For many UK small businesses, the data needed to predict next week’s sales already exists. It is simply scattered: contact records sit in a CRM, orders live in an ecommerce platform, invoices are in accounting software, and campaign results are buried in email or advertising dashboards. The result is a familiar management routine: lots of activity, plenty of reports, but no clear answer to the question that matters on Friday afternoon: what is likely to happen to sales and cash next week?
You do not need a costly data warehouse, a long transformation programme or a team of analysts to answer it. You need a lightweight, decision-first dashboard that brings together a few trusted measures, applies transparent assumptions and is reviewed every week. Its job is not to produce a perfect prediction. Its job is to give the owner or sales lead enough warning to act: follow up warmer leads, protect a campaign budget, prompt overdue customers or adjust purchasing.
This approach also fits the direction of UK business data use. The UK Business Data Survey 2026 found an association between AI use and more developed data practices, including collecting, analysing and sharing data. The practical lesson for smaller firms is not “buy AI first”. It is to create reliable, usable business data first, then use automation or AI carefully where it genuinely improves the weekly decision process.
Start with the decision, not the data project
Big Data can sound like a technology category reserved for enterprises. For a small business, it should mean something much simpler: using several operational data sources together to make a better commercial decision.
Begin by writing down the decisions your weekly forecast must support. Keep the list short. For example:
- Do we have enough qualified opportunities to reach next week’s sales target?
- Which channel deserves more or less marketing spend?
- Which open quotes or abandoned baskets need personal follow-up?
- How much cash is likely to arrive, and when?
- Should we reorder stock, book subcontractors or delay a discretionary cost?
These questions determine the dashboard. They prevent a common mistake: trying to connect every system and every field before anyone has agreed what they will do differently on Monday morning.
A good first dashboard can be a protected spreadsheet, a reporting feature in software you already use, or a simple business-intelligence view. The tool matters less than the operating discipline. One owner should be accountable for refreshing it, everyone should agree the definitions, and the meeting should end with named actions.
The four data sources that create a useful forecast
The goal is to link the customer journey from interest to cash. Most small firms can make a credible first version from four sources.
1. CRM: the future pipeline
Your CRM provides the leading indicators: new leads, source, sales stage, expected value, next action and expected close date. This is where a forecast begins, but CRM data is only useful when stages mean the same thing to everyone.
Define no more than five or six stages. A service business might use New enquiry, Contacted, Qualified, Proposal sent, Verbal yes and Won/Lost. An ecommerce-led firm may instead record Newsletter subscriber, First purchase, Repeat customer and Lapsed customer. Every stage should have an observable entry rule. “Qualified”, for instance, might mean the buyer has a defined need, budget range and decision timing.
Require a next action and next-action date for every material opportunity. A pipeline with no next action is not a forecast; it is a list of hopes.
2. Ecommerce or order data: what buyers actually did
Order data tells you how leads and customer intent turned into revenue. Pull weekly orders, revenue excluding VAT if that is how you manage sales, average order value, cancellations, refunds, product or service category, new versus returning customers and order date.
For a business with an online shop, add checkout starts and abandoned baskets if those measures are available. For a trade or professional-services firm, use accepted quotes, booked work, deposits and completed jobs instead. The principle is the same: capture the operational milestones that happen before revenue is fully recognised.
3. Invoicing and bank-facing data: when revenue becomes cash
Sales and cash are related, but they are not interchangeable. An invoice raised this week may be paid in 7, 30 or 60 days; an upfront deposit may arrive before work starts. Your cash forecast therefore needs invoice issue date, due date, invoice amount, payment status, actual payment date, credit notes and recurring payment dates.
Group unpaid invoices by due date: overdue, due this week, due next week and due later. Then apply the payment behaviour you have observed for each customer group. A dependable repeat client that usually pays within seven days should not be forecast the same way as a first-time corporate buyer with a 30-day process.
Include planned outgoings too: payroll, VAT, rent, supplier bills, subscriptions, loan repayments and tax payments. Official government guidance on supplier payment has long made the essential point that amounts due should appear in a cash-flow forecast, rather than becoming surprises when they fall due. A weekly view makes that discipline operational.
4. Campaign data: where demand is coming from
Campaign data helps you distinguish a healthy pipeline from one inflated by poor-quality leads. Use the channel, campaign name, spend, clicks or enquiries, leads, qualified leads, orders and attributable revenue where your tracking supports it.
Do not overstate precision. A buyer may see a social advert, search your brand later and purchase after an email. For a small business, consistent source tagging is usually more valuable than an elaborate attribution model. Choose a practical rule, document it and keep using it long enough to spot trends.
Create one shared weekly dataset
Before building charts, create a small combined table at a sensible level of detail. For many firms, one row per customer, opportunity, order or invoice is enough. Add a unique customer ID wherever possible. If systems do not share an ID, use a carefully cleaned email address, account name or phone number as a temporary matching key, while accepting that matches will need review.
Keep the first field list lean:
- Customer and source: customer ID, customer type, acquisition channel, campaign and region where relevant.
- Pipeline: lead date, opportunity stage, expected close date, expected sale value, owner and next action date.
- Orders: order date, order value, product or service group, new or repeat status, refund or cancellation flag.
- Invoices: invoice date, due date, amount, payment status, payment date and expected collection week.
- Calendar: week commencing, month, season or trading event.
Use a Monday-to-Sunday or Monday-to-Friday reporting week consistently. Write definitions beside the dashboard. For example, “new lead” means a unique prospect created during the week, not every form submission; “conversion” means a paid order or signed agreement; and “repeat order” means an order from a customer who previously bought.
Data quality is part of commercial management, not housekeeping. Remove duplicate contacts, fix clearly wrong dates, record the source of a lead and make sure closed-lost reasons are usable. The Information Commissioner’s Office says organisations should have processes to check data accuracy and record the source of data. Its guidance also stresses data minimisation: hold information that is adequate and relevant for the purpose, but no more than necessary. Read the ICO’s accuracy guidance and data minimisation guidance before widening access or exporting customer data into a new tool.
Forecast four measures that lead to action
A dashboard becomes decision-first when it concentrates on a handful of measures that answer specific questions. Build a baseline from the most recent eight to 13 comparable weeks, then adjust for known events such as bank holidays, school breaks, promotions, price changes, stock constraints and large contracts. Avoid comparing a quiet January week with a Black Friday week just because both are “last week”.
Leads: forecast the top of the funnel
Calculate the average weekly number of leads by channel, then look at the recent trend. If paid search generated 24, 27, 29 and 31 leads in the last four comparable weeks, a sensible short-term baseline may be around 28 to 30, not an optimistic extrapolation from the latest spike.
Show actual leads, forecast leads, target leads and the gap. Break the total down by source only when someone can act on it. A negative gap in high-quality referral leads may justify more partner outreach; a rise in cheap but unqualified social leads may justify changing the campaign rather than celebrating volume.
Conversions: translate opportunity into likely sales
Use historical conversion rates by meaningful segment. The basic calculation is simple: forecast sales equals forecast qualified leads multiplied by the historical qualified-lead-to-sale conversion rate. If you expect 30 qualified leads and your recent rate is 25%, the starting forecast is 7.5 sales. In practice, show seven or eight as a range, not a false point estimate.
For a CRM pipeline, use stage-weighted value. Assign a probability based on your own closed-won history, not generic software defaults. For example, opportunities at Proposal sent may historically close 40% of the time, while Verbal yes closes 75%. Multiply each opportunity value by the relevant probability and total the results expected to close in the week.
Keep an eye on pipeline ageing. An opportunity stuck at Proposal sent for 90 days should not retain the same probability as a fresh proposal. Add an ageing flag so the sales team either progresses, re-dates or closes it.
Repeat orders: make the customer base visible
Repeat demand often stabilises a small business forecast, but it is easy to miss when every order is treated as a separate transaction. Track repeat-order rate, number of active repeat customers, time since last purchase and customers due for a natural reorder.
For a consumable product, calculate typical reorder intervals. If customers who buy a 30-day supply commonly return between days 25 and 40, create a weekly “due to reorder” cohort. For a service business, use renewal dates, maintenance schedules, annual reviews or project completion dates. Forecast repeat orders from customers who are due, multiplied by the observed repurchase rate for similar customers.
This turns retention into an immediate commercial task. A list of 40 customers due to reorder is more useful than a broad instruction to “improve loyalty”. Assign outreach, automate a compliant reminder where appropriate and measure the resulting conversions.
Cash flow: forecast collection, not just revenue
Build the cash forecast from three layers: opening bank balance, expected cash in and committed cash out. Expected cash in should include invoices likely to be paid, deposits already agreed, recurring payments and a cautious share of near-term forecast sales where payment is immediate. Do not count an unpaid proposal as cash simply because it has a high sales probability.
Use payment timing categories. For example, include 95% of invoices from reliably prompt customers due this week, perhaps 60% of invoices from customers that usually pay after reminders, and none of an overdue disputed invoice until the issue is resolved. These are illustrative assumptions; replace them with your own payment history and update them when behaviour changes.
Show a base, downside and upside cash position. The downside case might delay marginal invoices by a week and reduce new sales conversion. The purpose is to reveal the lowest likely cash point, giving you time to chase payment, alter purchasing or speak to your adviser.
A worked weekly forecast example
Imagine a Manchester-based specialist homewares retailer selling online and to trade customers. On Friday, the owner refreshes a dashboard for the week beginning 7 September.
- The CRM shows 42 new leads expected from current campaign activity, including 18 trade enquiries and 24 consumer enquiries.
- Historical qualified-lead conversion is 30% for trade and 12% for consumer leads. After applying expected qualification rates, the business forecasts six trade orders and two consumer orders.
- The weighted value of open trade opportunities expected to close next week is £8,400. Recent ecommerce demand suggests £3,200 in online sales before refunds.
- Its repeat-order cohort contains 55 previous buyers whose usual reorder window falls next week. At a 20% expected repurchase rate and a £46 average order value, that adds about £506 of repeat revenue.
- Invoices due next week total £12,000. Based on payment history, £9,600 is treated as likely cash in. Payroll, stock purchases, rent and other committed payments total £10,850.
The dashboard does not tell the owner to assume that every opportunity will close. It shows where action changes the result: personally follow up the three largest trade proposals, run a replenishment email to the repeat-order cohort, pause a campaign producing weak consumer enquiries and call two customers before their invoices move overdue. That is a forecast doing its job.
Run a 30-minute weekly forecast meeting
Set a fixed time, ideally before the week begins. The dashboard owner refreshes the data and notes exceptions. The commercial lead, operations lead and owner then answer five questions:
- What changed from last week’s forecast, and why?
- Which sales are most likely to land this week?
- Which opportunities need intervention now?
- What cash is at risk, and who owns collection?
- What one marketing, sales or operational action will we test?
Finish with an action list containing an owner and due date. Next week, compare forecast with actual results. Do not use misses to blame people. Use them to refine stage probabilities, campaign assumptions, payment timings and data definitions. Forecast accuracy improves through this feedback loop.
Use AI and automation carefully after the basics work
Once the dashboard is trusted, automation can reduce manual effort. You might automate data refreshes, flag duplicate records, identify invoices approaching due date, summarise changes since last week or draft a manager’s briefing. AI can also help identify patterns worth investigating, such as declining conversion for a campaign or a customer segment whose reorder interval is lengthening.
But keep a human accountable for the forecast and for decisions affecting customers. The government survey reports that AI use is associated with more mature collection and analysis practices; it does not mean that an AI tool can compensate for inconsistent CRM stages, missing invoice dates or weak consent and governance. The survey findings are a reason to improve the foundations, not to outsource judgement.
When using external tools, limit the personal data shared, use access controls, document the purpose and check contractual and data-protection responsibilities. The ICO provides small-organisation guidance that is a useful starting point for owners building these processes.
Conclusion: build the smallest forecast that changes a decision
The most valuable small-business dashboard is rarely the flashiest one. It is the one that is refreshed every week, trusted by the people using it and connected to a clear action: chase that invoice, call that prospect, protect that stock line or change that campaign.
Start with CRM, orders, invoices and campaigns. Standardise the definitions, calculate a modest set of lead, conversion, repeat-order and cash measures, and create base, downside and upside views. After four to six weekly cycles, review what improved and what still causes surprises. Then add automation only where it removes friction without obscuring the underlying numbers.
Your next step: block out 90 minutes this week to list your four data sources, agree five dashboard measures and build a first weekly forecast for the next trading week. A simple forecast you use is worth far more than an ambitious data-platform project that never reaches the sales meeting.
