Stay ahead of the curve
Get exclusive insights delivered directly to your inbox. Join thousands of tech leaders.
TL;DR
- Invite vendors by job: two per job your decision needs, six to eight in total. Technology intelligence covers five jobs: patent search, landscapes and portfolios, scouting, market and company intelligence, and forecasting which technology wins.
- Most longlists stop at the first four, which describe what exists. Forecasting is the only category that ranks the technologies themselves.
- Test every finalist on your own case: a shift your industry already lived through, using only data public at the time, plus one live decision. Same brief, same deadline.
- Make security and data rights pass-or-fail gates, not weighted scores, and price every bid with your team's hours included.
Invite vendors by the job you need done: two per job, six to eight in total. Technology intelligence software does five jobs: patent search, landscapes and portfolios, scouting, market and company intelligence, and forecasting which technology wins. Most RFP longlists cover only the first four, so the RFP never tests the question an R&D budget turns on. Give every finalist a technology shift your industry already lived through, allow only data public at the time, and see who calls it. GetFocus publishes seven of these backtests. In the Kodak case, the rate separated digital from film in 1983, 17 years before film sales fell away.
Which vendors should we invite to a technology intelligence RFP?
Start from the decision, not the category label. "Technology intelligence" covers tools with little in common, and an RFP that mixes them ends up scoring a patent search engine against a startup database on one sheet. Group the market by job, pick the jobs your decision needs, and invite two vendors for each. For an R&D team choosing which technologies to fund, that is usually three or four jobs.
| Job | The question it answers | Examples to invite | What to test |
|---|---|---|---|
| Patent search and prior art | Does this exist? Is our idea new? | Derwent Patent Search, PatSnap Eureka | Recall on a case you already know |
| Landscapes and portfolios | Who owns what, and how strong is it? | LexisNexis PatentSight+, Questel Orbit Intelligence | Ownership and family counts in your field |
| Scouting | Who could we partner with, license from or buy? | Wellspring Scout, ITONICS | Partners your team did not already know |
| Market and company intelligence | Who is funding, buying and talking about what? | AlphaSense, CB Insights | Company and deal coverage in your sector |
| Forecasting which technology wins | Which route improves fastest, and when does it overtake? | GetFocus | A call on a shift that already happened |
The examples come from our October 2026 comparison of 18 R&D intelligence tools, which covers the full field, with pricing.
Write the scope into the first line of the RFP. Vendors such as Flexera and BuiltWith also sell "technology intelligence", meaning your own IT estate or the tech stacks behind other companies' websites. That is a different market, and bids from it waste a round.
In chemicals and materials, the forecasting slot also draws analyst houses and sector consultancies. Invite one if you want expert judgement in the mix, and hold it to the same test. You can see a forecasting output before writing a line of the RFP: our analysis of 14 routes to low-carbon ethylene is public.
This applies to R&D and innovation teams buying for technology decisions. If the buyer is the IP department running prior-art and freedom-to-operate searches, run a search-tool RFP and skip the forecasting test.
Why do most RFP longlists leave out forecasting?
Because most of the market is built to describe what exists. Search, landscape, scouting and market tools report activity: filing growth, citations to single patents, funding rounds, news. Activity shows where attention is going. It is a weak guide to which technology will perform. In a study of 28 technology domains, MIT researchers found that the number of patents in a domain did not significantly predict how fast it improved, while citation dynamics did (Benson and Magee, 2015).
So an RFP that scores documents indexed and dashboards delivered rewards the wrong thing for an R&D decision. GetFocus is built for the missing job. It measures each competing route's Technology Improvement Rate from the patent citation network, sets it next to maturity (TRL), and gives every route a verdict: fund, monitor, phase out or ignore. The method is public. GetFocus is not a prior-art or freedom-to-operate tool.
How do you evaluate technology intelligence vendors before signing?
Fix the weights before the first demo, and keep two criteria out of the weighting altogether. Security and exit terms are gates: a vendor that fails either is out, whatever its total.
| Criterion | Weight | What good looks like |
|---|---|---|
| Answers your decision | 25% | A ranked call with a verdict per technology, not a list of results |
| Evidence you can open | 20% | Every claim traces to the documents behind it; the method in writing |
| Coverage and freshness | 15% | Your field, languages and regions; new documents in weeks, not quarters |
| Monitoring | 15% | Alerts on thresholds you set; updates that report what changed |
| Usable by R&D managers | 15% | Plain-language questions, no Boolean strings or classification codes |
| Total cost, team time included | 10% | Licence, set-up, and the expert hours spent running and checking it |
| Security and data rights | Gate | Clear processing terms, no training on your questions, you own the outputs |
| Exit terms | Gate | Exports in usable formats, a renewal notice window, no lock-in |
Ask for the method in writing, not a tour of results. If a vendor uses generative AI to summarise, check that every claim links to its source. GetFocus's forecast is an explainable regression on two citation metrics, cycle time and knowledge flow, validated on 150+ domains at R² ≈ 0.7. Language models read and filter the patents; they do not make the forecast. Then call two reference customers with a use case like yours, one with at least a year on the platform.
How do you test a vendor on your own case?
A demo shows features. A test shows whether the vendor can answer your question. Give two or three finalists the same two tasks, the same brief and two to four weeks, and score the results blind where you can.
1. A shift your industry already lived through. Pick a technology change whose outcome you know, set a cut-off date before it was obvious, and allow only data public before that date. A forecasting vendor should call the winner from that data alone. One that cannot replay the past has no record to show on the future.
2. A live decision. Your real question, in plain language, such as "Which battery chemistry wins for low-cost EVs?" Judge whether the answer is a decision, with routes to fund, monitor, phase out or ignore, and whether your experts can open the evidence behind each call.
GetFocus publishes the first test for itself. The Kodak backtest replays the patent record using only what was public at each date: the improvement rate separated digital imaging from silver-halide film in 1983, 17 years before film sales fell away in 2000. It took two platform searches. The proof library runs the same replay on six more shifts, from hydrogen cars to LED lighting.
Backtest · Imaging · 1980–2026
The test a forecasting vendor should pass
Two searches on public patent data. The rate separated digital from film in 1983, 17 years before film sales fell away.
Improvement rate per year, computed from patents public at each date · GetFocus backtest: film vs digital
The live task costs nothing to run with us. A free 48-hour Decision Brief returns a first-pass radar, an improvement rate for each competing technology, the monitoring triggers we would set and a 30-minute readout with an analyst. Run it next to the other finalists' pilots and compare.
Which tools work for R&D managers without patent-search training?
The ones that take a question in plain language and answer in plain language, with the patents behind every answer. Boolean strings and classification codes are tools for IP professionals. An R&D manager should be able to ask "Is invention leaving the lab, and who controls the parts that matter?" and get a straight answer. Test it in the pilot: hand the live task to an R&D manager, not the IP team. GetFocus takes the question in plain language, with no query syntax, and reports in four decision quadrants.
What should the RFP ask about price and contracts?
Most vendors sell quote-based annual subscriptions, priced per seat, per credit or per technology area. Ask every bidder to price the same scope: the domains you will track for 12 months, the people who need access and the updates you expect. GetFocus prices per domain, with five platform seats, up to 50 technologies, monthly updates and quarterly reviews included.
Put the contract questions in the RFP itself: who owns the analysis, whether your questions or documents train any model, what you can export, and how renewal and exit work. Answers in writing before signature are worth more than promises after it.
FAQ
How many vendors should a technology intelligence RFP include?
Six to eight: two per job your decision needs. Take two or three finalists into a two-to-four-week test on your own case.
Is technology intelligence the same as IT asset or technographic intelligence?
No. In R&D it means external technologies, patents, research and companies. IT asset tools track the software your organisation runs; technographic tools track what other companies run. Name the one you mean in the RFP's first line.
How do you test a technology forecasting vendor?
Give it a shift your industry already lived through, with a cut-off date and only data public before it, and check whether it calls the winner. Then give it a live decision and check that the evidence behind each call can be opened.
What does technology intelligence software cost?
Mostly quote-based annual subscriptions, per seat, per credit or per technology area. Compare quotes on the same scope and add your team's hours to the licence. GetFocus prices per domain, with five seats included.
