How long should you keep competitor price history, and what can you actually learn from it?

What price history is actually for

Most teams keep competitor price history because their tool happens to store it, then use it once a quarter to draw a line on a slide. The real value is narrower and much more practical. History is what lets you tell the difference between a competitor doing something new and a competitor doing what it always does in September. Without a baseline, every move looks either alarming or invisible depending on who is reading the chart that day, and the meeting turns into a debate about vibes.

The second use is calibration. Alert thresholds are guesses until you know how much a given item normally moves in a typical month. Once you have a year or two behind you, "they cut eight percent" becomes a much more useful sentence: either they cut roughly that much every six weeks and this is routine, or this is the deepest cut on that item since you started watching. History also settles internal arguments cheaply. When someone proposes matching a competitor because they are "always cheaper", a dated record either supports that or quietly ends the discussion.

Keep reading: Competitive Intelligence for Small Brands, How to Read a Competitor's Promo Calendar, Price Positioning: Where You Sit in the Market. See how RivalPricr helps you competitor price and promo intelligence.

How long to keep it, and at what resolution

Retention and resolution are separate decisions, and conflating them is what makes price archives expensive. A workable default is full observation detail for a recent window of a few months, so you can debug a suspicious reading while the evidence is fresh, then a rolled-up daily record for the long term with the day's low, high, and closing price plus an observation count. Promotional events are tiny by comparison and worth keeping indefinitely as discrete records: start date, end date, mechanic, depth, and the items covered.

On duration, two years is the practical minimum because seasonality has to repeat before you can call it seasonality, and three years is where most teams start trusting a pattern enough to plan around it. Store change events rather than every single observation for the long tail, which cuts storage dramatically, but always keep an observation counter alongside it. That counter is what lets you distinguish a price that genuinely did not move from a stretch where your collection was broken, and those two look identical on a chart. Keep the context with the number too: region, seller, currency, promotional or regular, shipping and fees, stock status, source, and capture timestamp. A number without that context will eventually be misread as a price drop when it was really a different seller or a different regional page. Related: Price Positioning: Where You Sit in the Market

The patterns worth mining, and the ones that mislead

Four patterns repay the effort reliably. First, promotional cadence: how often a competitor discounts a category, how deep it typically goes, how long the promotion runs, and how quickly the price reverts afterward. Second, the distribution of move sizes, which is the honest basis for setting alert thresholds instead of picking a round number. Third, the split between anchor items they almost never discount and traffic items they cut aggressively, which tells you where their margin actually lives. Fourth, lead and lag behavior: over enough moves, you can see whether a competitor tends to move first in your category or tends to follow within days of your own changes. Related: Spotting a Price War Before It Starts

The failure mode is over-reading. Small samples, catalog changes, and improvements to your own product matching all create apparent trends that are really artifacts. A useful discipline is to require a pattern to appear at least three times across two or more years, and to have a plausible business reason behind it, before anyone builds a plan on it. Treat any claim about causation with suspicion, since two competitors reacting to the same supplier cost or the same seasonal demand will look like one following the other. Describe what you observe as a typical range rather than a precise figure, and say plainly when the sample is thin.

Keeping the archive honest and usable

Append, never overwrite. When you fix a bad match or change a normalization rule, write a new version rather than silently rewriting the past, and keep the raw capture separate from the normalized value so you can re-derive history when your rules improve. Record collection gaps explicitly as gaps. A flat line in a chart should be visually distinguishable from missing data, otherwise someone will eventually present an outage as evidence of a stable competitor price. Related: Turning Competitor Data Into Strategy

Write the retention rule down and review it once a year: what is kept at full detail, what is rolled up, when, and what is deleted. Include an export step in that policy so a tooling change never costs you the archive, since history is the one asset in price intelligence you cannot buy back after the fact. Strip anything that looks like personal data before it lands in the archive, respect the terms of the sources you collect from, and keep the record internal. The goal is an archive that a new analyst can read in a year and trust without having to ask which parts are real. Related: Building a Market Dashboard That Matters