Tracking competitor prices is not the same as pricing against them
Tracking produces a number; pricing requires a decision. Three questions sit in between: is it the same product, is the data complete, and does the buyer even perceive the gap.
Most teams set up competitor price tracking and consider the job done. Numbers flow across the screen, the tables are full, everyone can see who is charging what. And yet prices do not change — or when they do, the decision still comes from somebody's instinct.
Seeing a competitor's price and setting your price against it are not the same job. The distance between them is three questions, and most systems skip all three.
Tracking produces a number. Pricing requires a decision. To get from one to the other you first have to know how trustworthy that number is, and then whether it means anything at all.
Question one: is this actually the same product
The most dangerous output of a tracking system is a mismatched price. A wrong price raises no error. It simply persuades you into a wrong decision.
The problem is semantic more than technical. Two listings can share a brand, a model name and an almost identical title while one is the 5-litre tin and the other is 2.5 litres. One is the base line, the other the premium line. The difference may not appear in the title at all.
That is why matching should not be a binary same-or-different call. It needs three outcomes:
| Verdict | Meaning | Use in pricing |
|---|---|---|
| Same | Identical product | Compared directly |
| Variant | Different size, colour or pack of the same product | Held separately — needs unit-price comparison |
| Different | Only looks similar | Never used |
How confident that verdict is should be recorded too. Low-confidence matches should not be accepted automatically; they belong in a review queue. A system that cannot say "I am not sure" will look certain in exactly the places where it isn't.
Question two: never let unmeasured look like good news
This is the most damaging mistake in practice, and it is almost never noticed.
When you classify a product's competitive position you usually end up with boxes like cheapest, competitive, higher. So which box does a product with no competitor listings attached fall into?
The wrong answer is "competitive". Yet that is exactly the default behaviour in most systems — when there is nothing to compare against, the product is quietly filed as neutral. The result is that the "competitive products" count on your dashboard includes products that were never measured at all.
What was never measured must not look like a good result. "Unknown" is a separate answer and deserves its own box.
The fix is to keep unknown as a distinct position. That single change does two things at once: the dashboard starts telling the truth, and your coverage gap becomes visible enough to be worked on.
Question three: does the buyer even perceive this difference
Say your price is 0.8% above the cheapest competitor. Does that make you expensive?
This is where pricing meets psychophysics. Weber's law holds that a difference between two stimuli is only perceived once it exceeds a constant proportion of the baseline:
Applied to price, the consequence is blunt: a price difference below the perception threshold does not exist for the buyer. The standard review of this line of work in the marketing literature is Monroe (1973), and it remains the starting point for research on price thresholds.
The practical implication: the narrower you make your "competitive" band, the more alerts your system generates about differences no buyer would notice. A 1% band is probably narrower than the perception threshold — meaning it labels your products "expensive" over a gap nobody cares about.
The right width varies by category; paint and phones cannot share a constant. The cheap way to calibrate is to sweep the band from 1% to 5% and count how many products land in each box at each step. If most of your catalogue is bunched in that range, the band choice is determining the entire result — and that choice had better have a reason behind it.
If the data is incomplete, the price should not move
Price data degrades silently. When a source changes its page structure the system does not fail — it simply sees fewer products. No failure means no log entry, and prices quietly drift into being wrong.
Detecting this is surprisingly simple: compare the number of products a given update returns against the median of the last few successful updates from the same source. An update that falls well below that median is suspect.
Using the median is not incidental. The median has a breakdown point of 50%, the mean has 0%. A single unusual day drags the mean wherever it likes — and a single unusual day is precisely what you are trying to catch.
The real decision is here: suspect data should not raise a warning, it should stop the pipeline. On a suspect update, prices are not refreshed and recommendations are not recalculated. A record is written and somebody is told.
The difference is subtle and the consequence is not. A system that only warns still publishes the wrong price and relies on nobody ignoring the warning. A system that stops the pipeline makes sure the wrong price is never formed.
The path from tracking to a price
The gates a competitor price has to pass before it can become your pricing decision:
- Observation — the price is recorded with a timestamp.
- Integrity — is this update complete, or suspect.
- Matching — is it the same product, and with what confidence.
- Position — is there a contract violation, is it measurable, where do we stand.
- Significance — is the gap large enough for a buyer to notice.
- Decision — and only here, a price recommendation.
Most tracking tools stop at step one and leave the other five to the user. A full screen is not the same as a decision made.
Summary
- Tracking produces numbers; pricing requires decisions. They are different jobs.
- Matching needs three outcomes and a recorded confidence.
- An unmeasured product must not count as "competitive".
- A gap below the perception threshold does not exist for the buyer; pick your band for a reason.
- Suspect data should stop the pipeline, not just raise a warning.
References
- Monroe, K. B. (1973). Buyers' Subjective Perceptions of Price. Journal of Marketing Research, 10(1), 70–80. The standard review of price thresholds and the acceptable price range; applies Weber's law to price perception.
Frequently asked questions
What is the difference between competitor price tracking and competitive pricing?
Tracking observes and records competitor prices — its output is a number. Pricing turns that number into a decision, which needs three further things: knowing the match is correct, verifying the data is complete, and confirming the gap is large enough for a buyer to perceive.
How should a product with no competitor listings be classified?
It should sit in a separate 'unknown' position. Defaulting these products to 'competitive' makes never-measured products look like good results on the dashboard and hides the coverage gap.
Why should product matching not be a binary decision?
Because two listings can share a brand and model while differing in size, pack or product line. Matching needs three outcomes — same, variant, different — with a recorded confidence for each; low-confidence matches belong in a review queue rather than being accepted automatically.
How do you detect that price data has degraded?
Compare the number of products each update returns against the median of the last few successful updates from the same source. An update falling well below that median is suspect. The median is used because its breakdown point is 50%; a mean is moved by a single unusual day.