Platform
Who's it for?
Resources
Compare
A buyer's framework for evaluating pricing intelligence vendors. Published by ShopVision. Free to use and adapt, no attribution required. A sister template for MAP monitoring and enforcement is available separately.
Most pricing intelligence RFPs fail before the first vendor responds. They ask about feature checklists instead of data quality. They let vendors demo on curated sample data instead of the buyer's own catalog. And they treat the decision as a software purchase when it is really a data purchase with software attached.
This template is built around one principle: the data is the product. A pricing intelligence platform with a beautiful interface and bad product matching will cost you more than having no tool at all, because you will make pricing decisions with false confidence.
Use this template three ways:
One rule we'd urge you to keep: require every vendor to run their matching engine on a sample of your actual catalog before you sign anything. Section 14 shows how to structure that.
Pricing intelligence decisions stall when the wrong person owns the wrong question. Assign sections before you send the RFP.
Score every question 0 to 5. Resist the urge to average everything equally; weight the sections that match your use case.
Scoring rules that keep vendors honest:
Decide your weighting before sending, not after responses arrive, and total it in the scorecard at the end.
Why this matters: you are buying a multi-year data relationship, not a one-time tool. Vendor stability, focus, and roadmap direction determine whether the platform you buy is the platform you renew.
Describe your company: founding year, ownership structure, headcount, and headcount specifically dedicated to data operations and data quality.
What percentage of your revenue comes from pricing intelligence specifically, versus other product lines? Is pricing intelligence your core business or a module?
List three customers of similar size and category to us who have been live for more than 12 months, and confirm they are available for reference calls.
What is your gross customer retention rate over the past 24 months? If you will not disclose it, explain why.
Summarize your product roadmap for pricing intelligence for the next 12 months. What are the two largest investments you are making, and why?
Watch for: vendors who lead with logo walls but cannot produce two referenceable customers in your category. Also watch for pricing intelligence as a bolt-on to a platform whose real business is something else; data quality investment follows revenue.
Why this matters: coverage claims are the most inflated numbers in this category. "We track 500 million products" tells you nothing about whether the vendor tracks the 40 competitors and 3,000 SKUs you actually care about. Collection methods also carry legal and ethical exposure that lands on you, the buyer, if a vendor cuts corners.
For the competitor list attached to this RFP, confirm which sites you already track today versus which you would need to add. For sites you would add, what is the lead time and is there an added cost?
Describe how you collect data: your own collection infrastructure, third-party data providers, or a mix. If third parties are involved, name them and describe your quality controls over their output.
How do you handle sites with anti-bot protections? Describe your approach to collection compliance, including how you respect site terms of service and applicable law.
Do you cover marketplace sellers (Amazon 3P, Walmart Marketplace, eBay), and can you distinguish first-party from third-party offers, including Buy Box ownership?
What geographic markets, currencies, and languages do you cover? For zip-code,region-dependent or store level pricing, how do you capture location-specific prices?
What happens when a competitor redesigns their site? What is your average time to restore full data collection after a site structure change, are rectification steps human based or fully automated and how are we notified of data gaps?
Watch for: vague answers on 2.3. A vendor that will not describe its collection compliance posture in writing is transferring risk to you. And press hard on 2.6; every vendor breaks when sites change, the difference is whether recovery takes hours or weeks, and whether you find out from them or from your own broken reports.
Why this matters: this is the single highest-stakes section in this document. Every price comparison, every alert, every report sits on top of the vendor's ability to provide exact (SKU level) or high confidence similarity matching. Weak matching produces confident-looking dashboards full of wrong comparisons. If you evaluate only one thing deeply, evaluate this.
Describe your matching methodology end to end: what signals you use (GTIN/UPC, MPN, title, description, category classifications, images , attributes), how automated matching works, and where humans are involved.
Describe how you support both exact matching (i.e. matching exact SKU’s by model year, color, size etc.) and similarity matching (i.e. competitive products that share highly similar technical attributes such as color, fabric, dimension, style, function).
Describe how you support matching at the product level (i.e. these two products are the same or highly similar) and at the variant level (i.e. XXS, Perl Blue = XXS, Perl Blue)?
What match rate do you typically achieve for catalogs in our category? State it as exact matches, close equivalents, and unmatched, and be specific about how you define each tier.
Will you run your matching engine on a 500-SKU sample of our catalog, against three competitors we name, before contract signature? If not, why not?
How do you handle the hard cases: multipacks and bundles, size and unit-of-measure variations, colorway and style variants, private label equivalents, and refurbished versus new?
Do matches carry a confidence score visible to us? Can we set a confidence threshold below which matches are excluded from reports and alerts?
When we find a wrong or missing match, what is the correction workflow and turnaround time? Do corrections persist, or can they be overwritten by your next matching run?
How are matches re-verified over time as products are added, discontinued, repackaged, or relisted? On what cadence do you re-verify your matches (weekly, monthly, other) ?
Describe your matching technology, i.e. human (manual), embeddings or web agent / LLM.
Watch for: any vendor who resists 3.5. Running matching on your real catalog is the single most predictive test in this entire process, and a vendor confident in their engine will welcome it. Also distrust match rates quoted without tier definitions; "95% match rate" that counts loose equivalents as matches is a different product than 95% exact.
Why this matters: a price that is wrong is worse than a price that is missing, because someone on your team will act on it. Accuracy is measurable, so make vendors commit to numbers and to the methodology behind them.
What price accuracy rate do you commit to contractually, and how is it measured? Describe the audit methodology, sample sizes, and frequency.
How do you capture prices that only appear in-cart, after a configurator or customization workflow, behind a login, or after a coupon is applied? Is the price you report the price a shopper actually pays?
How do you represent shipping costs, taxes (inclusive vs exclusive), member pricing, setup fees, and financing offers in your effective price calculations?
Describe your automated anomaly detection. When your system captures a price that looks wrong (a $9 price on a $900 product), what happens before it reaches our dashboard?
Do we have audit rights? If our team spot-checks 100 prices against live sites, and accuracy falls below your committed rate, what is the remedy?
Show us your data quality reporting. Do we get ongoing visibility into collection success rates, match health, and known gaps, or do we only find out when something breaks?
Do you report both MSRP and sale prices? Do you report offers and promotions and detect temporary (promotion led) sale prices vs permanent price drops?
Watch for: accuracy claims with no methodology attached. "99% accurate" means nothing without knowing what was sampled, when, and against what definition of correct. The strongest vendors publish their QA process; the weakest describe accuracy as a feature.
Why this matters: pricing moves fastest exactly when it matters most. A daily refresh that slips to 48 hours during Black Friday is a tool that fails on the ten days a year that justify its cost. Get freshness commitments in writing, per site tier, including peak periods.
State your standard refresh frequency for each tier of site you track, and the typical latency between a price changing on a competitor's site and that change appearing in our account.
Can we request on-demand refreshes for specific products or competitors? What are the limits and turnaround times?
What are your refresh commitments during peak trading periods (Black Friday through Cyber Monday, key category events)? Have you sustained them in prior years, and can you show it?
How much historical data do we get at signing, and how far back does your archive go for the competitors we care about? Is history included or an add-on?
If refresh frequency degrades below the committed level for more than 48 hours, how are we notified, and what is the contractual remedy?
What alerting do we get when price freshes fail and you are unable to retrieve competitor pricing within your provided SLA’s
Watch for: one blanket refresh number for all sites. Real collection operations tier their sites; a vendor that claims everything refreshes hourly either has a small site universe or is describing their best case as their standard.
Why this matters: raw price feeds are commodity; the analytics layer is where a tool either changes decisions or becomes a spreadsheet export nobody opens. Evaluate this section against the specific decisions your team makes weekly, not against demo charts.
Walk through how a pricing manager would answer this question in your platform, live: "Where am I priced above the market on products where I lost share this month?"
What price position metrics do you support (price index versus a named competitor set, market position percentile, gap-to-lowest), and can we define our own competitor sets per category?
How does the platform handle assortment context: showing not just price gaps but whether the competitor is in stock, and how availability shifts change effective competition?
Can we analyze price history over time, at both the SKU level and the category level, and export any view we can see?
What repricing support exists, if any: rules-based suggestions, elasticity modeling, or integrations that push recommended prices into our systems? Where does the human approve?
Describe how insights reach people who do not log in: scheduled reports, digests, embedded dashboards in BI tools. What does adoption look like at your healthiest accounts?
Watch for: demos that never leave curated sample data. Insist that at least one analytics question in the live demo be answered on data the vendor has not staged. Also probe 6.6 honestly; most pricing tools are used by two people and ignored by twenty, and the vendor's answer tells you whether they know it.
Why this matters: list price is half the story. In promotional categories, the effective price after codes, stacked offers, and member discounts is the real competitive price, and it is much harder to capture. This is where data operations depth separates vendors.
What promotion types can you detect and structure: sitewide sales, category markdowns, coupon codes, bundle offers, loyalty and member pricing, financing promotions?
How do you calculate effective price when promotions stack? Show an example from a real site.
Can we see a competitor's promotional calendar over time: when they run offers, at what depth, in which categories, and how that pattern shifts year over year?
How quickly does a new promotion appear in the platform after it launches on a competitor's site?
Can we set alerts on promotional behavior specifically, such as a named competitor launching a discount deeper than a threshold in a category we define?
How do you report what marketing channels a promotion or offer was promoted through (i.e. on site via homepage / collection page banners, email, sms, push notifications, social media posts & paid media)
Watch for: vendors that detect that a promotion exists but cannot structure it. A banner screenshot is not promotion intelligence; depth, scope, mechanics, and duration in queryable form is.
Why this matters: alert fatigue kills pricing tools. A system that emails 200 price changes a day trains your team to ignore it within a month. The question is not whether alerts exist but whether the platform separates signal from noise, and fits the way your team already works.
How configurable are alerts: by product set, competitor, threshold type (percentage, dollar, position change), and frequency? Show a real customer's alert configuration.
What noise controls exist: digest bundling, materiality thresholds, deduplication when one competitor moves many prices at once?
What channels do alerts reach: email, Slack, Microsoft Teams, mobile? Can different team members subscribe to different slices?
Describe roles and permissions. Can we scope what a merchandiser sees versus an executive versus an external agency partner?
What does onboarding a new user take? If a category manager joins mid-contract, how long until they are productive without vendor help?
Watch for: ask the vendor for their alert click-through or engagement stats on real accounts. Vendors who track whether alerts get acted on have thought about this problem; vendors who count alerts sent have not.
Why this matters: pricing data creates most of its value outside the pricing tool, in your BI stack, your repricer, your planning systems. A platform you cannot get data out of, at the granularity you can see it in, is a rental you will regret. Exit-readiness starts at integration design.
Describe your API: coverage relative to the UI (can everything visible be pulled programmatically?), rate limits, authentication, and documentation. Provide the docs link.
What scheduled export options exist: formats, destinations (SFTP, cloud storage, email), and granularity?
What native BI connectors do you support (Looker, Power BI, Tableau, SnowFlake, Big Query or other direct warehouse shares)?
What ecommerce platform and system integrations exist today: Shopify, marketplaces, ERPs, repricing engines? Distinguish native integrations from "possible via API."
Do you support webhooks or event streams so our systems can react to price changes without polling?
On contract termination, what do we keep? Confirm in writing that we can export our full history, including matches we corrected and configurations we built, in a usable format.
How do you ingest our catalog data into your systems? Do you support direct data feed integration to our PIM / eCommerce platform, CSV import or direct scraping of products from our own website(s)?
Watch for: API access gated behind a higher tier or priced separately at a level that punishes usage. Also get 10.6 answered in the contract, not the RFP response; data hostage-taking at renewal time is a known pattern in this category.
Why this matters: this section did not exist in pricing RFPs three years ago. It matters now for two reasons. First, natural language access changes who in your organization can use competitive data, from two analysts to everyone. Second, your own AI roadmap will eventually want this data as an input, and vendors differ enormously in how ready they are for that.
Describe your AI-assisted analysis capabilities. Can a non-analyst ask a question in plain language ("which competitors dropped prices on running shoes this week?") and get a correct, current answer?
How are AI-generated answers grounded? Does every answer cite the underlying records and prices it drew from, so a user can verify it? What safeguards exist against fabricated or stale answers?
Do you offer agent-ready access to your data, such as an MCP server or equivalent, so our own AI tools and assistants can query it directly? Under what authentication and permission model?
Which AI models power your features, and can we understand or control where our queries and data are processed?
Is our data, or our usage, used to train models that serve other customers? State your policy plainly.
How are AI features priced: included, metered, or a separate SKU? What happens to our cost if our usage of AI features grows tenfold?
Watch for: AI features that are a chat window bolted onto stale exports. The test is 11.2: ask the vendor to have their AI answer a question live, then click through to the underlying data. If you cannot trace the answer to records, the feature is a demo, not a capability. Score unverifiable AI claims at 0, not 1.
Why this matters: pricing intelligence platforms hold your competitive strategy: the competitors you watch, the products you defend, the thresholds you act on. That map of your priorities deserves the same protection as customer data. Your security team should own this section.
Describe your security certifications and audit posture: which certifications you hold today (SOC 2, ISO 27001), which are in progress with dates, and whether recent audit reports or penetration test summaries are available under NDA.
How is our data isolated from other customers, including competitors of ours who may also be your customers? Be specific about the isolation model.
Describe encryption at rest and in transit, secrets management, and access controls within your own organization: who at your company can see our account data, and how is that logged?
What SSO and identity options do you support (SAML, OIDC)? Is SSO included or a paid tier?
List your subprocessors and hosting providers, with data residency locations. How are we notified when the list changes?
Describe your incident response commitments: breach notification timelines, communication process, and your track record of security incidents in the past 36 months.
Watch for: on 12.1, treat certification as disclosure, not a pass/fail gate, especially with newer vendors whose engineering may be strong ahead of their paperwork. What should be non-negotiable is a straight answer. A vendor who is vague about what they hold today versus what is in progress will be vague about worse things later.
Why this matters: time-to-value in this category should be measured in weeks, not quarters. The heavy lift is matching and configuration, and who carries it (you or the vendor) determines whether your team sees value before the skeptics harden.
Provide a week-by-week implementation plan for an account of our size: matching setup, competitor onboarding, alert configuration, user training. Who does each step, us or you?
What is your median time from contract to first business decision made with the data, across recent customers of our size?
Describe ongoing support: channels, response time commitments by severity, and hours coverage. Is a named customer success contact included at our tier?
What does the ongoing cadence look like: business reviews, usage health checks, match quality reviews? How do you flag when our account is underusing what we bought?
What training and enablement exists for new team members after launch, without paid services?
Watch for: implementation plans that put matching setup on your team. The vendor built the matching engine; the vendor should own match quality, at launch and continuously. Buyers who accept matching as their own responsibility inherit the vendor's hardest problem.
Why this matters: the sticker price is rarely where this category gets expensive. Costs hide in the growth levers: added competitors, added SKUs, added users, API calls, AI usage. Model your year-two cost at realistic growth before comparing vendors on year-one price.
State your pricing model precisely: what units drive cost (competitor sites, SKUs tracked, seats, refresh frequency, API volume), and the price at each break point.
Using the scope attached to this RFP, provide year-one pricing, and then year-two pricing assuming we add five competitor sites, 25% more SKUs, and five more users.
What are your contract minimums: term length, minimum commitments, and payment terms? What discount applies to annual prepay or multi-year terms?
What are your renewal terms: is there a cap on annual price increases in writing? What is your average net price change at renewal across your customer base?
What is included versus priced separately? List every add-on SKU: API access, AI features, additional markets, historical data, premium support, professional services.
Are there any additional / one-time fees for implementation or is implementation included in your software subscription fees? Do you have certified implementation partners or is implementation led exclusively by your own team(s)?
Watch for: pricing models where the unit of cost is the unit of your growth. Per-SKU pricing on a growing catalog, or metered API pricing on a team building automation, means your bill rises exactly when the tool is working. Prefer models where success does not trigger repricing conversations.
Why this matters: nothing in the written responses above outweighs a structured pilot on your own data. A vendor's willingness to be measured before you commit is itself a signal. Structure the pilot so it produces a number, not a feeling.
Propose a paid or free pilot structure: duration, scope (SKUs, competitors, users), and cost, with a clear conversion path to a full contract.
Agree to these acceptance criteria, or propose alternatives with reasoning:
Who staffs the pilot from your side, and what does a mid-pilot review look like?
If the pilot fails an acceptance criterion, what happens: do we walk away clean, does the pilot extend, or is there a remediation commitment?
Watch for: vendors who accept the pilot but resist the acceptance criteria. The criteria are the pilot. A pilot without pass/fail conditions is a long demo.
Total the weighted scores after all responses and demos are complete. Set your weights before sending, not after responses arrive. A promotion-heavy retailer should raise the weight on Section 7; a brand that rarely discounts can lower it.
Section 14 is not scored; it is pass/fail, and it happens with your finalist only.
