures how each competitor frames its pricing - the headline over the table and which plan the page steers visitors toward.
- Focus a run on one dimension across the whole market, such as how everyone meters usage, when a specific packaging decision is on the table.
- Have the agent append a short commentary tab summarizing what the quarter's changes suggest about where each rival is finding or losing traction.
Frequently asked questions
How does it handle pricing hidden behind a contact form?
It records exactly what the page shows: which tiers publish numbers and which say contact sales. That opacity is data in itself, and the sheet tracks it over time - a competitor that stops publishing a price is often repositioning upmarket, and the diff will catch the moment it happens.
How are the historical comparisons sourced?
The mission stores each run's capture as the baseline for the next diff, so comparisons come from your own dated record of what the pages said. That makes every flagged change verifiable: the sheet shows the prior value, the new value, and the dates both were observed.
Can this misread a complicated pricing page?
Normalization involves judgment, so the agent shows its work. Any mapping it was unsure about carries a note in the affected cell, and the source tab links the live page for instant verification. Most teams spot-check the first run and then trust the diffs, which are mechanical.