Picture the forecast meeting at a large apparel or golf brand.
Forty reps each have a number. Operations has the booked orders. Finance has the shipments. Your specialty retailers, the pro shops and independents in your network, have sell-through you can't see.
Everyone brought data. None of it matches.
The usual response is to blame the model and shop for a better one. But the model is rarely where it breaks. It breaks upstream, in the inputs. For a brand with a big sales team and a specialty retail network, those inputs are mostly people and small accounts, and they split in three places. This post covers each split and what to do about it.
One note on scope. This post is about forecast inputs. It doesn't describe a forecasting tool, and it doesn't assume you're buying one.
Reps know things the system doesn't. Which buyer is nervous. Which region just got rained out. Which retailer loved the line at market. That's real information. It's also the part of the forecast that's hardest to audit.
The research on judgment is mixed, and that's the useful part. A study by Fildes and colleagues analyzed more than 60,000 forecasts from four supply-chain companies and found that judgmental adjustments increased accuracy on average in three of the four companies. But positive adjustments were much less likely to improve accuracy than negative ones, which the authors say suggests a general bias towards optimism. A 2024 follow-up pooled about 147,000 forecasts from six studies. Adjustments improved bias and accuracy for just over half of SKUs, positive adjustments were more likely to make things worse, and debiasing the adjusted forecasts improved performance. These are supply-chain planners, not wholesale reps. Treat the pattern as a warning, not a rate for your team.
Confidence is a separate problem. Gartner's 2020 State of Sales Operations survey found that only 45% of sales leaders and sellers had high confidence in their organization's forecast accuracy. Gartner reported 47% rating their data quality high and 13% rating it poor. The press release doesn't state a sample size, and the survey covers B2B sales organizations broadly, not apparel or golf wholesale. Use it as context.
Then there's the target problem. A CEL Consulting survey, reported by Supply & Demand Chain Executive in 2023, found that about 20% of companies don't distinguish demand forecasts from sales targets. The sample size isn't stated, so treat it as directional. But for a team where reps carry quotas, the blur is easy to picture. A rep's number can mean "what I expect" or "what I'm paid on," and your forecast can't tell which.
At a brand with a big sales team, some orders are entered by reps and some are placed by retailers themselves. Both land in your totals. They aren't the same signal.
Rep-entered orders tend to follow appointments, market weeks and sales pushes. Self-serve orders tend to arrive when a retailer needs something. When the mix between the two shifts, a forecast built on last year's history can read the shift as a change in demand. That's our reasoning, not a published finding. But the shift itself is well documented in RepSpark customers. At Peter Millar, reorder value grew 148% as unassisted retail ordering overtook rep-entered orders. RepSpark's blog doesn't state a time period for that figure. At L*Space, RepSpark let the brand enable specific buyers to process their own pre-books, and sales reps are notified once orders are placed.
Reps without digital tools add a third stream. In RepSpark's 2026 State of Wholesale survey of 26 lifestyle apparel brands, with revenue from $1M to more than $50M, 33% said they have no digital tools for their sales reps at all. That's a small sample that includes brands well below enterprise scale, so it's directional. But when a rep's orders and notes live in email, texts and spreadsheets, the forecast sees them late, or not at all.
Pro shops and independents are your reach. They're also the accounts where you know the least about what happens after the sale. You see what they buy from you. You often don't see what they sell.
Even the market-level picture has gaps. MyGolfSpy, reviewing a Circana golf retail report, noted that Circana's online numbers "do not, in most cases, reflect sales from direct-to-consumer brands". Even market-level data can leave out part of a brand's own business.
Sell-through is still worth chasing. But don't treat it as the cure. A study of a major Australian food company's fast-moving consumer goods supply chain, covering 684 SKU-locations over 112 weeks, found that order-based forecasts beat point-of-sale-based ones by 6% to 15%. That's food manufacturing, not apparel or golf, so don't carry the number over. The lesson that does carry over is that your own order history is a strong baseline. Add sell-through to it as you get it. Don't wait for sell-through before you fix what you already have.
Golf and apparel both run on seasons, and your reps run on regions. That matters, because regional demand can swing apart from the national number.
A 2022 National Golf Foundation analysis noted that the first four months of the year account for about 25% of play and May through August about 45%. Golf Datatech's May 2026 rounds report showed rounds up 5.6% nationally. By region, May ran from +0.2% in the Mountain region to +13.1% in the Mid Atlantic. The report's own weather rules of thumb come with a note that results may vary by region.
Channels also diverge. Golf Datatech's 2018 apparel review, reported by The Golf Wire, found the on-course (green grass) channel down 9% while the off-course channel rose 6%. That's an older data point, and it's golf apparel only. But it makes the same point. A national forecast can hide opposite moves in the two channels.
The practical meaning: your reps hold the earliest regional signal you have. If it lives in their heads, it reaches the forecast late.
You don't need a new model to start. You need clearer inputs.
Write down what the forecast is for. A forecast is what you expect to happen. A target is what you want to happen. If reps carry both in one number, split them.
Commitments made at line showings, market weeks and store visits should be recorded in your ordering platform, not in notes and spreadsheets. Decide what counts as a commitment, who records it and when. The goal is for what reps know to arrive in the same form as what retailers place.
Track whether adjustments improve the forecast or hurt it. Fildes and colleagues call this forecast value added, and their 2024 work found that debiasing adjusted forecasts improved performance. Keep the overrides that help. Retire the ones that don't.
Review rep-entered and retailer-placed orders as separate series before you combine them. When the mix shifts, you'll see it as a mix shift instead of a demand swing.
Start with each territory and roll up. If one region is running far ahead of the others, you'll see it before it disappears into an average.
Oliva and Watson studied forecasting at a consumer-electronics company and described bias that was intentional, driven by misaligned incentives, and unintentional, from informational and procedural blindspots. They argue that an independent group can stabilize the political side of forecasting enough to allow process improvement. It's one company's case, so treat it as an idea to test. The principle is simple. Someone should own how the number is built, even if many people contribute to it.
RepSpark's role here is the input layer, not the model. Reps and retailers place orders in the same system, connected to your ERP, so rep-entered and retailer-placed orders are recorded the same way. RepSpark supports 20+ ERPs and 35+ integrations, and 250+ brands run on the platform.
The Orders page can be filtered by order source, which it describes as where the order originated, so you can look at how each order came in. Flow's Core Reporting Suite, completed in the July 2026 release, includes Orders Overview, Orders Cross-tab, Salesperson Performance, Top Accounts and Top Products. Golf brands that sell to pro shops have a dedicated page, linked below.
To go deeper, read about RepSpark for enterprise brands, what the platform offers sales teams and operations teams, how ERP integrations work, and RepSpark for golf brands. For recent releases, see Features and Updates.
Our earlier post on why forecasting breaks down when wholesale data lives in too many places covers how the demand signal gets corrupted at the system level. For the rep side, see rep reporting and team communication for wholesale.
If you want to see rep-entered and retailer-placed orders side by side in your own setup, schedule a demo.
Because the forecast is only as good as the inputs it's built from. When rep numbers, rep-entered orders, retailer-placed orders and shipments are recorded in different places, in different ways and at different times, the forecast starts from mismatched data. A better model can't fix inputs that disagree.
Research suggests they're useful but need checking. A study of more than 60,000 supply-chain forecasts found that adjustments improved accuracy in three of four companies, but positive adjustments were much less likely to help. A 2024 study of about 147,000 forecasts found adjustments helped for just over half of SKUs. Those studies cover supply-chain planners, not wholesale reps, so score your own team's overrides.
Look at them as separate streams before combining them. Rep-entered orders tend to follow appointments and market weeks, while retailer-placed orders arrive when retailers need product. If the mix shifts from one to the other, separate series show you a change in channel mix instead of a change in demand.
It helps, but you don't have to wait for it. A study of one food company's fast-moving consumer goods supply chain found order-based forecasts beat point-of-sale-based ones by 6% to 15%. That study isn't about apparel or golf, so treat it as a caution against waiting. Build the best baseline you can from your own orders, then add sell-through as it becomes available.
You can see what they buy from you but often not what they sell. Market-level data has gaps too. MyGolfSpy noted that Circana's online numbers generally don't reflect sales from direct-to-consumer brands. Small accounts also mean more individual buyers, each with their own habits, and regional conditions affect each one differently.
It matters, and it varies by region. A 2022 National Golf Foundation analysis noted that the first four months account for about 25% of play and May through August about 45%. Golf Datatech's May 2026 report showed rounds up 5.6% nationally, ranging from +0.2% in the Mountain region to +13.1% in the Mid Atlantic. Regional signals from reps matter because a national number hides them.
A forecast is what you expect to happen. A target is what you want to happen. A CEL Consulting survey found about 20% of companies don't distinguish the two, though the sample size isn't stated. If your reps carry quotas, keep the number they're paid on separate from the number you plan on.
Time one forecast cycle from start to finish. List every place demand data lives, who owns each number, and when it's updated. Then separate the forecast from the target, decide where rep commitments get recorded, and review rep-entered and retailer-placed orders as separate streams.
Sources: