Research at Grounds is not a separate product. It is the analytical engine behind every engagement. When we research for a client, we draw on:
We read the public filings, disclosures, and market data of comparable companies to build an independent expectation you can defend, the number a client's own figures get measured against.
For work that turns on external prices, we track the commodities, rates, and indices underneath a client's numbers, so an assumption can be tested against the market rather than taken on faith.
Government and regulatory data is largely free, rigorously collected, and mostly unused. We pull the series that matter into a client's analysis and reconcile them against their own numbers.
For clients whose results move with a place, we read the specific markets they serve, not a national average that describes none of them.
When information is scattered across holdings that never shared a system, we tie it together into one read, each position set against the outside world.
When the number a decision needs does not exist yet, we build it to a standard, so a figure created for the question is as defensible as one pulled from a filing.
Every figure we cite has to clear the same bar a firm applies to its own work: traced to its source, current, backed by a real method, and honest about its limits. Here is what that means in practice, and some of what it makes us leave out.
We cite a figure only when we have read it in the original, the filing, the dataset, the paper, not in an article about it.
In practice: A widely-repeated statistic on companies abandoning AI traces, when you follow it back, to a source page that no longer opens; nearly everyone who quotes it is quoting a summary of a summary. We leave it out. A number you cannot open at its source is a number you cannot defend to a client.
Studies, surveys, and prices age at different rates, and a figure past its shelf life is refreshed or dropped rather than quietly carried forward.
In practice: A market-sizing number from three years ago and a wage index from last month are not the same kind of fact. We date each one, check it still says what it said, and treat a price as good for far less time than a peer-reviewed result.
No stated sample and no methodology means no citation, however precisely the number fits the point.
In practice: One of the most cited documents in this field, aimed squarely at firms like the ones we serve, contains no methodology section at all; its figures come from interviews and observation with no sample behind them. We read it for perspective and attach no number to it. The better a weak source fits, the more dangerous it is.
Every figure we do use carries what it does and does not support, and the population it actually describes.
In practice: A survey of 351 companies is not the market, and a study of large public filers is not a read on a private firm. We say which is which, because for this reader the limit is part of the fact, not the fine print under it.
It is the same standard behind the studies on our AI research page , and behind every number we hand a client.
When an engagement needs an independent number, we source it, test it, and state its limits, so what reaches your client holds up.
Schedule a Conversation →