The price index formula, and how to read price position

The price index formula is your price divided by a reference price, multiplied by 100. At 100 you are exactly at the reference. At 96.4 you are 3.6 percent below it, and at 110.7 you are 10.7 percent above it. The reference is normally the average of in-stock competitor offers for the same matched product. Everything difficult about the metric happens after that: which offers count, how SKUs are weighted, and what an index number does not tell you.

The price index formula

Price index = (your price / reference price) x 100

reference price is one of:
  mean of in-stock competitor offers      most common
  median of in-stock competitor offers    safer with fewer than 5 references
  lowest in-stock competitor offer        aggressive, matches how shoppers compare
  one named competitor                    a rival index, not a market index

Category index = weighted mean of SKU indexes
  weight by revenue    default for pricing decisions
  weight by units      what a typical unit sold experiences
  weight by margin     exposure of profit dollars
  no weighting         only for a like-for-like basket

Single-SKU example. Northline Supply lists a cordless drill kit at $229.00. Three competitors have it in stock at $239.00, $234.00 and $239.50. The reference is (239.00 + 234.00 + 239.50) / 3 = 712.50 / 3 = $237.50. The index is 229.00 / 237.50 x 100 = 96.4, so the drill kit is priced 3.6 percent below the market average.

Mean or median, and why it matters at small reference counts

With three or four references, one outlier controls the mean. Take a pressure washer you list at $379.00 with competitor offers at $389.00, $399.00 and a marketplace seller clearing stock at $199.00. The mean is (389.00 + 399.00 + 199.00) / 3 = 987.00 / 3 = $329.00, which gives an index of 379.00 / 329.00 x 100 = 115.2. You look 15 percent expensive. The median of those three offers is $389.00, which gives 379.00 / 389.00 x 100 = 97.4. You are slightly cheap. Same day, same data, an 18-point difference decided by one liquidation listing.

A full worked example, with the arithmetic shown

Six SKUs from the fictional Northline Supply demo tenant, each with three in-stock competitor offers from Voltbay, Harborline and PrimeDeck. Every index below is the SKU price divided by the market average of those three offers, times 100.

Read the column and the story writes itself: three big items priced below market, three small accessories priced well above it. That is a deliberate and extremely common structure. Now watch what happens when it gets summarised into one number.

Weighting: four answers from the same six SKUs

The unweighted average of those six index values is (96.4 + 99.6 + 97.0 + 110.7 + 117.1 + 113.0) / 6 = 633.8 / 6 = 105.6. The revenue-weighted index is 98.3. The same six products, the same day, the same competitor prices, and the two numbers disagree about whether this business is 5.6 percent expensive or 1.7 percent cheap.

Two things follow. First, an unweighted index is only defensible on a deliberately constructed like-for-like basket, such as twenty comparable items in one category at similar price points. Across a mixed catalog it reports the tail. Second, a mean of ratios is not the same as a ratio of means, which is why the basket ratio of 98.9 sits far from the unweighted 105.6. If someone hands you an index without telling you the weighting, you have been handed a number, not a measurement.

In-stock filtering, and the rules that keep an index honest

A competitor price on a product they cannot ship is not a price. It is a historical artefact displayed on a page. Including out-of-stock offers in a reference average systematically drags the reference toward whatever the cheapest unavailable listing happens to say, which makes you look more expensive than you are and invites a price cut you did not need.

• Exclude out-of-stock offers from the reference entirely. Not down-weight, exclude. An offer you cannot buy has no bearing on a purchase decision.
• Require a minimum reference count. Two in-stock offers minimum, three preferred. Below that, drop the SKU from the index rather than reporting a comparison against one rival dressed as a market.
• Report the basket size on every index. If 220 of 260 SKUs qualified this week and 244 qualified last week, the index moved partly because the basket moved.
• Compute a constant-basket index for period comparison. Restrict to SKUs present in both periods, then report both numbers. The full-basket index tells you where you are; the constant-basket index tells you what changed.
• Cap reference age. A price collected nine days ago does not belong in this morning's average. Reference age should be shorter than the reporting interval, which makes refresh frequency a reporting decision and not only a data decision.
• Decide advertised price or delivered price, then never mix. If a competitor shows $199.00 plus $14.00 shipping and you show $209.00 delivered, an advertised-price index says you are 5.0 percent expensive and a delivered-price index says you are 1.9 percent cheap. Both are computable. Only one can be your standard.

Check the shipping arithmetic above, because it is the kind of thing that gets waved through in a meeting. Advertised: 209.00 / 199.00 = 1.0503, so 105.0. Delivered: 209.00 / 213.00 = 0.9812, so 98.1. That is a 6.9-point swing produced entirely by a definition.

Reading price position mix

An index is a single number, and single numbers hide distributions. Price position mix is the companion metric: the share of your SKUs that are the lowest offer available, matched within a tolerance band, or above the market. Define the comparison explicitly, because position against the lowest competitor offer and position against the market average produce different pictures of the same catalog.

Take the six SKUs above and measure position against the lowest in-stock competitor offer, with matched defined as within 1.0 percent. The drill kit at $229.00 beats the lowest offer of $234.00, so it is lowest. The pressure washer at $379.00 beats $384.00, so it is lowest. The shop vacuum at $149.00 sits above the lowest offer of $145.00 by 2.8 percent, so it is above, not matched. The three accessories are all above. That gives lowest on 2 of 6, matched on 0, above on 4 of 6.

By SKU count that reads badly: lowest on only 33.3 percent of the catalog. Weight it by revenue and the same data reads differently. The two SKUs where you are lowest carry $27,480 + $17,055 = $44,535 of the $61,815 total, which is 72.0 percent of revenue. You are the cheapest option on 72 percent of the money and the most expensive option on a handful of accessories. Both sentences are true, and only one of them should drive a pricing decision.

Eight ways a price index quietly lies

1. 1Unweighted across a mixed catalog. The failure shown above: 105.6 versus 98.3 on identical data. If a report does not state its weighting, treat the number as unverified.
2. 2A basket that changes between periods. Add 40 accessories to monitoring and the index rises even though no price changed. Always report basket size, and always compute a constant-basket comparison alongside.
3. 3Out-of-stock references included. Unavailable listings drag the reference down and make you look expensive against products nobody can buy.
4. 4Bad matches in the reference set. One 6 quart pot matched to your 8 quart pot puts a phantom low price into the average. See the product matching guide for how to score and review matches before they reach a report.
5. 5Advertised price compared against delivered price. Shipping, on-page coupons, membership pricing and bundle inclusions all change what is being compared. Pick one definition and enforce it in the pipeline, not in the analysis.
6. 6Stale references. Averaging today's price against a competitor price collected last week is a comparison across time presented as a comparison across the market. Cap reference age below your reporting interval.
7. 7Averaging averages. Computing sub-category indexes and then averaging those indexes reproduces the unweighted problem one level up. Weight at every level of aggregation, using the same weight.
8. 8A market of one. An index against a single competitor is a rival comparison. It can be useful and it should be labelled as what it is, never as a market index.

Setting a target index band

An index is a diagnostic, not a goal. Nobody should be asked to hit 100. What works is a band per SKU role, reviewed quarterly, with the repricing rules that enforce the band carrying hard floors underneath them.

Size the prize before you chase a band. On the six-SKU example, moving the revenue-weighted index from 98.3 to 99.3 is a price increase of roughly 1.0 percent, because 99.3 / 98.3 = 1.0102. On $61,815 of monthly revenue that is about $628 more revenue in the month, and because unit costs do not change, close to all of it lands as gross margin. That figure holds only if unit volume holds, which is an elasticity question rather than an index question, and the honest version of this calculation always states that condition.

From there, the band becomes the input to a rule rather than a slide. Rules that chase an index need a margin floor, a cost floor, a MAP floor and a maximum daily change underneath them, so that a broken match or an outlier reference can never move a real price beyond a boundary you set. That mechanism is covered on repricing software, and the reporting side lives in pricing analytics. If you want to see position mix on your own catalog before committing to any of it, the price position analyzer runs on a CSV.