Skip to content

Which deals in this quarter's forecast are actually real?

Audit the committed number with evidence instead of opinion. Join Salesforce opportunities and stage history, HubSpot engagement, and Amplitude product usage in Querri, then split the forecast into high, medium, and low confidence tiers with dollar totals your CFO can act on.

Open Querri

What you'll need

Querri (Free trial) to join the three systems, score every open deal, and build the executive view

Salesforce, HubSpot, and Amplitude exports — Step 1 lists exactly what to pull from each

An account ID and your quarterly targets — to line the three systems up and measure against the number

Free resources

Start your free trial here →

Sample data sets

Salesforce Opportunities Salesforce Opportunity Field History Salesforce Opportunity Contact Roles HubSpot Engagement Activity HubSpot Contacts Amplitude Account Weekly Usage Account Product Access Quarterly Sales Targets

Synthetic demo datasets covering 900 opportunities across five quarters, not real customer data. Safe to download and explore. Follow along and you'll get the same numbers you see below.

Need help?

If you have any questions, you can request a demo or email our team.

Before we begin

Every forecast review runs on the same currency: what the sales team believes. Stage, probability, forecast category, and a close date that has quietly moved twice. It's not dishonest, it's just unverified, and by week ten of the quarter nobody can tell the difference between a deal that is closing and a deal that is being talked about.

This playbook builds the other view. One row per open opportunity, enriched with evidence from outside the CRM: whether the buyer is still responding, whether the decision-maker has shown up in the last month, whether the account is actually using the product, and how deals with this exact signal pattern have historically closed. Then it splits the quarter into high, medium, and low confidence with dollar totals, and names the deals driving the gap.

How it works:

  • • Upload Salesforce opportunities with stage and close-date history, HubSpot engagement activity, and Amplitude account usage with a product-access flag
  • • Clean the pipeline: flag stale close dates, missing next steps, repeated pushes, and stage-time outliers
  • • Score each open deal on buyer engagement and product usage, then test every signal against your own closed deals before you trust it
  • • Split the forecast into high, medium, and low confidence tiers with dollar totals
  • • Produce the executive view: the committed number, the at-risk number, the upside, and the specific deals driving the gap

Follow the steps

Open Querri →
1 Step 1:

Upload Salesforce, HubSpot, and Amplitude

Bring all three systems into one project. Each one answers a different question, so bring the pieces that let it do that:

  • •

    Salesforce — open and closed opportunities (the closed ones are your evidence base), opportunity field history so you can reconstruct stage and close-date changes, and contact roles so you know who is actually on the deal.

  • •

    HubSpot — engagement activity at the event level. One row per email, call, meeting, or form, each with a direction and an outcome.

  • •

    Amplitude — account-level weekly usage, plus a product-access table that says whether usage applies to that account at all.

Add your quarterly revenue targets and that's everything. Querri profiles every file and works out how they relate.

Tip: Include at least four quarters of closed opportunities. You cannot call a deal high confidence until you can show which signal combinations actually closed, and that requires closed-won and closed-lost history, not just the open pipeline. This sample spans five quarters and 900 opportunities.

2 Step 2:

Clean the pipeline before you score it

Start with the hygiene problems that inflate every forecast: close dates that have already passed, deals with no documented next step, close dates that have been pushed more than once, and deals sitting in a stage far longer than deals that actually won. Ask Querri to reconstruct each deal's history and flag them:

Prompt

"For every open opportunity closing in 2026-Q3, use the opportunity field history to calculate days in the current stage, the number of times the close date has been pushed out, and total days pushed. Flag close dates that have already passed, missing next steps, and any deal whose days in stage exceed the 75th percentile for that stage among closed-won deals. Show deal counts and dollar amounts per flag."

The quarter's open pipeline is 140 deals worth $12,473,000, and 130 of them carry at least one flag. Only 10 deals, $644,500, come through clean. The single biggest problem isn't the close date, it's the buying committee: 122 deals worth $10,598,000 have had no decision-maker activity in 30 days.

Clean the pipeline before you score it
3 Step 3:

Score the evidence, then test it against your own closed deals

Now score engagement and usage per deal, and separate buyer engagement from sales effort. Ten unanswered outbound emails are not evidence. A reply, a held meeting, a connected call, or a form submission is. Then prove the signals work by scoring your closed deals as they looked 14 days before they closed:

Prompt

"For each opportunity, count buyer-initiated engagement only: inbound email replies, meetings held, connected calls, and form submissions. Calculate days since the last buyer response, meetings held in the last 30 days, unique engaged contacts, and whether an economic buyer or decision-maker engaged in the last 30 days. Then score every closed opportunity as of its close date minus 14 days and show the win rate for each signal band."

Across 720 closed deals the base win rate is 60.7%. Engagement recency splits it violently: a buyer response inside 7 days is worth 89%, past 30 days it collapses to 5%. Decision-maker involvement is just as stark, 94.6% against 28.6%, and three or more engaged contacts wins 99% of the time where single-threaded deals win 19.5%. The CRM's own category was informative too: every deal still sitting in Best Case 14 days out lost, all 118 of them.

Score the evidence, then test it against your own closed deals
4 Step 4:

Split the forecast into confidence tiers

Now tier the open pipeline on the signals that survived the test: engagement recency, decision-maker involvement, multi-threading, close-date stability, and stage duration against the won-deal norm. High confidence requires recent buyer contact, an active decision-maker, at most one push, and a stage duration inside the norm. Low confidence is where the evidence is missing, not merely thin:

Prompt

"Score every open 2026-Q3 opportunity on engagement recency, decision-maker activity, engaged contacts, close-date pushes, and stage duration versus the won-deal benchmark. Assign High, Medium, or Low confidence, then show deal counts and dollar totals per tier, cross-tabbed against the CRM forecast category, and attach the historical win rate for each tier."

$3,420,000 of the open pipeline, 27% of it, sits in the bottom tier, where comparable historical deals won 14.3% of the time. And the tiers cut across the CRM categories rather than tracking them: $432,500 of Commit scores Low, while $3,055,000 of high-confidence business sits outside Commit entirely.

Split the forecast into confidence tiers
5 Step 5:

Produce the executive view

The last step turns the tiers into the four numbers an executive team actually needs, then names the deals driving the gap with the specific missing evidence next to each one, so the meeting is about actions rather than adjectives.

Booked

$3,148,500

41 deals won, 105% of target

Committed

$2,905,500

43 open deals on top of booked

At risk

$1,895,500

24 of those deals, no evidence

Upside

$3,055,000

31 strong deals not in Commit

The narrative Querri writes:

The quarter's target is already covered by booked business. The exposure is in what has been promised on top of it.

  • • Only 58% of that committed number ($1,689,000 across 24 deals) is backed by recent buyer engagement, an active decision-maker and a stable close date.
  • • The at-risk block sits in the tier where comparable deals historically won 14.3% of the time, and $3,055,000 of high-confidence business isn't in Commit at all. The forecast is wrong in both directions.
  • • The fastest fix isn't a pricing approval. It's a meeting with an economic buyer on those 24 deals.

What you can create or export:

  • A one-page executive forecast view: booked, committed, at risk, upside, and the named gap drivers
  • The deal-level scoring table with every signal and flag, exportable to Excel, CSV, or Google Sheets
  • A live dashboard your revenue team can open on Monday, refreshed on a schedule
  • A narrative-driven presentation for the forecast call or board meeting, exported to PDF or PowerPoint
  • A saved project you can re-run weekly, so tier movement week over week becomes its own early-warning signal
Produce the executive view

Tips for a forecast review people actually trust

Separate buyer engagement from sales effort

Ten unanswered outbound emails make a deal look active and change nothing about whether it will close. Count only buyer-initiated evidence: inbound replies, meetings held, connected calls, high-intent form submissions. In this dataset that one distinction is the difference between an 89% win rate and a 5% one.

Validate every signal before it moves a dollar

Score your closed deals as they looked two weeks before they closed, then check which signal bands actually separated won from lost. Here engagement recency and decision-maker involvement were decisive and product usage was not. Had we scored usage on instinct, the tiers would have been confidently wrong.

Report the composition, not just the total

Weighted two ways, this quarter's open pipeline comes out a million dollars apart, and the tiers cut across the CRM's categories in both directions: Commit deals with no supporting evidence, high-confidence deals nobody has committed. A total that happens to look reasonable, built from deals that are misfiled, will still cost you the quarter. Show the tiers, not the sum.

Run it the same way every week

Consistency beats sophistication. Save the analysis as a project in Querri and schedule it, so the tier a deal was in last week is comparable to this week. A deal sliding from High to Low is often a clearer warning than any single flag, and it arrives while you can still do something about it.

Have a Data Dictionary? Bring it. Don't have one? Build a first draft in Querri.

A Data Dictionary documents what the fields in your data mean, how important business terms are defined, and how different systems relate to one another. Forecast analysis is unusually dependent on it, because almost every term is contested: what counts as buyer engagement, which roles qualify as decision-makers, when a close-date change counts as a push, and what "Commit" is supposed to mean in your company. The sample datasets below use field names close to the Salesforce, HubSpot, and Amplitude defaults, so Querri can infer a lot on its own, but the definitions that decide a forecast are yours to state.

Show me how to bring or build one

If you already have a Data Dictionary

Upload it alongside the data and tell Querri how you want it used. Querri can work across multiple Sources in one project, investigate their structure and contents, identify data types and relationships, and join Sources when needed.

Prompt

"Use our Data Dictionary as the reference for interpreting the Salesforce, HubSpot, and Amplitude data in this project. Use its definitions for stages, forecast categories, contact roles, and engagement outcomes. If the data appears to conflict with the dictionary, flag the conflict rather than making an assumption."

If you don't have a Data Dictionary

Use Querri to create a useful first draft. Upload the Sources you want documented, then ask Querri to inspect them and build a table with everything a good dictionary needs:

  • Source name
  • Field or column name
  • Inferred data type
  • Example values
  • Likely business meaning
  • Possible relationships to fields in other Sources
  • Questions or ambiguities that need human clarification
Prompt

"Review the Sources in this project and create a first-pass Data Dictionary. For every field, show the Source, field name, inferred data type, example values, likely business meaning, and any fields that appear useful for joining Sources. Flag definitions you cannot determine confidently instead of guessing."

Then add the business context only your team knows

Querri can infer a great deal from the data itself, but the definitions that decide a forecast come from your team. Querri can see that an activity has an outcome of "Replied" and a direction of "Inbound"; only you can say that this is what your company counts as buyer engagement. Rules like these can't be learned from the data alone:

  • • Buyer engagement means inbound replies, meetings held, connected calls, and form submissions, not email opens or sequence enrollments
  • • Economic Buyer, Decision Maker, and Executive Sponsor are the roles that count as decision-maker involvement
  • • A close-date change only counts as a push when the new date is later than the old one
  • • Product usage applies to customers, trials, and pilots, and never to accounts with no product access
  • • Stage-duration outliers are measured against the 75th percentile for closed-won deals, not against an average

Save your standard definitions once

When a definition is a company standard, save it once instead of restating it in every chat. Add organization-wide rules in Context Settings, and Querri applies them automatically to every analysis your team runs. For the bigger picture, the Querri Library is where your company's data context lives: it learns your business and keeps your sources, definitions, and business rules consistent everywhere. Once your dictionary looks right, you can export it to CSV or Excel for use outside Querri.

A good Data Dictionary gets better with human review

Think of Querri's generated dictionary as a starting point, not the final authority. Querri does the tedious part: profiling fields and proposing definitions. Your team confirms the business meaning, corrects the ambiguous ones, and adds the rules that can't be learned from the data. In a forecast review that shared reference is worth as much as the analysis, because it moves the argument off what the numbers mean and onto what to do about them.

Frequently asked questions

What does "forecast confidence" actually mean?
Forecast confidence is the amount of independent evidence supporting a deal, separate from what the sales team believes about it. Salesforce tells you what the team is forecasting. HubSpot tells you whether the buyer is still engaging. Amplitude tells you whether the account is getting product value. A deal is high confidence when those sources agree and the deal's own history is clean. It is low confidence when the CRM says one thing and every other source says nothing.
Why not just use the CRM probability field?
Because probability is usually a function of stage, and stage is a claim rather than an observation. Weight this quarter's open pipeline by CRM probability and you get $6,759,240. Weight it by what comparable deals in each evidence tier historically did and you get $7,806,756. Look inside those two totals and the disagreement is stark: the CRM carries $1,792,035 of expectation on the bottom tier, where the evidence says $489,060, and only $2,577,045 on the top tier, where the evidence says $4,663,352. It is optimistic about the deals that historically lose and pessimistic about the ones that win. The totals are one argument. The composition is the one you act on.
Which signals actually predicted wins?
Tested against 720 closed deals with every signal frozen 14 days before close, buyer engagement dominated. Deals whose buyer had responded within 7 days won 89% of the time; deals with no response in 31+ days won 5%. Decision-maker activity in the last 30 days split outcomes 94.6% to 28.6%. Three or more engaged contacts won 99%, single-threaded deals 19.5%. Close-date pushes and stage timeouts mattered less but still moved the number: never pushed 63%, pushed twice 37%.
Should product usage lower a deal's confidence score?
Only where the account can actually use the product, and only after you have tested it. In this dataset usage trend pointed the wrong way: accounts whose usage was flat or declining won 65.7% of the time against 58.4% for accounts whose usage was growing, and the correlation between recent activity and winning was r = -0.035, p = 0.35. Statistically nothing. So usage stays in the analysis as context and as a renewal risk flag, but it does not move the tier. That is the real lesson: validate a signal against your own closed deals before you let it change the number you give your CEO.
How do you avoid look-ahead bias when you test the scoring rules?
Freeze every signal at a fixed point before the outcome. Here each closed deal is scored as of its close date minus 14 days, using only the stage, forecast category, amount, activity, and usage that existed on that date, reconstructed from Salesforce field history. Without that discipline you end up measuring the closing motion (contracts signed, meetings held in the final week) and your model looks far more accurate than it would have been in real life.
What if the three systems don't share an account ID?
Join in priority order: a shared system ID first, then a CRM ID stored as a property in the other system, then website or email domain, then a normalized company name, then a manually maintained exception table. Never match on company name alone. Always report your match rate, and treat unmatched accounts as unknown rather than as evidence of no engagement or no usage.
Do I need historical forecast snapshots to do this?
You can build the whole thing from opportunity field history, which is what this playbook does. Weekly snapshots make it better: they let you see how much the forecast moved this week, which deals were added late in the quarter, which ones oscillate between Best Case and Commit, and which reps consistently over-forecast. If you don't have snapshots yet, start writing a weekly copy of your open pipeline to a file now. In three months you'll have the dataset.
How often should this run?
Weekly, and the same way every week. The first build is the work: connecting the sources, defining what counts as buyer engagement, setting the stage-duration benchmarks. After that, save it as a project in Querri and schedule it. The value compounds because the tier changes week over week are themselves a signal, and because a number nobody argues about on Monday is worth more than a better number nobody trusts.
What do I do with the low-confidence deals?
Not necessarily remove them. Low confidence means the evidence isn't there yet, so the action is usually to go get it: a meeting with the economic buyer, a second stakeholder, a real next step with a date. Then re-score. Deals that stay in the bottom tier for two or three consecutive weeks with no new evidence are the ones to take out of the committed number, and the exec conversation is much easier when you can name the specific missing evidence rather than arguing about a gut feel.

Other popular resources

Playbook

Weekly pipeline coverage

Build a coverage report that strips out false signal.

Playbook

Weekly RevOps exec report

Replace the Monday morning CRM scramble.

Demo

Request a walkthrough

See Querri in action with your own data.

Pricing

Compare plans

Find the right plan for individuals, teams, or enterprise.