The short answer
A shopper asks for a kettle under £50. Your assistant finds a good one and quotes £44.99. The shopper clicks through and the retailer says £52. Nothing was technically wrong with the data. It was just a price from last week, shown as if it were today's. This guide covers how fresh price data really needs to be, and how to be honest when it is not.
There is no single right answer
Asking how fresh a price needs to be is like asking how fresh a weather forecast needs to be. It depends what you are using it for. Two things decide it: how fast prices move in the category, and what the shopper is about to do with the number.
A price used to say "this is roughly a £300 dishwasher, not a £900 one" can be older than a price used to say "this is the cheapest place to buy it right now". The first is about the shape of the market. The second is a claim about a specific moment, and it goes stale fast.
Factor one: how volatile is the category?
Some categories move often. Consumer electronics, popular trainers and seasonal items tend to see frequent promotions and short sales. Others move slowly. A replacement kettle filter or a basic bath towel may hold the same price for months. You do not need numbers to see the pattern in your own data: compare snapshots of the same product over time and note how often the price changes.
- Fast-moving: headphones, laptops, TVs, trainers in popular colourways, anything in a big sale period.
- Medium: small kitchen appliances, prams, furniture, mid-range clothing.
- Slow-moving: consumables, spare parts, basic household items.
Treat these as starting buckets, not rules. Your own snapshot history is a better guide than any general list.
Factor two: browsing or deciding?
Think about where the shopper is in their decision. Early on, they are browsing. They want to know the options and the rough price band. Later, they are deciding. They want to know what it costs today and where.
| Shopper is… | What the price is for | How much age is acceptable | How to show it |
|---|---|---|---|
| Browsing options | Setting a rough price band | More tolerant, especially in slow categories | Show the price with its snapshot date |
| Filtering by budget | Including or excluding products | Moderate; items near the budget edge need care | Flag items close to the limit as "check current price" |
| Comparing retailers | Claiming one is cheaper | Low; only compare snapshots from a similar time | Show each retailer's snapshot date side by side |
| About to buy | The amount they will pay | None; the retailer's live price is the only true answer | Send them to the retailer to confirm today's price |
The last row matters most. No stored price is ever the final word on what a shopper pays. Delivery, vouchers, member pricing and stock all change the total. The retailer's own page is the source of truth at the moment of purchase.
Always keep the snapshot date
A price without a date is hard to use responsibly. You cannot tell the shopper how old it is, and your assistant cannot decide whether to trust it. Store every price as a snapshot: the amount, the currency, the retailer, and when it was seen. Keep those four together as one record rather than separate fields that can drift apart.
We cover where a price comes from and how to record when it was checked in product price source and checked time. The short version: keep the date with the price all the way to the shopper's screen, and do not let it drop off in the middle of your system.
How should an assistant word an older price?
Plain wording does most of the work. A shopper does not mind a price being a few days old. They mind being told it is current when it is not.
Recent snapshot
"£44.99 at the retailer when last checked on 14 October. Check the retailer for today's price."
Older snapshot
"Last seen at around £45 in early October. Prices in this category change often, so the current price may differ."
No date available
"Listed price: £44.99. We don't have a check date for this price, so please confirm on the retailer's site."
Avoid words like "now", "today" or "currently" unless you actually checked the price today. Avoid "cheapest" and "best price" unless every price compared was captured at a similar time.
Send the shopper to the retailer for today's price
Every price should come with a link to the retailer's product page. This is how the shopper confirms what they will actually pay. It also protects you: when the stored price and the live price differ, the shopper sees the difference before they commit, not after.
- Store price as a dated snapshotAmount, currency, retailer and snapshot date together.
- Set freshness rules per categoryUse your own snapshot history to see which categories move fastest.
- Match wording to price ageNo "now" or "today" for prices that were not checked today.
- Be careful near budget limitsA product just under budget last week may be just over it today.
- Only compare like with likeCompare retailer prices captured at a similar time.
- Always link to the retailerThe retailer page is the final word on what the shopper pays.
How Simlir handles price
Simlir returns each product's price as a dated snapshot: amount, currency, retailer and snapshot date when available, along with a retailer link. Your assistant can use the date to choose its wording and send the shopper to the retailer to confirm. Simlir returns links; it does not complete checkout. The field details are in the API docs, and the wider set of fields worth keeping is covered in product data schema for AI shopping assistants. If you are auditing data you already hold, the retailer product data quality checklist is a good companion.
