Pricing analytics built on each seller's recorded prices

01Price history on every product page

Each product page charts its prices per region over 30 days, 90 days or a year. The main line is the lowest market price per day, worked out like the product's current price: sellers' offers only, the cheapest offer that isn't sold out, and re-checks can only lower it. Next to it run lines for the 5 sellers seen most often, so you can see who moved first and how far.

Ads and search results that name no seller stay off the chart, because they are often a different product. Individual observations are kept for 90 days. After that each seller's day is kept as a low, an average and a high, which is what the longer chart ranges draw.

Above the chart, the product's trend shows how far the current market price sits from its 30-day average in standard deviations (a z-score) and how long ago a seller last changed its price.

02Four views of one product

• Product trend: the market's average price per day, moving averages over the last 7 and 30 days that have data, bands 2 standard deviations either side of the 30-day average, momentum since the start of the window, and the days since the last price change and the last sale.
• Sale analysis: how often the product's sellers are on sale, the average and deepest discount, each seller's sale runs with their median and 90th-percentile length, the average gap between sales, and a projected date for the next one.
• Leader and follower detection: for every pair of sellers, the lag at which their daily prices line up best, searched up to 14 days either way by default. The seller whose moves the other repeats days later is the leader, and each seller gets a count of the pairs it leads.
• Price forecast: a straight-line regression through the market's daily average price (60 days by default), projected 7 days ahead by default and up to 30, with a 95% band and the R² of the fit. It needs at least 3 days of data.

The leader and follower view answers a practical question: if one shop cuts, will the others follow, and after how long? A pair with a strong correlation at a 2-day lag suggests the follower tends to match the leader's moves about 2 days later.

03Seller scorecards across your catalog

The seller scorecard profiles every seller across all your monitored products over 7, 30 or 90 days. For each one it gives the number of your products it sells, its average price, its price volatility, the share of its observations on sale, its average discount, and its undercut rate: the share of product-days on which it was the cheapest seller. First and last seen dates show whether a seller is still active.

A scatter chart plots average price against sale frequency, which separates shops that are cheap every day from shops that are dear but discount often. Through the API the scorecard also filters by category, region or one shop's domain, and takes custom date ranges.

For a single product, the competitor pricing view gives each seller's current, average, lowest and highest price over the last 30 days, with a daily series per seller.

04Market views: categories, regions and timing

• Category price index: the average market price of each category, subcategories included, over a window you choose, with the lowest and highest price, the spread, the share of observations on sale, and the 5 products that rose and the 5 that fell most against a month earlier. It compares the index with the same window a month, a quarter and a year back.
• Regional arbitrage: products whose average price differs by at least 5% between your regions, ranked by the gap, with the cheapest and dearest region named. Regions are compared only within one currency, so California is set against Texas and Germany against France, never dollars against euros.
• Timing patterns: the average price, the share of observations on sale and the average discount by day of the week, day of the month or hour of the day, for one product or the whole catalog.

Regional arbitrage is most useful when you track the same product in several states or countries. Regional price tracking explains how regions and slots work.

05Three more views through the API

Market positioning places the latest market price of each product within the range of prices recorded for it: a percentile and a label, from aggressive low pricing (below the 25th percentile) through competitive low and mid-market to premium (the 75th and above), with the mean, standard deviation and coefficient of variation. The product pricing overview gives each product's latest, lowest, highest and mean market price and whether the trend is up, down or stable. Competitor pricing is the per-seller view described above.

Every analytics view is in the REST API, which is on every plan, and the price history, products and alerts export as CSV. API and integrations covers keys and rate limits.

06What counts as the market

Market statistics (the trend, forecast, category index, regional arbitrage, timing patterns and positioning) use sellers' offers plus search results that name no seller, such as Amazon keyword results. They leave out Google Shopping ads, Amazon Sponsored Products and other placements, because an ad is often for another product. They also leave out the re-check samples, because a day with only a handful of re-read sellers would skew an average.

Seller views (scorecards, sale analysis, leader and follower, competitor pricing) use sellers' offers only, and there the re-checks do count, since they are those sellers' own prices. Every view works in one currency at a time, your most common one unless you pick another, and most price statistics drop sold-out offers unless you ask to include them.

07Data quality, in the open

A data-quality page shows how the numbers were gathered: the mix of sources and how many rows are offers, ads or listings; monitored products whose last check is older than your plan's cadence plus a day; products with no observations in 60 days; product-days where sources disagree by more than 5%; thin coverage by product and region; and how often sources state stock at all. Data sources and data quality describes each source.

08How much history the views need

The analytics read individual observations, which are kept for 90 days, and several views need a fair number of checks before they say anything useful:

• Weekly re-checks on Starter and Plus give each known seller about 13 readings in 90 days. Leader and follower detection needs at least 15 days on which both sellers were seen, so in practice it relies on the daily re-checks on Growth and up.
• Market statistics leave re-checks out, so on standard plans they rest mostly on the monthly searches: about 3 search days in 90. The trend, the forecast (which needs at least 3 days of data) and the category index fill in slowly there.
• The category index compares with the same window a month, a quarter and a year earlier. With 90 days of observations, only the month-on-month change has a full earlier window; the quarter and year comparisons come back empty or thin.
• Timing patterns group prices by when they were recorded and, like other market statistics, leave re-checks out. A monthly search reads each product on about one day a month, so day-of-week and hour-of-day patterns say more with the daily or hourly searches of the Custom plan.
• The forecast is a straight line through recent daily averages. It shows direction; it doesn't know about promotions or seasons.

Analytics describe the market; they don't change prices. On plans that include it, repricing turns rules into suggested prices, and price alerts tell you when a single seller moves. For how other tools report on the same data, the Prisync comparison is a starting point: Prisync's feature page lists Excel reports.

Questions people ask